diff --git a/.coderabbit.yaml b/.coderabbit.yaml
new file mode 100644
index 00000000..46a4f0b9
--- /dev/null
+++ b/.coderabbit.yaml
@@ -0,0 +1,5 @@
+reviews:
+  auto_review:
+    enabled: true
+    auto_incremental_review: true
+    drafts: false
diff --git a/.gemini/config.yaml b/.gemini/config.yaml
new file mode 100644
index 00000000..3b5e413e
--- /dev/null
+++ b/.gemini/config.yaml
@@ -0,0 +1,23 @@
+have_fun: false
+
+ignore_patterns:
+  - "**/charts/**"
+  - "**/vendor/**"
+  - "**/zz_generated.*.go"
+  - "**/pkg/generated/**"
+  - "**/_out/**"
+  - "**/*.tgz"
+  - "**/dashboards/**/*.json"
+  - "**/*.patch"
+  - "**/*.diff"
+  - "**/images/*.json"
+
+code_review:
+  disable: false
+  comment_severity_threshold: LOW
+  max_review_comments: 50
+  pull_request_opened:
+    help: false
+    summary: true
+    code_review: true
+    include_drafts: false
diff --git a/.gemini/styleguide.md b/.gemini/styleguide.md
new file mode 100644
index 00000000..7185a625
--- /dev/null
+++ b/.gemini/styleguide.md
@@ -0,0 +1,110 @@
+# Cozystack Review Guidelines
+
+## Project Architecture
+
+Cozystack is a Kubernetes-based PaaS built on Helm umbrella charts and FluxCD.
+Packages live in `packages/{core,system,apps,extra,library,tests}/`. The `library/` group holds reusable helper charts; the `tests/` group exists because library charts are not directly testable.
+Each package wraps one or more upstream Helm charts.
+Go code in `cmd/`, `internal/`, `pkg/` implements Kubernetes controllers and API server.
+Static CRDs are generated by `hack/update-codegen.sh` from types under `api/v1alpha1/`, `api/backups/`, and `api/dashboard/`.
+App Kinds (like `Postgres`, `Kafka`) live in `api/apps/v1alpha1/` but are registered dynamically at runtime from `ApplicationDefinition` resources — they are not static CRDs.
+
+## Vendored Code — Critical Rules
+
+**Never suggest editing files inside any `charts/` directory under `packages/`.**
+Those are upstream Helm charts vendored via `make update` (which runs `helm pull`).
+Any direct edit is overwritten on the next update and provides zero value.
+
+If you find an issue that appears to live in vendored chart code:
+
+- For configuration-level changes: suggest overrides in the package root `values.yaml`.
+- For structural changes: suggest a patch file in `packages/<name>/patches/` applied by the Makefile.
+- For source-code changes in images: suggest a patch in `packages/<name>/images/<name>/patches/`.
+- For true upstream bugs: point to the upstream repository and suggest an upstream issue/PR.
+- Do NOT suggest creating `charts/patches/` — patches never live inside `charts/`.
+
+Similarly, never propose edits to:
+
+- `vendor/` — Go dependencies. Changes go through `go get` and `go mod tidy`.
+- `zz_generated.*.go` — regenerated by `make generate`.
+- `pkg/generated/` — auto-generated Kubernetes client code.
+- Image digest values in `values.yaml` — set by CI via `make image`, not by humans.
+- `go.mod` / `go.sum` by hand — use `go get` and `go mod tidy`.
+
+## Commit and PR Requirements
+
+Each commit must follow [Conventional Commits](https://www.conventionalcommits.org/) format: `type(scope): brief description`.
+
+Valid types: `feat`, `fix`, `docs`, `style`, `refactor`, `perf`, `test`, `build`, `ci`, `chore`.
+
+Valid scopes:
+
+- System: `dashboard`, `platform`, `cilium`, `kube-ovn`, `linstor`, `fluxcd`, `cluster-api`
+- Apps: `postgres`, `mariadb`, `redis`, `kafka`, `clickhouse`, `kubernetes`, `virtual-machine`
+- Meta: `api`, `hack`, `tests`, `ci`, `docs`
+- Package-specific: any `<package-name>` matching a directory under `packages/`
+
+Breaking changes: append `!` after type/scope (`feat(api)!: ...`) or add a `BREAKING CHANGE:` footer.
+
+Each commit must have a `Signed-off-by:` trailer (produced by `git commit --signoff`).
+
+PR body must contain a release note block:
+
+````text
+```release-note
+type(scope): human-readable changelog entry
+```
+````
+
+Flag any PR whose commits lack the Conventional Commits format or signoff, or whose body has no release-note block.
+
+## Helm Chart Conventions
+
+Packages follow an umbrella chart pattern:
+
+- `charts/` — vendored upstream, read-only
+- `templates/` — Cozystack-specific extra manifests
+- `values.yaml` — override values for the upstream chart
+- `values.schema.json` — JSON Schema for dashboard UI and input validation
+
+When reviewing `values.schema.json`:
+
+- `make generate` regenerates this file and strips `title`, `description`, and `x-*` custom annotations. Do not suggest adding fields that will be stripped on the next regeneration.
+- Focus on type correctness, `required` fields, `enum` values, and default values.
+- Flag breaking changes: removing a field, changing its type, or narrowing its enum.
+
+## Sensitive Components
+
+**`packages/core/platform/`**: the platform chart that deploys everything. Changes here can require updates to the migration flow — the Helm hook in `packages/core/platform/templates/migration-hook.yaml` and the runner image with scripts in `packages/core/platform/images/migrations/`. Flag any change to this package that lacks a corresponding migration update or an explicit note that backward compatibility is preserved.
+
+**`api/apps/v1alpha1/`**: app CRD Kinds are registered at runtime from `ApplicationDefinition` resources and the matching `values.schema.json`. Suggest changes to the relevant package schema rather than hand-editing generated types.
+
+**RBAC, ServiceAccounts, and SecurityContext**: flag overly broad RBAC (`*` on resources or verbs without justification), missing `securityContext`, containers running as root without an explicit reason, and `hostPath`/`hostNetwork` usage without clear rationale.
+
+## Go Code Standards
+
+- Use controller-runtime patterns for reconcilers.
+- Use structured logging via `logr` — flag `fmt.Print*` and `log.Print*` in controller code.
+- Handle errors explicitly. Discarding meaningful errors with `_` is a bug.
+- Propagate `context.Context` through call chains. Flag `context.Background()` created inside a reconciler or request handler.
+- Prefer `ctrl.Result{RequeueAfter: ...}` over empty requeue for predictable reconciliation loops.
+- Tests live beside the code (`*_test.go`). New behavior without tests is worth flagging.
+
+## What to Review Carefully
+
+- Logic errors, off-by-one bugs, nil dereferences.
+- Missing error handling, especially in reconcilers and API handlers.
+- Helm template correctness: missing `quote`, incorrect indentation, wrong scope in `with`/`range`.
+- Security: permissive RBAC, privileged containers, secrets in environment variables, hardcoded credentials.
+- Missing resource requests/limits on new workloads.
+- Breaking changes in `values.schema.json`: removed fields, tightened types, narrower enums.
+
+## Anti-patterns — Do Not Flag These
+
+- Large JSON files in `dashboards/` — imported from upstream Grafana sources, not hand-written.
+- Files under any `charts/` directory — vendored upstream, left as-is intentionally.
+- Whitespace or formatting in `*.patch` / `*.diff` files — machine-applied, not authored.
+- Missing comments on generated code.
+- `go.sum` changes accompanying `go.mod` changes — expected and correct.
+- Fork relationships for vendored tooling images — intentional (e.g., `cozystack/kilo` fork is expected).
+- Absence of unit tests for vendored chart overrides — covered by E2E tests in `hack/e2e-apps/`.
diff --git a/.github/CODEOWNERS b/.github/CODEOWNERS
index 2045ec5a..2482712d 100644
--- a/.github/CODEOWNERS
+++ b/.github/CODEOWNERS
@@ -1 +1 @@
-* @kvaps @lllamnyp @lexfrei @androndo @IvanHunters @sircthulhu
+* @kvaps @lllamnyp @lexfrei @androndo @IvanHunters @sircthulhu @myasnikovdaniil
diff --git a/.github/ISSUE_TEMPLATE/bug_report.md b/.github/ISSUE_TEMPLATE/bug_report.md
index 1a119259..fefdab82 100644
--- a/.github/ISSUE_TEMPLATE/bug_report.md
+++ b/.github/ISSUE_TEMPLATE/bug_report.md
@@ -1,7 +1,7 @@
 ---
 name: Bug report
 about: Create a report to help us improve
-labels: 'bug'
+labels: 'kind/bug'
 assignees: ''
 
 ---
diff --git a/.github/PULL_REQUEST_TEMPLATE.md b/.github/PULL_REQUEST_TEMPLATE.md
index 9a52457c..044dd6a0 100644
--- a/.github/PULL_REQUEST_TEMPLATE.md
+++ b/.github/PULL_REQUEST_TEMPLATE.md
@@ -1,8 +1,11 @@
 <!-- Thank you for making a contribution! Here are some tips for you:
-- Start the PR title with the [label] of Cozystack component:
-  - For system components: [platform], [system], [linstor], [cilium], [kube-ovn], [dashboard], [cluster-api], etc.
-  - For managed apps: [apps], [tenant], [kubernetes], [postgres], [virtual-machine] etc.
-  - For development and maintenance: [tests], [ci], [docs], [maintenance].
+- Use Conventional Commits for the PR title: `type(scope): description`
+  - Types: feat, fix, docs, style, refactor, perf, test, build, ci, chore
+  - Scopes are not an exhaustive list — pick the most specific scope for the change and extend the list when a genuinely new area appears. Examples:
+    - System components: dashboard, platform, operator, cilium, kube-ovn, linstor, fluxcd, cluster-api
+    - Managed apps: postgres, mariadb, redis, kafka, clickhouse, virtual-machine, kubernetes
+    - Development and maintenance: api, hack, tests, ci, docs, maintenance
+  - Breaking changes: append `!` after type/scope (`feat(api)!: ...`) or add a `BREAKING CHANGE:` footer
 - If it's a work in progress, consider creating this PR as a draft.
 - Don't hesistate to ask for opinion and review in the community chats, even if it's still a draft.
 - Add the label `backport` if it's a bugfix that needs to be backported to a previous version.
@@ -11,14 +14,19 @@
 ## What this PR does
 
 
+### Screenshots
+
+<!-- REQUIRED for UI changes: attach screenshots or screen recordings demonstrating
+the visual impact of your changes. PRs with UI changes without screenshots will not be merged. -->
+
 ### Release note
 
 <!--  Write a release note:
 - Explain what has changed internally and for users.
-- Start with the same [label] as in the PR title
+- Start with the same `type(scope):` prefix as in the PR title
 - Follow the guidelines at https://github.com/kubernetes/community/blob/master/contributors/guide/release-notes.md.
 -->
 
 ```release-note
-[]
+
 ```
\ No newline at end of file
diff --git a/.github/labels.yml b/.github/labels.yml
new file mode 100644
index 00000000..2ce04130
--- /dev/null
+++ b/.github/labels.yml
@@ -0,0 +1,371 @@
+# Cozystack repository labels
+#
+# Label conventions follow the Kubernetes scheme:
+# https://github.com/kubernetes/test-infra/blob/master/label_sync/labels.md
+#
+# Synced into the repository by .github/workflows/labels.yaml
+# (EndBug/label-sync@v2). Edit this file via pull request — UI changes
+# will be overwritten on the next sync.
+#
+# Constraints (enforced by the validate job in labels.yaml):
+#   - description ≤ 100 characters (GitHub REST API limit)
+#   - color is a 6-character hex string (no leading #)
+#   - label names are unique
+#   - aliases do not collide with top-level names
+#
+# Categories:
+#   kind/         issue or PR type
+#   priority/     urgency
+#   triage/       review state
+#   lifecycle/    issue or PR lifecycle
+#   area/         subsystem; extensible — add when 3+ open issues exist
+#   do-not-merge/ PR merge blockers
+#   security/     security-finding severity and status (Cozystack-specific)
+#   size:         PR size (auto-applied)
+#
+# `aliases:` lets EndBug/label-sync rename existing labels without losing
+# references on already-tagged issues and PRs.
+#
+# GitHub-default labels not migrated here (`wontfix`, `invalid`) currently
+# carry zero issues/PRs in this repo and will be removed in a follow-up
+# cleanup PR rather than aliased to a different namespace.
+
+# ──────────────────────────────────────────────
+# kind/ — issue or PR type
+# ──────────────────────────────────────────────
+
+- name: kind/bug
+  color: 'd73a4a'
+  description: Categorizes issue or PR as related to a bug
+  aliases: ['bug']
+
+- name: kind/feature
+  color: 'a2eeef'
+  description: Categorizes issue or PR as related to a new feature
+  aliases: ['enhancement']
+
+- name: kind/documentation
+  color: '0075ca'
+  description: Categorizes issue or PR as related to documentation
+  aliases: ['documentation']
+
+- name: kind/support
+  color: 'd876e3'
+  description: Categorizes issue as a support question
+  aliases: ['question']
+
+- name: kind/cleanup
+  color: 'c7def8'
+  description: Categorizes issue or PR as related to cleanup of code, process, or technical debt
+
+- name: kind/regression
+  color: 'e11d21'
+  description: Categorizes issue or PR as related to a regression from a prior release
+
+- name: kind/flake
+  color: 'f7c6c7'
+  description: Categorizes issue or PR as related to a flaky test
+
+- name: kind/failing-test
+  color: 'e11d21'
+  description: Categorizes issue or PR as related to a consistently or frequently failing test
+
+- name: kind/api-change
+  color: 'c7def8'
+  description: Categorizes issue or PR as related to adding, removing, or otherwise changing an API
+
+- name: kind/breaking-change
+  color: 'e11d21'
+  description: Indicates the change introduces a breaking API or behaviour change
+
+# ──────────────────────────────────────────────
+# priority/ — urgency
+# ──────────────────────────────────────────────
+
+- name: priority/critical-urgent
+  color: 'e11d21'
+  description: Highest priority. Must be actively worked on as someone's top priority right now
+
+- name: priority/important-soon
+  color: 'eb6420'
+  description: Must be staffed and worked on either currently, or very soon, ideally in time for the next release
+
+- name: priority/important-longterm
+  color: 'fbca04'
+  description: Important over the long term, but may not be staffed and/or may need multiple releases to complete
+
+- name: priority/backlog
+  color: 'fef2c0'
+  description: General backlog priority. Lower than priority/important-longterm
+
+# ──────────────────────────────────────────────
+# triage/ — review state
+# ──────────────────────────────────────────────
+
+- name: triage/needs-triage
+  color: 'ededed'
+  description: Indicates an issue needs triage by a maintainer
+
+- name: triage/accepted
+  color: '0e8a16'
+  description: Indicates an issue is ready to be actively worked on
+
+- name: triage/needs-information
+  color: 'fbca04'
+  description: Indicates an issue needs more information in order to work on it
+
+- name: triage/not-reproducible
+  color: 'fbca04'
+  description: Indicates an issue can not be reproduced as described
+
+- name: triage/duplicate
+  color: 'cfd3d7'
+  description: Indicates an issue is a duplicate of another issue
+  aliases: ['duplicate']
+
+- name: triage/unresolved
+  color: 'cfd3d7'
+  description: Indicates an issue that can not or will not be resolved
+
+# ──────────────────────────────────────────────
+# lifecycle/ — issue or PR lifecycle
+# ──────────────────────────────────────────────
+
+- name: lifecycle/active
+  color: '1d76db'
+  description: Indicates that an issue or PR is actively being worked on by a contributor
+
+- name: lifecycle/frozen
+  color: 'db5dd6'
+  description: Indicates that an issue or PR should not be auto-closed due to staleness
+  aliases: ['frozen']
+
+- name: lifecycle/stale
+  color: 'dadada'
+  description: Denotes an issue or PR has remained open with no activity and has become stale
+  aliases: ['stale']
+
+- name: lifecycle/rotten
+  color: '795548'
+  description: Denotes an issue or PR that has aged beyond stale and will be auto-closed
+
+# ──────────────────────────────────────────────
+# area/ — subsystem (extensible)
+# Add a new area/* when there are 3+ open issues on the topic.
+# ──────────────────────────────────────────────
+
+- name: area/api
+  color: 'bfd4f2'
+  description: Issues or PRs related to the cozystack-api aggregated API server
+
+- name: area/ai
+  color: 'bfd4f2'
+  description: Issues or PRs related to AI agent guides, AGENTS.md, docs/agents/
+
+- name: area/build
+  color: 'bfd4f2'
+  description: Issues or PRs related to image build infrastructure, multi-arch support
+
+- name: area/ci
+  color: 'bfd4f2'
+  description: Issues or PRs related to CI workflows, GitHub Actions, automation
+
+- name: area/dashboard
+  color: 'bfd4f2'
+  description: Issues or PRs related to the dashboard / UI
+
+- name: area/extra
+  color: 'bfd4f2'
+  description: Issues or PRs related to tenant-specific modules (packages/extra/)
+
+- name: area/database
+  color: 'bfd4f2'
+  description: Issues or PRs related to managed databases (postgres, mariadb, redis, etcd, kafka, clickhouse)
+
+- name: area/kubernetes
+  color: 'bfd4f2'
+  description: Issues or PRs related to the tenant Kubernetes app
+
+- name: area/monitoring
+  color: 'bfd4f2'
+  description: Issues or PRs related to the monitoring stack (vlogs, vmstack, grafana, workloadmonitor)
+
+- name: area/networking
+  color: 'bfd4f2'
+  description: Issues or PRs related to networking (ingress, gateway, vpn, metallb, cilium, kube-ovn)
+
+- name: area/platform
+  color: 'bfd4f2'
+  description: Issues or PRs related to platform infrastructure (bundle, flux, talos, installer)
+
+- name: area/release
+  color: 'bfd4f2'
+  description: Issues or PRs related to release tooling (changelog, backport, release pipeline)
+
+- name: area/storage
+  color: 'bfd4f2'
+  description: Issues or PRs related to storage (linstor, seaweedfs, bucket, velero, harbor)
+
+- name: area/testing
+  color: 'bfd4f2'
+  description: Issues or PRs related to testing (e2e, bats, unit tests)
+
+- name: area/virtualization
+  color: 'bfd4f2'
+  description: Issues or PRs related to virtualization (kubevirt, cdi, vmi, vm-import)
+
+- name: area/uncategorized
+  color: 'fbca04'
+  description: PR auto-labeler could not map title scope to a known area/*; please review
+
+# ──────────────────────────────────────────────
+# do-not-merge/ — PR merge blockers (Prow convention)
+# ──────────────────────────────────────────────
+
+- name: do-not-merge/work-in-progress
+  color: 'e11d21'
+  description: Indicates that a PR should not merge because it is a work in progress
+  # Both legacy spellings collapse here. EndBug processes aliases sequentially;
+  # the second rename hits a name collision and logs a warning — the legacy
+  # label survives and gets cleaned up in the follow-up dedup PR.
+  aliases: ['do-not-merge', 'do not merge']
+
+- name: do-not-merge/hold
+  color: 'e11d21'
+  description: Indicates that a PR should not merge because someone has issued /hold
+
+# ──────────────────────────────────────────────
+# Cozystack-specific (preserved)
+# ──────────────────────────────────────────────
+
+- name: epic
+  color: 'a335ee'
+  description: A large development increment that brings definite value to Cozystack users
+
+- name: community
+  color: '97458a'
+  description: Community contributions are welcome in this issue
+
+- name: help wanted
+  color: '008672'
+  description: Extra attention is needed
+
+- name: good first issue
+  color: '7057ff'
+  description: Good for newcomers
+
+- name: quality-of-life
+  color: 'aaaaaa'
+  description: QoL improvements
+
+- name: upstream-issue
+  color: 'aaaaaa'
+  description: Requires resolving an issue in an upstream project
+
+- name: backport
+  color: 'fbca04'
+  description: Should change be backported on previous release
+
+- name: backport-previous
+  color: 'fbd876'
+  description: Backport target — previous release line
+
+- name: release
+  color: 'aaaaaa'
+  description: Releasing a new Cozystack version
+
+- name: automated
+  color: 'ededed'
+  description: Created by automation
+
+- name: debug
+  color: '704479'
+  description: Debugging in progress
+
+- name: sponsored
+  color: '00ff00'
+  description: Sponsored work
+
+- name: lgtm
+  color: '238636'
+  description: This PR has been approved by a maintainer
+
+- name: ok-to-test
+  color: '00ff00'
+  description: Indicates a non-member PR is safe to run CI on
+
+# ──────────────────────────────────────────────
+# size: — PR size (auto-applied by sizing bot)
+# ──────────────────────────────────────────────
+
+- name: 'size:XS'
+  color: '00ff00'
+  description: This PR changes 0-9 lines, ignoring generated files
+
+- name: 'size:S'
+  color: '77b800'
+  description: This PR changes 10-29 lines, ignoring generated files
+
+- name: 'size:M'
+  color: 'ebb800'
+  description: This PR changes 30-99 lines, ignoring generated files
+
+- name: 'size:L'
+  color: 'eb9500'
+  description: This PR changes 100-499 lines, ignoring generated files
+
+- name: 'size:XL'
+  color: 'ff823f'
+  description: This PR changes 500-999 lines, ignoring generated files
+
+- name: 'size:XXL'
+  color: 'ffb8b8'
+  description: This PR changes 1000+ lines, ignoring generated files
+
+# ──────────────────────────────────────────────
+# security/ — security-finding severity and status
+# ──────────────────────────────────────────────
+
+- name: security
+  color: 'aaaaaa'
+  description: Security-related issues and features
+
+- name: security/critical
+  color: 'd73a4a'
+  description: Critical security vulnerability
+
+- name: security/high
+  color: 'e99695'
+  description: High severity security finding
+
+- name: security/medium
+  color: 'f9c513'
+  description: Medium severity security finding
+
+- name: security/low
+  color: '0e8a16'
+  description: Low severity security finding
+
+- name: security/triage-needed
+  color: 'fbca04'
+  description: Needs security triage
+
+- name: security/confirmed
+  color: '1d76db'
+  description: Confirmed vulnerability
+
+- name: security/false-positive
+  color: 'c5def5'
+  description: Triaged as false positive
+
+- name: security/accepted-risk
+  color: 'bfd4f2'
+  description: Risk accepted with justification
+
+- name: security/in-progress
+  color: '0075ca'
+  description: Fix in progress
+
+- name: security/fixed
+  color: '0e8a16'
+  description: Fix released
diff --git a/.github/workflows/auto-release.yaml b/.github/workflows/auto-release.yaml
index 19d4b460..44ca8ca6 100644
--- a/.github/workflows/auto-release.yaml
+++ b/.github/workflows/auto-release.yaml
@@ -20,6 +20,14 @@ jobs:
       pull-requests: read
 
     steps:
+      - name: Generate GitHub App token
+        id: app-token
+        uses: actions/create-github-app-token@v1
+        with:
+          app-id: ${{ secrets.COZYSTACK_CI_APP_ID }}
+          private-key: ${{ secrets.COZYSTACK_CI_PRIVATE_KEY }}
+          owner: cozystack
+
       - name: Checkout code
         uses: actions/checkout@v4
         with:
@@ -28,27 +36,27 @@ jobs:
 
       - name: Configure git
         env:
-          GH_PAT: ${{ secrets.GH_PAT }}
+          APP_TOKEN: ${{ steps.app-token.outputs.token }}
         run: |
-          git config user.name  "cozystack-bot"
-          git config user.email "217169706+cozystack-bot@users.noreply.github.com"
-          git remote set-url origin https://cozystack-bot:${GH_PAT}@github.com/${GITHUB_REPOSITORY}
+          git config user.name  "cozystack-ci[bot]"
+          git config user.email "274107086+cozystack-ci[bot]@users.noreply.github.com"
+          git remote set-url origin https://x-access-token:${APP_TOKEN}@github.com/${GITHUB_REPOSITORY}
           git config --unset-all http.https://github.com/.extraheader || true
 
       - name: Process release branches
         uses: actions/github-script@v7
         env:
-          GH_PAT: ${{ secrets.GH_PAT }}
+          APP_TOKEN: ${{ steps.app-token.outputs.token }}
         with:
-          github-token: ${{ secrets.GH_PAT }}
+          github-token: ${{ steps.app-token.outputs.token }}
           script: |
             const { execSync } = require('child_process');
 
-            // Configure git to use PAT for authentication
-            execSync('git config user.name "cozystack-bot"', { encoding: 'utf8' });
-            execSync('git config user.email "217169706+cozystack-bot@users.noreply.github.com"', { encoding: 'utf8' });
-            execSync(`git remote set-url origin https://cozystack-bot:${process.env.GH_PAT}@github.com/${process.env.GITHUB_REPOSITORY}`, { encoding: 'utf8' });
-            // Remove GITHUB_TOKEN extraheader to ensure PAT is used (needed to trigger other workflows)
+            // Configure git to use GitHub App token for authentication
+            execSync('git config user.name "cozystack-ci[bot]"', { encoding: 'utf8' });
+            execSync('git config user.email "274107086+cozystack-ci[bot]@users.noreply.github.com"', { encoding: 'utf8' });
+            execSync(`git remote set-url origin https://x-access-token:${process.env.APP_TOKEN}@github.com/${process.env.GITHUB_REPOSITORY}`, { encoding: 'utf8' });
+            // Remove GITHUB_TOKEN extraheader to ensure App token is used (needed to trigger other workflows)
             execSync('git config --unset-all http.https://github.com/.extraheader || true', { encoding: 'utf8' });
 
             // Get all release-X.Y branches
diff --git a/.github/workflows/codegen-drift.yml b/.github/workflows/codegen-drift.yml
new file mode 100644
index 00000000..a26caf5e
--- /dev/null
+++ b/.github/workflows/codegen-drift.yml
@@ -0,0 +1,61 @@
+name: Codegen Drift Check
+
+on:
+  pull_request:
+    types: [opened, synchronize, reopened]
+    paths:
+      - 'api/**'
+      - 'pkg/apis/**'
+      - 'pkg/generated/**'
+      - 'internal/crdinstall/manifests/**'
+      - 'packages/system/cozystack-controller/definitions/**'
+      - 'packages/system/application-definition-crd/definition/**'
+      - 'packages/system/backup-controller/definitions/**'
+      - 'packages/system/backupstrategy-controller/definitions/**'
+      - 'hack/update-codegen.sh'
+      - 'hack/boilerplate.go.txt'
+      - 'Makefile'
+      - 'go.mod'
+      - 'go.sum'
+      - '.github/workflows/codegen-drift.yml'
+
+concurrency:
+  group: codegen-drift-${{ github.workflow }}-${{ github.event.pull_request.number }}
+  cancel-in-progress: true
+
+jobs:
+  codegen-drift:
+    name: Verify generated code is up to date
+    runs-on: ubuntu-22.04
+    steps:
+      - name: Checkout code
+        uses: actions/checkout@v4
+
+      - name: Set up Go
+        uses: actions/setup-go@v5
+        with:
+          go-version-file: go.mod
+          cache: true
+
+      - name: Pre-fetch k8s.io/code-generator module
+        # hack/update-codegen.sh sources kube_codegen.sh from the Go module cache.
+        # The module is not declared in go.mod, so fetch it explicitly at the
+        # version pinned in the script.
+        run: |
+          version=$(grep -oP 'code-generator@\Kv[0-9.]+' hack/update-codegen.sh)
+          tmpdir=$(mktemp -d)
+          cd "$tmpdir"
+          go mod init codegen-fetch
+          go get "k8s.io/code-generator@${version}"
+
+      - name: Run make generate
+        run: make generate
+
+      - name: Fail on drift
+        run: |
+          if [ -n "$(git status --porcelain)" ]; then
+            echo "::error::'make generate' produced changes. Run 'make generate' locally and commit the result."
+            git status --short
+            git diff --color=always
+            exit 1
+          fi
diff --git a/.github/workflows/labels.yaml b/.github/workflows/labels.yaml
new file mode 100644
index 00000000..69c18329
--- /dev/null
+++ b/.github/workflows/labels.yaml
@@ -0,0 +1,84 @@
+name: Labels
+
+on:
+  pull_request:
+    paths:
+      - .github/labels.yml
+      - .github/workflows/labels.yaml
+  push:
+    branches: [main]
+    paths:
+      - .github/labels.yml
+      - .github/workflows/labels.yaml
+  workflow_dispatch:
+  schedule:
+    - cron: '17 4 * * 1'  # Mondays at 04:17 UTC
+
+permissions:
+  contents: read
+
+concurrency:
+  group: labels-sync
+  cancel-in-progress: false
+
+jobs:
+  validate:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Validate labels.yml schema
+        run: |
+          python3 - <<'PY'
+          import re, sys, yaml
+
+          path = '.github/labels.yml'
+          data = yaml.safe_load(open(path))
+          errors = []
+
+          # 1. description ≤ 100 chars (GitHub REST API limit)
+          for label in data:
+              desc = label.get('description', '') or ''
+              if len(desc) > 100:
+                  errors.append(f"{label['name']}: description {len(desc)} chars (max 100)")
+
+          # 2. color is 6-char hex without leading #
+          for label in data:
+              color = label.get('color', '') or ''
+              if not re.match(r'^[0-9A-Fa-f]{6}$', color):
+                  errors.append(f"{label['name']}: bad color {color!r} (must be 6-char hex without #)")
+
+          # 3. unique top-level names
+          names = [label['name'] for label in data]
+          dups = sorted({n for n in names if names.count(n) > 1})
+          for n in dups:
+              errors.append(f"duplicate name: {n}")
+
+          # 4. aliases do not collide with any top-level name
+          name_set = set(names)
+          for label in data:
+              for alias in (label.get('aliases') or []):
+                  if alias in name_set:
+                      errors.append(f"alias {alias!r} (under {label['name']!r}) collides with a top-level name")
+
+          if errors:
+              for err in errors:
+                  print(f"::error::{err}")
+              sys.exit(1)
+
+          print(f"labels.yml schema OK ({len(data)} labels)")
+          PY
+
+  sync:
+    needs: validate
+    if: github.event_name != 'pull_request'
+    runs-on: ubuntu-latest
+    permissions:
+      issues: write
+      pull-requests: write
+    steps:
+      - uses: actions/checkout@v4
+      - uses: EndBug/label-sync@v2
+        with:
+          config-file: .github/labels.yml
+          delete-other-labels: false
diff --git a/.github/workflows/pr-labeler.yaml b/.github/workflows/pr-labeler.yaml
new file mode 100644
index 00000000..6e6f3017
--- /dev/null
+++ b/.github/workflows/pr-labeler.yaml
@@ -0,0 +1,206 @@
+name: PR Auto-Label
+
+on:
+  pull_request_target:
+    types: [opened, edited, reopened, synchronize]
+
+permissions:
+  contents: read
+  pull-requests: write
+
+jobs:
+  label:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Apply labels from PR title
+        uses: actions/github-script@v7
+        with:
+          script: |
+            // Conventional Commits types accepted by Cozystack (per docs/agents/contributing.md):
+            //   feat, fix, docs, style, refactor, perf, test, build, ci, chore, revert
+            // Mapping below maps a subset to kind/* — types not listed do not produce a kind/*.
+            const typeToKind = {
+              feat:     'kind/feature',
+              fix:      'kind/bug',
+              docs:     'kind/documentation',
+              chore:    'kind/cleanup',
+              refactor: 'kind/cleanup',
+              // style, perf, test, build, ci, revert — no kind mapping
+            };
+
+            // scope -> area/* mapping. Keys are the scopes observed in cozystack issues
+            // and PRs. Add new entries when a scope recurs (3+ times).
+            const scopeToArea = {
+              // area/api
+              'api':                 'area/api',
+              'cozystack-api':       'area/api',
+
+              // area/ai
+              'agents':              'area/ai',
+              'ai':                  'area/ai',
+
+              // area/build
+              'build':               'area/build',
+
+              // area/ci
+              'ci':                  'area/ci',
+
+              // area/dashboard
+              'dashboard':           'area/dashboard',
+
+              // area/database
+              'postgres':            'area/database',
+              'postgres-operator':   'area/database',
+              'mariadb':             'area/database',
+              'mariadb-operator':    'area/database',
+              'redis':               'area/database',
+              'etcd':                'area/database',
+              'kafka':               'area/database',
+              'clickhouse':          'area/database',
+
+              // area/extra
+              'extra':               'area/extra',
+
+              // area/kubernetes
+              'kubernetes':          'area/kubernetes',
+              'apps/kubernetes':     'area/kubernetes',
+
+              // area/monitoring
+              'monitoring':          'area/monitoring',
+              'vlogs':               'area/monitoring',
+              'vmstack':             'area/monitoring',
+              'grafana':             'area/monitoring',
+              'workloadmonitor':     'area/monitoring',
+
+              // area/networking
+              'ingress':             'area/networking',
+              'ingress-nginx':       'area/networking',
+              'gateway':             'area/networking',
+              'vpn':                 'area/networking',
+              'metallb':             'area/networking',
+              'cilium':              'area/networking',
+              'kube-ovn':            'area/networking',
+              'tcp-balancer':        'area/networking',
+              'securitygroups':      'area/networking',
+              'cozy-proxy':          'area/networking',
+
+              // area/platform
+              'platform':            'area/platform',
+              'bundle':              'area/platform',
+              'flux':                'area/platform',
+              'fluxcd':              'area/platform',
+              'cluster-api':         'area/platform',
+              'talos':               'area/platform',
+              'installer':           'area/platform',
+              'cozyctl':             'area/platform',
+              'cozystack-engine':    'area/platform',
+              'cozy-lib':            'area/platform',
+
+              // area/release
+              'backport':            'area/release',
+              'release':             'area/release',
+
+              // area/storage
+              'seaweedfs':           'area/storage',
+              'seaweedfs-cosi-driver': 'area/storage',
+              'bucket':              'area/storage',
+              'linstor':             'area/storage',
+              'velero':              'area/storage',
+              'harbor':              'area/storage',
+              'backups':             'area/storage',
+
+              // area/testing
+              'tests':               'area/testing',
+              'e2e':                 'area/testing',
+
+              // area/virtualization
+              'kubevirt':            'area/virtualization',
+              'cdi':                 'area/virtualization',
+              'vmi':                 'area/virtualization',
+              'vm-import':           'area/virtualization',
+              'virtual-machine':     'area/virtualization',
+              'hami':                'area/virtualization',
+              'gpu-operator':        'area/virtualization',
+            };
+
+            const pr = context.payload.pull_request;
+            const title = pr.title || '';
+            const body  = pr.body  || '';
+            const existing = new Set((pr.labels || []).map(l => l.name));
+            const toAdd = new Set();
+
+            // 1. Strip "[Backport release-1.x]" prefix if present.
+            const backportMatch = title.match(/^\[Backport ([^\]]+)\]\s+(.+)$/);
+            const cleanTitle = backportMatch ? backportMatch[2] : title;
+            if (backportMatch) {
+              toAdd.add('area/release');
+              toAdd.add('backport');
+            }
+
+            // 2. Try Conventional Commits form: type(scope)?(!)?: description
+            const conv = cleanTitle.match(/^([a-z]+)(?:\(([^)]+)\))?(!)?:\s*.+$/);
+            // 3. Fall back to bracket form: [scope] description
+            const bracket = !conv && cleanTitle.match(/^\[([^\]]+)\]\s+.+$/);
+
+            let type = null, scopeStr = null, breaking = false;
+            if (conv) {
+              type = conv[1];
+              scopeStr = conv[2] || null;
+              breaking = !!conv[3];
+            } else if (bracket) {
+              scopeStr = bracket[1];
+            }
+
+            // 4. Detect BREAKING CHANGE: or BREAKING-CHANGE: footer in body.
+            // Conventional Commits 1.0 spec item 16 treats them as synonymous.
+            if (/^BREAKING[ -]CHANGE:/m.test(body)) {
+              breaking = true;
+            }
+
+            // 5. Apply kind/* from type.
+            if (type) {
+              if (typeToKind[type]) {
+                toAdd.add(typeToKind[type]);
+              } else {
+                core.warning(`type "${type}" has no kind/* mapping — typo or new type? See .github/workflows/pr-labeler.yaml typeToKind`);
+              }
+            }
+
+            // 6. Apply area/* from scope. Composite scopes split on comma.
+            const scopes = (scopeStr || '')
+              .split(/,\s*/)
+              .map(s => s.trim())
+              .filter(Boolean);
+            for (const s of scopes) {
+              if (scopeToArea[s]) {
+                toAdd.add(scopeToArea[s]);
+              } else {
+                core.warning(`scope "${s}" has no area/* mapping — consider extending scopeToArea in .github/workflows/pr-labeler.yaml if it recurs`);
+              }
+            }
+
+            // 7. kind/breaking-change.
+            if (breaking) {
+              toAdd.add('kind/breaking-change');
+            }
+
+            // 8. Fallback: no area/* applied -> area/uncategorized.
+            const hasArea = [...toAdd].some(l => l.startsWith('area/'));
+            if (!hasArea) {
+              toAdd.add('area/uncategorized');
+            }
+
+            // 9. Additive only — never remove existing labels.
+            const newLabels = [...toAdd].filter(l => !existing.has(l));
+            if (newLabels.length === 0) {
+              core.info('No new labels to apply');
+              return;
+            }
+
+            await github.rest.issues.addLabels({
+              owner: context.repo.owner,
+              repo: context.repo.repo,
+              issue_number: pr.number,
+              labels: newLabels,
+            });
+            core.info(`Applied labels: ${newLabels.join(', ')}`);
diff --git a/.github/workflows/pull-requests-release.yaml b/.github/workflows/pull-requests-release.yaml
index 72f31b54..bb7f5da5 100644
--- a/.github/workflows/pull-requests-release.yaml
+++ b/.github/workflows/pull-requests-release.yaml
@@ -23,6 +23,14 @@ jobs:
       contains(github.event.pull_request.labels.*.name, 'release')
 
     steps:
+      - name: Generate GitHub App token
+        id: app-token
+        uses: actions/create-github-app-token@v1
+        with:
+          app-id: ${{ secrets.COZYSTACK_CI_APP_ID }}
+          private-key: ${{ secrets.COZYSTACK_CI_PRIVATE_KEY }}
+          owner: cozystack
+
       # Extract tag from branch name  (branch = release-X.Y.Z*)
       - name: Extract tag from branch name
         id: get_tag
@@ -47,11 +55,11 @@ jobs:
 
       - name: Create tag on merge commit
         env:
-          GH_PAT: ${{ secrets.GH_PAT }}
+          APP_TOKEN: ${{ steps.app-token.outputs.token }}
         run: |
-          git config user.name  "cozystack-bot"
-          git config user.email "217169706+cozystack-bot@users.noreply.github.com"
-          git remote set-url origin https://cozystack-bot:${GH_PAT}@github.com/${GITHUB_REPOSITORY}
+          git config user.name  "cozystack-ci[bot]"
+          git config user.email "274107086+cozystack-ci[bot]@users.noreply.github.com"
+          git remote set-url origin https://x-access-token:${APP_TOKEN}@github.com/${GITHUB_REPOSITORY}
           git tag -f ${{ steps.get_tag.outputs.tag }} ${{ github.sha }}
           git push -f origin ${{ steps.get_tag.outputs.tag }}
 
@@ -59,7 +67,7 @@ jobs:
       - name: Ensure maintenance branch release-X.Y
         uses: actions/github-script@v7
         with:
-          github-token: ${{ secrets.GH_PAT }}
+          github-token: ${{ steps.app-token.outputs.token }}
           script: |
             const tag = '${{ steps.get_tag.outputs.tag }}';  // e.g. v0.1.3 or v0.1.3-rc3
             const match = tag.match(/^v(\d+)\.(\d+)\.\d+(?:[-\w\.]+)?$/);
diff --git a/.github/workflows/pull-requests.yaml b/.github/workflows/pull-requests.yaml
index 76dbfc17..d9898616 100644
--- a/.github/workflows/pull-requests.yaml
+++ b/.github/workflows/pull-requests.yaml
@@ -99,6 +99,14 @@ jobs:
       disk_id:      ${{ steps.fetch_assets.outputs.disk_id }}
 
     steps:
+      - name: Generate GitHub App token
+        id: app-token
+        uses: actions/create-github-app-token@v1
+        with:
+          app-id: ${{ secrets.COZYSTACK_CI_APP_ID }}
+          private-key: ${{ secrets.COZYSTACK_CI_PRIVATE_KEY }}
+          owner: cozystack
+
       - name: Checkout code
         if: contains(github.event.pull_request.labels.*.name, 'release')
         uses: actions/checkout@v4
@@ -125,7 +133,7 @@ jobs:
         id: fetch_assets
         uses: actions/github-script@v7
         with:
-          github-token: ${{ secrets.GH_PAT }}
+          github-token: ${{ steps.app-token.outputs.token }}
           script: |
             const tag = '${{ steps.get_tag.outputs.tag }}';
             const releases = await github.rest.repos.listReleases({
@@ -159,6 +167,15 @@ jobs:
     if: ${{ always() && (needs.build.result == 'success' || needs.resolve_assets.result == 'success') }}
 
     steps:
+      - name: Generate GitHub App token
+        if: contains(github.event.pull_request.labels.*.name, 'release')
+        id: app-token
+        uses: actions/create-github-app-token@v1
+        with:
+          app-id: ${{ secrets.COZYSTACK_CI_APP_ID }}
+          private-key: ${{ secrets.COZYSTACK_CI_PRIVATE_KEY }}
+          owner: cozystack
+
       # ▸ Checkout and prepare the codebase
       - name: Checkout code
         uses: actions/checkout@v4
@@ -188,11 +205,11 @@ jobs:
         if: contains(github.event.pull_request.labels.*.name, 'release')
         run: |
           mkdir -p _out/assets
-          curl -sSL -H "Authorization: token ${GH_PAT}" -H "Accept: application/octet-stream" \
+          curl -sSL -H "Authorization: token ${APP_TOKEN}" -H "Accept: application/octet-stream" \
             -o _out/assets/nocloud-amd64.raw.xz \
             "https://api.github.com/repos/${GITHUB_REPOSITORY}/releases/assets/${{ needs.resolve_assets.outputs.disk_id }}"
         env:
-          GH_PAT: ${{ secrets.GH_PAT }}
+          APP_TOKEN: ${{ steps.app-token.outputs.token }}
 
       - name: Set sandbox ID
         run: echo "SANDBOX_NAME=cozy-e2e-sandbox-$(echo "${GITHUB_REPOSITORY}:${GITHUB_WORKFLOW}:${GITHUB_REF}" | sha256sum | cut -c1-10)" >> $GITHUB_ENV
diff --git a/.github/workflows/tags.yaml b/.github/workflows/tags.yaml
index 2bbdb7a6..1b68ee32 100644
--- a/.github/workflows/tags.yaml
+++ b/.github/workflows/tags.yaml
@@ -25,6 +25,14 @@ jobs:
       actions: write
 
     steps:
+      - name: Generate GitHub App token
+        id: app-token
+        uses: actions/create-github-app-token@v1
+        with:
+          app-id: ${{ secrets.COZYSTACK_CI_APP_ID }}
+          private-key: ${{ secrets.COZYSTACK_CI_PRIVATE_KEY }}
+          owner: cozystack
+
       # Check if a non-draft release with this tag already exists
       - name: Check if release already exists
         id: check_release
@@ -115,11 +123,11 @@ jobs:
       - name: Commit release artifacts
         if: steps.check_release.outputs.skip == 'false'
         env:
-          GH_PAT: ${{ secrets.GH_PAT }}
+          APP_TOKEN: ${{ steps.app-token.outputs.token }}
         run: |
-          git config user.name  "cozystack-bot"
-          git config user.email "217169706+cozystack-bot@users.noreply.github.com"
-          git remote set-url origin https://cozystack-bot:${GH_PAT}@github.com/${GITHUB_REPOSITORY}
+          git config user.name  "cozystack-ci[bot]"
+          git config user.email "274107086+cozystack-ci[bot]@users.noreply.github.com"
+          git remote set-url origin https://x-access-token:${APP_TOKEN}@github.com/${GITHUB_REPOSITORY}
           git config --unset-all http.https://github.com/.extraheader || true
           git add .
           git commit -m "Prepare release ${GITHUB_REF#refs/tags/}" -s || echo "No changes to commit"
@@ -128,13 +136,13 @@ jobs:
       - name: Tag API submodule
         if: steps.check_release.outputs.skip == 'false'
         env:
-          GH_PAT: ${{ secrets.GH_PAT }}
+          APP_TOKEN: ${{ steps.app-token.outputs.token }}
         run: |
           VTAG="${{ steps.tag.outputs.tag }}"
           SUBTAG="api/apps/v1alpha1/${VTAG}"
-          git config user.name  "cozystack-bot"
-          git config user.email "217169706+cozystack-bot@users.noreply.github.com"
-          git remote set-url origin https://cozystack-bot:${GH_PAT}@github.com/${GITHUB_REPOSITORY}
+          git config user.name  "cozystack-ci[bot]"
+          git config user.email "274107086+cozystack-ci[bot]@users.noreply.github.com"
+          git remote set-url origin https://x-access-token:${APP_TOKEN}@github.com/${GITHUB_REPOSITORY}
           TARGET="$(git rev-parse "${VTAG}^{}")"
           git tag -f "${SUBTAG}" "$TARGET"
           git push -f origin "refs/tags/${SUBTAG}"
@@ -183,11 +191,11 @@ jobs:
       - name: Create release branch
         if: steps.check_release.outputs.skip == 'false'
         env:
-          GH_PAT: ${{ secrets.GH_PAT }}
+          APP_TOKEN: ${{ steps.app-token.outputs.token }}
         run: |
-          git config user.name  "cozystack-bot"
-          git config user.email "217169706+cozystack-bot@users.noreply.github.com"
-          git remote set-url origin https://cozystack-bot:${GH_PAT}@github.com/${GITHUB_REPOSITORY}
+          git config user.name  "cozystack-ci[bot]"
+          git config user.email "274107086+cozystack-ci[bot]@users.noreply.github.com"
+          git remote set-url origin https://x-access-token:${APP_TOKEN}@github.com/${GITHUB_REPOSITORY}
           BRANCH="release-${GITHUB_REF#refs/tags/v}"
           git branch -f "$BRANCH"
           git push -f origin "$BRANCH"
@@ -197,7 +205,7 @@ jobs:
         if: steps.check_release.outputs.skip == 'false'
         uses: actions/github-script@v7
         with:
-          github-token: ${{ secrets.GH_PAT }}
+          github-token: ${{ steps.app-token.outputs.token }}
           script: |
             const version = context.ref.replace('refs/tags/v', '');
             const base    = '${{ steps.get_base.outputs.branch }}';
@@ -215,7 +223,7 @@ jobs:
                 repo:  context.repo.repo,
                 head,
                 base,
-                title: `Release v${version}`,
+                title: `chore(release): cut v${version}`,
                 body:  `This PR prepares the release \`v${version}\`.`,
                 draft: false
               });
@@ -239,6 +247,29 @@ jobs:
       pull-requests: write
     if: needs.prepare-release.result == 'success'
     steps:
+      - name: Generate GitHub App token
+        id: app-token
+        uses: actions/create-github-app-token@v1
+        with:
+          app-id: ${{ secrets.COZYSTACK_CI_APP_ID }}
+          private-key: ${{ secrets.COZYSTACK_CI_PRIVATE_KEY }}
+          owner: cozystack
+
+      # Read-only token for the AI step. Minting a separate scoped token
+      # means the Generate changelog using AI step cannot push branches,
+      # open PRs, or mutate any repository even with --allow-all-tools,
+      # regardless of whether the agent follows the prompt's instructions.
+      - name: Generate read-only GitHub App token
+        id: app-token-read
+        uses: actions/create-github-app-token@v1
+        with:
+          app-id: ${{ secrets.COZYSTACK_CI_APP_ID }}
+          private-key: ${{ secrets.COZYSTACK_CI_PRIVATE_KEY }}
+          owner: cozystack
+          permission-contents: read
+          permission-pull-requests: read
+          permission-metadata: read
+
       - name: Parse tag
         id: tag
         uses: actions/github-script@v7
@@ -261,7 +292,7 @@ jobs:
           ref: main
           fetch-depth: 0
           fetch-tags: true
-          token: ${{ secrets.GH_PAT }}
+          token: ${{ steps.app-token.outputs.token }}
 
       - name: Check if changelog already exists
         id: check_changelog
@@ -287,56 +318,65 @@ jobs:
 
       - name: Generate changelog using AI
         if: steps.check_changelog.outputs.exists == 'false'
+        timeout-minutes: 30
         env:
           COPILOT_GITHUB_TOKEN: ${{ secrets.COPILOT_GITHUB_TOKEN }}
-          GH_TOKEN: ${{ secrets.GH_PAT }}
+          GH_TOKEN: ${{ steps.app-token-read.outputs.token }}
+          VERSION: ${{ steps.tag.outputs.version }}
         run: |
-          copilot --prompt "prepare changelog file for tagged release v${{ steps.tag.outputs.version }}, use @docs/agents/changelog.md for it. Create the changelog file at docs/changelogs/v${{ steps.tag.outputs.version }}.md" \
+          copilot \
+            --prompt "Generate the release changelog for tag v${VERSION}. Follow the instructions in @docs/agents/changelog.md exactly, including the 'Scope and boundaries' section at the top. Your deliverable is the single file docs/changelogs/v${VERSION}.md — write it and exit; this workflow handles branching, committing, pushing, and opening the PR." \
             --allow-all-tools --allow-all-paths < /dev/null
 
       - name: Create changelog branch and commit
         if: steps.check_changelog.outputs.exists == 'false'
         env:
-          GH_PAT: ${{ secrets.GH_PAT }}
+          APP_TOKEN: ${{ steps.app-token.outputs.token }}
+          VERSION: ${{ steps.tag.outputs.version }}
         run: |
-          git config user.name  "cozystack-bot"
-          git config user.email "217169706+cozystack-bot@users.noreply.github.com"
-          git remote set-url origin https://cozystack-bot:${GH_PAT}@github.com/${GITHUB_REPOSITORY}
-          
-          CHANGELOG_FILE="docs/changelogs/v${{ steps.tag.outputs.version }}.md"
-          CHANGELOG_BRANCH="changelog-v${{ steps.tag.outputs.version }}"
-          
-          if [ -f "$CHANGELOG_FILE" ]; then
-            # Fetch latest main branch
-            git fetch origin main
-            
-            # Delete local branch if it exists
-            git branch -D "$CHANGELOG_BRANCH" 2>/dev/null || true
-            
-            # Create and checkout new branch from main
-            git checkout -b "$CHANGELOG_BRANCH" origin/main
-            
-            # Add and commit changelog
-            git add "$CHANGELOG_FILE"
-            if git diff --staged --quiet; then
-              echo "⚠️ No changes to commit (file may already be committed)"
-            else
-              git commit -m "docs: add changelog for v${{ steps.tag.outputs.version }}" -s
-              echo "✅ Changelog committed to branch $CHANGELOG_BRANCH"
-            fi
-            
-            # Push the branch (force push to update if it exists)
-            git push -f origin "$CHANGELOG_BRANCH"
-          else
-            echo "⚠️ Changelog file was not generated"
+          set -euo pipefail
+
+          CHANGELOG_FILE="docs/changelogs/v${VERSION}.md"
+          CHANGELOG_BRANCH="changelog-v${VERSION}"
+
+          if [ ! -f "$CHANGELOG_FILE" ]; then
+            echo "::error::Changelog file $CHANGELOG_FILE was not produced by the Generate changelog using AI step"
             exit 1
           fi
+          if [ ! -s "$CHANGELOG_FILE" ]; then
+            echo "::error::Changelog file $CHANGELOG_FILE is empty"
+            exit 1
+          fi
+
+          # Snapshot the file across the branch switch — the checkout below
+          # resets tracked files to match origin/main.
+          TEMP_FILE="$(mktemp)"
+          trap 'rm -f "$TEMP_FILE"' EXIT
+          cp "$CHANGELOG_FILE" "$TEMP_FILE"
+
+          git config user.name  "cozystack-ci[bot]"
+          git config user.email "274107086+cozystack-ci[bot]@users.noreply.github.com"
+          git remote set-url origin "https://x-access-token:${APP_TOKEN}@github.com/${GITHUB_REPOSITORY}"
+
+          git fetch origin main
+          git branch -D "$CHANGELOG_BRANCH" 2>/dev/null || true
+          git checkout -b "$CHANGELOG_BRANCH" origin/main
+
+          mkdir -p "$(dirname "$CHANGELOG_FILE")"
+          cp "$TEMP_FILE" "$CHANGELOG_FILE"
+
+          # The `check_changelog` step gated this job on the file being absent
+          # from origin/main, so `git add` + `git commit` must produce a diff.
+          # If they don't, something is wrong (e.g. empty file) — fail loud.
+          git add "$CHANGELOG_FILE"
+          git commit -m "docs: add changelog for v${VERSION}" -s
+          git push -f origin "$CHANGELOG_BRANCH"
 
       - name: Create PR for changelog
         if: steps.check_changelog.outputs.exists == 'false'
         uses: actions/github-script@v7
         with:
-          github-token: ${{ secrets.GH_PAT }}
+          github-token: ${{ steps.app-token.outputs.token }}
           script: |
             const version = '${{ steps.tag.outputs.version }}';
             const changelogBranch = `changelog-v${version}`;
@@ -371,7 +411,7 @@ jobs:
                 repo:  context.repo.repo,
                 head:  changelogBranch,
                 base:  baseBranch,
-                title: `docs: add changelog for v${version}`,
+                title: `docs(release): add changelog for v${version}`,
                 body:  `This PR adds the changelog for release \`v${version}\`.\n\n✅ Changelog has been automatically generated in \`docs/changelogs/v${version}.md\`.`,
                 draft: false
               });
@@ -381,7 +421,7 @@ jobs:
                 owner: context.repo.owner,
                 repo:  context.repo.repo,
                 issue_number: pr.data.number,
-                labels: ['documentation', 'automated']
+                labels: ['kind/documentation', 'automated']
               });
               
               console.log(`Created PR #${pr.data.number} for changelog`);
@@ -395,6 +435,14 @@ jobs:
     permissions:
       contents: read
     steps:
+      - name: Generate GitHub App token
+        id: app-token
+        uses: actions/create-github-app-token@v1
+        with:
+          app-id: ${{ secrets.COZYSTACK_CI_APP_ID }}
+          private-key: ${{ secrets.COZYSTACK_CI_PRIVATE_KEY }}
+          owner: cozystack
+
       - name: Parse tag
         id: tag
         uses: actions/github-script@v7
@@ -414,20 +462,61 @@ jobs:
         uses: actions/checkout@v4
         with:
           repository: cozystack/website
-          token: ${{ secrets.GH_PAT }}
+          token: ${{ steps.app-token.outputs.token }}
           ref: main
 
+      # Decide whether this release promotes the `next/` trunk to a new released
+      # version directory. Per the website repo's contract (see website#495):
+      #   - `make release-next` runs only for new minor/major final releases
+      #     (tag matches `vX.Y.Z` with no prerelease suffix AND `vX.Y/` does
+      #     not yet exist on disk).
+      #   - Prereleases and patch releases skip this step and let
+      #     `make update-all` handle routing on its own.
+      - name: Determine if this release promotes next/
+        id: promote
+        env:
+          TAG: ${{ steps.tag.outputs.tag }}
+        run: |
+          if [[ ! "$TAG" =~ ^v([0-9]+)\.([0-9]+)\.([0-9]+)$ ]]; then
+            echo "promote=false" >> "$GITHUB_OUTPUT"
+            echo "Prerelease tag '$TAG' — skipping release-next."
+            exit 0
+          fi
+          MAJOR="${BASH_REMATCH[1]}"
+          MINOR="${BASH_REMATCH[2]}"
+          if [[ "$MAJOR" == "0" ]]; then
+            DOC_VERSION="v0"
+          else
+            DOC_VERSION="v${MAJOR}.${MINOR}"
+          fi
+          if [[ -d "content/en/docs/${DOC_VERSION}" ]]; then
+            echo "promote=false" >> "$GITHUB_OUTPUT"
+            echo "content/en/docs/${DOC_VERSION}/ already exists — patch release, skipping release-next."
+          else
+            echo "promote=true" >> "$GITHUB_OUTPUT"
+            echo "New minor/major release ${DOC_VERSION} — will promote next/ → ${DOC_VERSION}/."
+          fi
+
+      - name: Promote next/ to released version
+        if: steps.promote.outputs.promote == 'true'
+        env:
+          GH_TOKEN: ${{ steps.app-token.outputs.token }}
+          TAG: ${{ steps.tag.outputs.tag }}
+        run: make release-next RELEASE_TAG="$TAG"
+
       - name: Update docs from release branch
         env:
-          GH_TOKEN: ${{ secrets.GH_PAT }}
-        run: make update-all BRANCH=release-${{ steps.tag.outputs.version }} RELEASE_TAG=${{ steps.tag.outputs.tag }}
+          GH_TOKEN: ${{ steps.app-token.outputs.token }}
+          TAG: ${{ steps.tag.outputs.tag }}
+          VERSION: ${{ steps.tag.outputs.version }}
+        run: make update-all BRANCH="release-$VERSION" RELEASE_TAG="$TAG"
 
       - name: Commit and push
         id: commit
         run: |
-          git config user.name  "cozystack-bot"
-          git config user.email "217169706+cozystack-bot@users.noreply.github.com"
-          git add content
+          git config user.name  "cozystack-ci[bot]"
+          git config user.email "274107086+cozystack-ci[bot]@users.noreply.github.com"
+          git add content hugo.yaml
           if git diff --cached --quiet; then
             echo "No changes to commit"
             echo "changed=false" >> $GITHUB_OUTPUT
@@ -443,7 +532,7 @@ jobs:
       - name: Open pull request
         if: steps.commit.outputs.changed == 'true'
         env:
-          GH_TOKEN: ${{ secrets.GH_PAT }}
+          GH_TOKEN: ${{ steps.app-token.outputs.token }}
         run: |
           BRANCH="update-docs-v${{ steps.tag.outputs.version }}"
           pr_state=$(gh pr view "$BRANCH" --repo cozystack/website --json state --jq .state 2>/dev/null || echo "")
diff --git a/.gitignore b/.gitignore
index 4e1e3119..61003de5 100644
--- a/.gitignore
+++ b/.gitignore
@@ -83,3 +83,4 @@ tmp/
 
 # build revision marker (generated by make image-packages)
 packages/core/platform/.build-revision
+.claude/
diff --git a/.pre-commit-config.yaml b/.pre-commit-config.yaml
index ac0f7e30..6492b92d 100644
--- a/.pre-commit-config.yaml
+++ b/.pre-commit-config.yaml
@@ -1,6 +1,17 @@
 repos:
 - repo: local
   hooks:
+    - id: run-make-generate-root
+      name: Run 'make generate' at repo root
+      entry: |
+        flock -x .git/pre-commit.lock sh -c '
+          echo "Running make generate at repo root"
+          make generate || exit $?
+          git diff --color=always | cat
+        '
+      language: system
+      files: ^(api/|pkg/apis/|pkg/generated/|internal/crdinstall/manifests/|packages/system/cozystack-controller/definitions/|packages/system/application-definition-crd/definition/|packages/system/backup-controller/definitions/|packages/system/backupstrategy-controller/definitions/|hack/update-codegen\.sh$|hack/boilerplate\.go\.txt$|Makefile$)
+      pass_filenames: false
     - id: run-make-generate
       name: Run 'make generate' in all app directories
       entry: |
diff --git a/AGENTS.md b/AGENTS.md
index eb1febf7..1e9c4d15 100644
--- a/AGENTS.md
+++ b/AGENTS.md
@@ -27,6 +27,12 @@ working with the **Cozystack** project.
   - Read: [`contributing.md`](./docs/agents/contributing.md)
   - Action: Read the file to understand git workflow, commit format, PR process
 
+- **Issue and PR labeling, triage** (e.g., "label this issue", "what label should I use", "triage this", "categorize")
+  - Read: [`.github/labels.yml`](./.github/labels.yml)
+  - Action: Use labels defined there. Conventions follow the Kubernetes scheme — `kind/*` (type), `area/*` (subsystem), `priority/*` (urgency), `triage/*` (review state), `lifecycle/*` (auto-close), `do-not-merge/*` (PR blockers), `security/*` (severity)
+  - For `area/*`: accuracy outweighs reuse. If no existing `area/*` truly fits the change, propose a new one via PR (extend `.github/labels.yml` and the scope mapping in `.github/workflows/pr-labeler.yaml`) — do not shoehorn the change into a wrong area. `area/uncategorized` is the auto-labeler fallback; treat it as a signal to pick a fit, create a new area, or correct the PR title
+  - PR titles: a Conventional Commits header (`type(scope): description`, types from [`contributing.md`](./docs/agents/contributing.md)) auto-applies `kind/*` and `area/*` via `.github/workflows/pr-labeler.yaml`. Append `!` (or add a `BREAKING CHANGE:` footer) to apply `kind/breaking-change`
+
 **Important rules:**
 - ✅ **ONLY read the file if the task matches the documented process scope** - do not read files for tasks that don't match their purpose
 - ✅ **ALWAYS read the file FIRST** before starting the task (when applicable)
@@ -53,7 +59,7 @@ working with the **Cozystack** project.
 ### Conventions
 - **Helm Charts**: Umbrella pattern, vendored upstream charts in `charts/`
 - **Go Code**: Controller-runtime patterns, kubebuilder style
-- **Git Commits**: `[component] Description` format with `--signoff`
+- **Git Commits**: Conventional Commits (`type(scope): description`) with `--signoff`
 
 ### What NOT to Do
 - ❌ Edit `/vendor/`, `zz_generated.*.go`, upstream charts directly
diff --git a/MAINTAINERS.md b/MAINTAINERS.md
index 9c44daab..d7061941 100644
--- a/MAINTAINERS.md
+++ b/MAINTAINERS.md
@@ -10,3 +10,5 @@
 | Timur Tukaev | [@tym83](https://github.com/tym83) | Ænix | Cozystack Website, Marketing, Community Management |
 | Kirill Klinchenkov | [@klinch0](https://github.com/klinch0) | Ænix | Core Maintainer |
 | Nikita Bykov | [@nbykov0](https://github.com/nbykov0) | Ænix | Maintainer of ARM and stuff |
+| Matthieu Robin | [@matthieu-robin](https://github.com/matthieu-robin) | Hidora | Managed Applications, Platform Quality & Benchmarking |
+| Mattia Eleuteri | [@mattia-eleuteri](https://github.com/mattia-eleuteri) | Hidora | CSI, Storage, Networking & Security |
diff --git a/Makefile b/Makefile
index 59a55bfb..acc9b6f6 100644
--- a/Makefile
+++ b/Makefile
@@ -1,4 +1,4 @@
-.PHONY: manifests assets unit-tests helm-unit-tests
+.PHONY: manifests assets unit-tests helm-unit-tests bats-unit-tests preflight
 
 include hack/common-envs.mk
 
@@ -22,6 +22,7 @@ build: build-deps
 	make -C packages/system/lineage-controller-webhook image
 	make -C packages/system/cilium image
 	make -C packages/system/linstor image
+	make -C packages/system/linstor-gui image
 	make -C packages/system/kubeovn-webhook image
 	make -C packages/system/kubeovn-plunger image
 	make -C packages/system/dashboard image
@@ -82,11 +83,46 @@ test:
 	make -C packages/core/testing apply
 	make -C packages/core/testing test
 
-unit-tests: helm-unit-tests
+unit-tests: helm-unit-tests bats-unit-tests go-unit-tests
 
 helm-unit-tests:
 	hack/helm-unit-tests.sh
 
+# Scoped go test over the cozystack-api surface that this repo owns. Kept
+# narrow intentionally - running `go test ./...` pulls in generated code
+# round-trip suites whose behavior depends on tool versions outside this
+# repo's control (kubebuilder, openapi-gen, etc.) and is better exercised
+# from their generator workflows.
+go-unit-tests:
+	go test ./pkg/registry/... ./pkg/config/... ./pkg/cmd/server/...
+
+# Discover every hack/*.bats file that is NOT an e2e test and run it
+# through cozytest.sh. Drop a new *.bats file in hack/ and it is picked
+# up automatically on the next `make unit-tests` run.
+#
+# Caveat: $(wildcard ...) returns space-separated names, so a filename
+# containing a literal space would split into multiple tokens here. All
+# current bats files use hyphen-separated names; if the project ever
+# introduces whitespace-bearing filenames this recipe must be rewritten
+# (e.g. to use `find ... -print0 | xargs -0`).
+BATS_UNIT_FILES := $(filter-out hack/e2e-%.bats,$(wildcard hack/*.bats))
+
+bats-unit-tests:
+	@if [ -z "$(BATS_UNIT_FILES)" ]; then \
+		echo "ERROR: no hack/*.bats unit test files found"; \
+		exit 1; \
+	fi
+	@for f in $(BATS_UNIT_FILES); do \
+		echo "--- running $$f ---"; \
+		hack/cozytest.sh "$$f" || exit 1; \
+	done
+
+# Operator-facing host preflight check. Warns about a standalone
+# containerd.service or docker.service running alongside the embedded
+# k3s runtime. Safe to run at any time; always exits 0.
+preflight:
+	@hack/check-host-runtime.sh
+
 prepare-env:
 	make -C packages/core/testing apply
 	make -C packages/core/testing prepare-cluster
diff --git a/README.md b/README.md
index dc1b0634..5bc8d7be 100644
--- a/README.md
+++ b/README.md
@@ -11,7 +11,7 @@
 
 # Cozystack
 
-**Cozystack** is a free PaaS platform and framework for building clouds.
+**Cozystack** is a free platform and framework for building clouds.
 
 Cozystack is a [CNCF Sandbox Level Project](https://www.cncf.io/sandbox-projects/) that was originally built and sponsored by [Ænix](https://aenix.io/).
 
diff --git a/api/apps/v1alpha1/kubernetes/types.go b/api/apps/v1alpha1/kubernetes/types.go
index 774fcdc5..88a3c97e 100644
--- a/api/apps/v1alpha1/kubernetes/types.go
+++ b/api/apps/v1alpha1/kubernetes/types.go
@@ -36,6 +36,9 @@ type ConfigSpec struct {
 	// Kubernetes control-plane configuration.
 	// +kubebuilder:default:={}
 	ControlPlane ControlPlane `json:"controlPlane"`
+	// Optional image overrides for air-gapped or rate-limited registries.
+	// +kubebuilder:default:={}
+	Images Images `json:"images"`
 }
 
 type APIServer struct {
@@ -66,6 +69,9 @@ type Addons struct {
 	// NVIDIA GPU Operator.
 	// +kubebuilder:default:={}
 	GpuOperator GPUOperatorAddon `json:"gpuOperator"`
+	// HAMi GPU virtualization middleware.
+	// +kubebuilder:default:={}
+	Hami HAMiAddon `json:"hami"`
 	// Ingress-NGINX controller.
 	// +kubebuilder:default:={}
 	IngressNginx IngressNginxAddon `json:"ingressNginx"`
@@ -157,6 +163,21 @@ type GatewayAPIAddon struct {
 	Enabled bool `json:"enabled"`
 }
 
+type HAMiAddon struct {
+	// Enable HAMi (requires GPU Operator).
+	// +kubebuilder:default:=false
+	Enabled bool `json:"enabled"`
+	// Custom Helm values overrides.
+	// +kubebuilder:default:={}
+	ValuesOverride k8sRuntime.RawExtension `json:"valuesOverride"`
+}
+
+type Images struct {
+	// Image used by the wait-for-kubeconfig init container. Empty falls back to images/busybox.tag.
+	// +kubebuilder:default:=""
+	WaitForKubeconfig string `json:"waitForKubeconfig,omitempty"`
+}
+
 type IngressNginxAddon struct {
 	// Enable the controller (requires nodes labeled `ingress-nginx`).
 	// +kubebuilder:default:=false
diff --git a/api/apps/v1alpha1/kubernetes/zz_generated.deepcopy.go b/api/apps/v1alpha1/kubernetes/zz_generated.deepcopy.go
index 2cbe3ba1..e021fbaa 100644
--- a/api/apps/v1alpha1/kubernetes/zz_generated.deepcopy.go
+++ b/api/apps/v1alpha1/kubernetes/zz_generated.deepcopy.go
@@ -49,6 +49,7 @@ func (in *Addons) DeepCopyInto(out *Addons) {
 	in.Fluxcd.DeepCopyInto(&out.Fluxcd)
 	out.GatewayAPI = in.GatewayAPI
 	in.GpuOperator.DeepCopyInto(&out.GpuOperator)
+	in.Hami.DeepCopyInto(&out.Hami)
 	in.IngressNginx.DeepCopyInto(&out.IngressNginx)
 	in.MonitoringAgents.DeepCopyInto(&out.MonitoringAgents)
 	in.Velero.DeepCopyInto(&out.Velero)
@@ -135,6 +136,7 @@ func (in *ConfigSpec) DeepCopyInto(out *ConfigSpec) {
 	}
 	in.Addons.DeepCopyInto(&out.Addons)
 	in.ControlPlane.DeepCopyInto(&out.ControlPlane)
+	out.Images = in.Images
 }
 
 // DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new ConfigSpec.
@@ -260,6 +262,37 @@ func (in *GatewayAPIAddon) DeepCopy() *GatewayAPIAddon {
 	return out
 }
 
+// DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
+func (in *HAMiAddon) DeepCopyInto(out *HAMiAddon) {
+	*out = *in
+	in.ValuesOverride.DeepCopyInto(&out.ValuesOverride)
+}
+
+// DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new HAMiAddon.
+func (in *HAMiAddon) DeepCopy() *HAMiAddon {
+	if in == nil {
+		return nil
+	}
+	out := new(HAMiAddon)
+	in.DeepCopyInto(out)
+	return out
+}
+
+// DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
+func (in *Images) DeepCopyInto(out *Images) {
+	*out = *in
+}
+
+// DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new Images.
+func (in *Images) DeepCopy() *Images {
+	if in == nil {
+		return nil
+	}
+	out := new(Images)
+	in.DeepCopyInto(out)
+	return out
+}
+
 // DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
 func (in *IngressNginxAddon) DeepCopyInto(out *IngressNginxAddon) {
 	*out = *in
diff --git a/api/apps/v1alpha1/postgresql/types.go b/api/apps/v1alpha1/postgresql/types.go
index fa85d51f..a2ff77a8 100644
--- a/api/apps/v1alpha1/postgresql/types.go
+++ b/api/apps/v1alpha1/postgresql/types.go
@@ -92,6 +92,9 @@ type Bootstrap struct {
 	// Timestamp (RFC3339) for point-in-time recovery; empty means latest.
 	// +kubebuilder:default:=""
 	RecoveryTime string `json:"recoveryTime,omitempty"`
+	// Barman server name (S3 path prefix) used by the original cluster when writing backups. Set this only when the original cluster had an explicit barmanObjectStore.serverName that differed from its Kubernetes resource name.
+	// +kubebuilder:default:=""
+	ServerName string `json:"serverName,omitempty"`
 }
 
 type Database struct {
diff --git a/api/apps/v1alpha1/vmdisk/types.go b/api/apps/v1alpha1/vmdisk/types.go
index aefac1e4..c493f49c 100644
--- a/api/apps/v1alpha1/vmdisk/types.go
+++ b/api/apps/v1alpha1/vmdisk/types.go
@@ -32,21 +32,28 @@ type ConfigSpec struct {
 }
 
 type Source struct {
+	// Clone an existing vm-disk.
+	Disk *SourceDisk `json:"disk,omitempty"`
 	// Download image from an HTTP source.
 	Http *SourceHTTP `json:"http,omitempty"`
-	// Use image by name.
+	// Use image by name from default collection.
 	Image *SourceImage `json:"image,omitempty"`
 	// Upload local image.
 	Upload *SourceUpload `json:"upload,omitempty"`
 }
 
+type SourceDisk struct {
+	// Name of the vm-disk to clone.
+	Name string `json:"name"`
+}
+
 type SourceHTTP struct {
 	// URL to download the image.
 	Url string `json:"url"`
 }
 
 type SourceImage struct {
-	// Name of the image to use (uploaded as "golden image" or from the list: `ubuntu`, `fedora`, `cirros`, `alpine`, `talos`).
+	// Name of the image to use.
 	Name string `json:"name"`
 }
 
diff --git a/api/apps/v1alpha1/vmdisk/zz_generated.deepcopy.go b/api/apps/v1alpha1/vmdisk/zz_generated.deepcopy.go
index e8ca986e..301bb1d9 100644
--- a/api/apps/v1alpha1/vmdisk/zz_generated.deepcopy.go
+++ b/api/apps/v1alpha1/vmdisk/zz_generated.deepcopy.go
@@ -70,6 +70,11 @@ func (in *ConfigSpec) DeepCopy() *ConfigSpec {
 // DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
 func (in *Source) DeepCopyInto(out *Source) {
 	*out = *in
+	if in.Disk != nil {
+		in, out := &in.Disk, &out.Disk
+		*out = new(SourceDisk)
+		**out = **in
+	}
 	if in.Http != nil {
 		in, out := &in.Http, &out.Http
 		*out = new(SourceHTTP)
@@ -97,6 +102,21 @@ func (in *Source) DeepCopy() *Source {
 	return out
 }
 
+// DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
+func (in *SourceDisk) DeepCopyInto(out *SourceDisk) {
+	*out = *in
+}
+
+// DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new SourceDisk.
+func (in *SourceDisk) DeepCopy() *SourceDisk {
+	if in == nil {
+		return nil
+	}
+	out := new(SourceDisk)
+	in.DeepCopyInto(out)
+	return out
+}
+
 // DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
 func (in *SourceHTTP) DeepCopyInto(out *SourceHTTP) {
 	*out = *in
diff --git a/api/apps/v1alpha1/vminstance/types.go b/api/apps/v1alpha1/vminstance/types.go
index 2bb059c6..9c48724a 100644
--- a/api/apps/v1alpha1/vminstance/types.go
+++ b/api/apps/v1alpha1/vminstance/types.go
@@ -26,6 +26,9 @@ type ConfigSpec struct {
 	// Ports to forward from outside the cluster.
 	// +kubebuilder:default:={22}
 	ExternalPorts []int `json:"externalPorts,omitempty"`
+	// Whether to accept ICMP traffic to the VM in PortList mode (preserves ping and PMTU discovery). No effect in WholeIP mode. Default true so ping behaves as users expect even when port filtering is in effect.
+	// +kubebuilder:default:=true
+	ExternalAllowICMP bool `json:"externalAllowICMP"`
 	// Requested running state of the VirtualMachineInstance
 	// +kubebuilder:default:="Always"
 	RunStrategy RunStrategy `json:"runStrategy"`
diff --git a/api/apps/v1alpha1/vminstance/zz_generated.deepcopy.go b/api/apps/v1alpha1/vminstance/zz_generated.deepcopy.go
index 00b2a710..86d5ddfc 100644
--- a/api/apps/v1alpha1/vminstance/zz_generated.deepcopy.go
+++ b/api/apps/v1alpha1/vminstance/zz_generated.deepcopy.go
@@ -63,9 +63,14 @@ func (in *ConfigSpec) DeepCopyInto(out *ConfigSpec) {
 		*out = make([]Disk, len(*in))
 		copy(*out, *in)
 	}
+	if in.Networks != nil {
+		in, out := &in.Networks, &out.Networks
+		*out = make([]Network, len(*in))
+		copy(*out, *in)
+	}
 	if in.Subnets != nil {
 		in, out := &in.Subnets, &out.Subnets
-		*out = make([]Subnet, len(*in))
+		*out = make([]Network, len(*in))
 		copy(*out, *in)
 	}
 	if in.Gpus != nil {
@@ -121,6 +126,21 @@ func (in *GPU) DeepCopy() *GPU {
 	return out
 }
 
+// DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
+func (in *Network) DeepCopyInto(out *Network) {
+	*out = *in
+}
+
+// DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new Network.
+func (in *Network) DeepCopy() *Network {
+	if in == nil {
+		return nil
+	}
+	out := new(Network)
+	in.DeepCopyInto(out)
+	return out
+}
+
 // DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
 func (in *Resources) DeepCopyInto(out *Resources) {
 	*out = *in
@@ -138,18 +158,3 @@ func (in *Resources) DeepCopy() *Resources {
 	in.DeepCopyInto(out)
 	return out
 }
-
-// DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
-func (in *Subnet) DeepCopyInto(out *Subnet) {
-	*out = *in
-}
-
-// DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new Subnet.
-func (in *Subnet) DeepCopy() *Subnet {
-	if in == nil {
-		return nil
-	}
-	out := new(Subnet)
-	in.DeepCopyInto(out)
-	return out
-}
diff --git a/api/apps/v1alpha1/vpc/zz_generated.deepcopy.go b/api/apps/v1alpha1/vpc/zz_generated.deepcopy.go
index cd718167..3cda29c4 100644
--- a/api/apps/v1alpha1/vpc/zz_generated.deepcopy.go
+++ b/api/apps/v1alpha1/vpc/zz_generated.deepcopy.go
@@ -58,6 +58,16 @@ func (in *ConfigSpec) DeepCopyInto(out *ConfigSpec) {
 		*out = make([]Subnet, len(*in))
 		copy(*out, *in)
 	}
+	if in.Peers != nil {
+		in, out := &in.Peers, &out.Peers
+		*out = make([]Peer, len(*in))
+		copy(*out, *in)
+	}
+	if in.Routes != nil {
+		in, out := &in.Routes, &out.Routes
+		*out = make([]Route, len(*in))
+		copy(*out, *in)
+	}
 }
 
 // DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new ConfigSpec.
@@ -70,6 +80,36 @@ func (in *ConfigSpec) DeepCopy() *ConfigSpec {
 	return out
 }
 
+// DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
+func (in *Peer) DeepCopyInto(out *Peer) {
+	*out = *in
+}
+
+// DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new Peer.
+func (in *Peer) DeepCopy() *Peer {
+	if in == nil {
+		return nil
+	}
+	out := new(Peer)
+	in.DeepCopyInto(out)
+	return out
+}
+
+// DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
+func (in *Route) DeepCopyInto(out *Route) {
+	*out = *in
+}
+
+// DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new Route.
+func (in *Route) DeepCopy() *Route {
+	if in == nil {
+		return nil
+	}
+	out := new(Route)
+	in.DeepCopyInto(out)
+	return out
+}
+
 // DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
 func (in *Subnet) DeepCopyInto(out *Subnet) {
 	*out = *in
diff --git a/api/backups/v1alpha1/zz_generated.deepcopy.go b/api/backups/v1alpha1/zz_generated.deepcopy.go
index b7387772..61b8f839 100644
--- a/api/backups/v1alpha1/zz_generated.deepcopy.go
+++ b/api/backups/v1alpha1/zz_generated.deepcopy.go
@@ -1,5 +1,21 @@
 //go:build !ignore_autogenerated
 
+/*
+Copyright 2025 The Cozystack Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/
+
 // Code generated by controller-gen. DO NOT EDIT.
 
 package v1alpha1
@@ -604,6 +620,11 @@ func (in *RestoreJobSpec) DeepCopyInto(out *RestoreJobSpec) {
 		*out = new(v1.TypedLocalObjectReference)
 		(*in).DeepCopyInto(*out)
 	}
+	if in.Options != nil {
+		in, out := &in.Options, &out.Options
+		*out = new(runtime.RawExtension)
+		(*in).DeepCopyInto(*out)
+	}
 }
 
 // DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new RestoreJobSpec.
@@ -645,4 +666,3 @@ func (in *RestoreJobStatus) DeepCopy() *RestoreJobStatus {
 	in.DeepCopyInto(out)
 	return out
 }
-
diff --git a/api/v1alpha1/packagesource_types.go b/api/v1alpha1/packagesource_types.go
index c0b943e8..75bf217b 100644
--- a/api/v1alpha1/packagesource_types.go
+++ b/api/v1alpha1/packagesource_types.go
@@ -132,6 +132,16 @@ type ComponentInstall struct {
 	// DependsOn is a list of component names that must be installed before this component
 	// +optional
 	DependsOn []string `json:"dependsOn,omitempty"`
+
+	// UpgradeCRDs controls how CRDs from the chart's crds/ directory are
+	// handled on HelmRelease upgrades. Maps to HelmRelease.Spec.Upgrade.CRDs.
+	// Empty string (default) preserves the helm-controller default (Skip).
+	// Use "CreateReplace" for operators that evolve their CRD set between
+	// versions. Warning: CreateReplace overwrites CRDs and may cause data
+	// loss if upstream drops fields from a CRD with live objects.
+	// +optional
+	// +kubebuilder:validation:Enum=Skip;Create;CreateReplace
+	UpgradeCRDs string `json:"upgradeCRDs,omitempty"`
 }
 
 // Component defines a single Helm release component within a package source
diff --git a/cmd/cozystack-controller/main.go b/cmd/cozystack-controller/main.go
index 77c4f9dc..bcaca9c9 100644
--- a/cmd/cozystack-controller/main.go
+++ b/cmd/cozystack-controller/main.go
@@ -42,6 +42,7 @@ import (
 	"github.com/cozystack/cozystack/internal/telemetry"
 
 	helmv2 "github.com/fluxcd/helm-controller/api/v2"
+	cosiv1alpha1 "sigs.k8s.io/container-object-storage-interface-api/apis/objectstorage/v1alpha1"
 	// +kubebuilder:scaffold:imports
 )
 
@@ -56,6 +57,7 @@ func init() {
 	utilruntime.Must(cozystackiov1alpha1.AddToScheme(scheme))
 	utilruntime.Must(dashboard.AddToScheme(scheme))
 	utilruntime.Must(helmv2.AddToScheme(scheme))
+	utilruntime.Must(cosiv1alpha1.AddToScheme(scheme))
 	// +kubebuilder:scaffold:scheme
 }
 
diff --git a/cmd/kubeovn-plunger/main.go b/cmd/kubeovn-plunger/main.go
index 210405ce..611d030b 100644
--- a/cmd/kubeovn-plunger/main.go
+++ b/cmd/kubeovn-plunger/main.go
@@ -29,6 +29,7 @@ import (
 	utilruntime "k8s.io/apimachinery/pkg/util/runtime"
 	clientgoscheme "k8s.io/client-go/kubernetes/scheme"
 	ctrl "sigs.k8s.io/controller-runtime"
+	"sigs.k8s.io/controller-runtime/pkg/cache"
 	"sigs.k8s.io/controller-runtime/pkg/healthz"
 	"sigs.k8s.io/controller-runtime/pkg/log/zap"
 	"sigs.k8s.io/controller-runtime/pkg/metrics"
@@ -131,6 +132,11 @@ func main() {
 		HealthProbeBindAddress: probeAddr,
 		LeaderElection:         enableLeaderElection,
 		LeaderElectionID:       "29a0338b.cozystack.io",
+		Cache: cache.Options{
+			DefaultNamespaces: map[string]cache.Config{
+				kubeOVNNamespace: {},
+			},
+		},
 		// LeaderElectionReleaseOnCancel defines if the leader should step down voluntarily
 		// when the Manager ends. This requires the binary to immediately end when the
 		// Manager is stopped, otherwise, this setting is unsafe. Setting this significantly
diff --git a/dashboards/gpu/gpu-efficiency.json b/dashboards/gpu/gpu-efficiency.json
new file mode 100644
index 00000000..28a24568
--- /dev/null
+++ b/dashboards/gpu/gpu-efficiency.json
@@ -0,0 +1,819 @@
+{
+  "uid": "gpu-efficiency",
+  "title": "GPU Efficiency Score",
+  "description": "Tensor saturation, util/watt and throttling — reveals inefficient GPU workloads",
+  "tags": [
+    "gpu",
+    "efficiency",
+    "finops"
+  ],
+  "timezone": "browser",
+  "editable": true,
+  "graphTooltip": 1,
+  "time": {
+    "from": "now-1h",
+    "to": "now"
+  },
+  "fiscalYearStartMonth": 0,
+  "schemaVersion": 42,
+  "panels": [
+    {
+      "type": "row",
+      "collapsed": false,
+      "title": "Overall efficiency metrics",
+      "gridPos": {
+        "h": 1,
+        "w": 24,
+        "x": 0,
+        "y": 0
+      },
+      "id": 1,
+      "panels": []
+    },
+    {
+      "type": "stat",
+      "id": 2,
+      "targets": [
+        {
+          "expr": "avg(gpu:tensor_saturation:avg5m) * 100",
+          "refId": "A"
+        }
+      ],
+      "title": "Avg Tensor Saturation",
+      "description": "Mean tensor core saturation across all GPUs. \u003c10% means GPUs are used inefficiently (workloads could move to CPU or optimize their code). Cluster-wide — DCGM metrics cannot be attributed to workload namespaces.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 5,
+        "w": 8,
+        "x": 0,
+        "y": 1
+      },
+      "repeatDirection": "h",
+      "options": {
+        "graphMode": "area",
+        "colorMode": "background",
+        "justifyMode": "auto",
+        "textMode": "value",
+        "wideLayout": true,
+        "showPercentChange": false,
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ]
+        },
+        "percentChangeColorMode": "standard",
+        "orientation": ""
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "percent",
+          "min": 0,
+          "max": 100,
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "value": null,
+                "color": "red"
+              },
+              {
+                "value": 10,
+                "color": "orange"
+              },
+              {
+                "value": 30,
+                "color": "yellow"
+              },
+              {
+                "value": 60,
+                "color": "green"
+              }
+            ]
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "stat",
+      "id": 3,
+      "targets": [
+        {
+          "expr": "avg(gpu:util_per_watt:avg5m)",
+          "refId": "A"
+        }
+      ],
+      "title": "Avg Utilization per Watt",
+      "description": "NVML utilization % per watt across all GPUs. Higher value = more efficient workload. Cluster-wide — DCGM metrics cannot be attributed to workload namespaces.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 5,
+        "w": 8,
+        "x": 8,
+        "y": 1
+      },
+      "repeatDirection": "h",
+      "options": {
+        "graphMode": "area",
+        "colorMode": "background",
+        "justifyMode": "auto",
+        "textMode": "value",
+        "wideLayout": true,
+        "showPercentChange": false,
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ]
+        },
+        "percentChangeColorMode": "standard",
+        "orientation": ""
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "none",
+          "decimals": 3,
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "value": null,
+                "color": "red"
+              },
+              {
+                "value": 0.5,
+                "color": "yellow"
+              },
+              {
+                "value": 1,
+                "color": "green"
+              }
+            ]
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "stat",
+      "id": 4,
+      "targets": [
+        {
+          "expr": "avg(gpu:power_throttle_fraction:rate5m)",
+          "refId": "A"
+        }
+      ],
+      "title": "Avg Power Throttling",
+      "description": "Fraction of time GPUs hit the TDP cap and lose performance. \u003e5% means tenants underutilize billed FLOPS. Cluster-wide — rule aggregates by (Hostname, gpu, UUID) and drops namespace label.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 5,
+        "w": 8,
+        "x": 16,
+        "y": 1
+      },
+      "repeatDirection": "h",
+      "options": {
+        "graphMode": "area",
+        "colorMode": "background",
+        "justifyMode": "auto",
+        "textMode": "value",
+        "wideLayout": true,
+        "showPercentChange": false,
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ]
+        },
+        "percentChangeColorMode": "standard",
+        "orientation": ""
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "percentunit",
+          "decimals": 2,
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "value": null,
+                "color": "green"
+              },
+              {
+                "value": 0.05,
+                "color": "yellow"
+              },
+              {
+                "value": 0.2,
+                "color": "red"
+              }
+            ]
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "row",
+      "collapsed": false,
+      "title": "NVML vs Tensor (mismatch detector)",
+      "gridPos": {
+        "h": 1,
+        "w": 24,
+        "x": 0,
+        "y": 6
+      },
+      "id": 10,
+      "panels": []
+    },
+    {
+      "type": "timeseries",
+      "id": 11,
+      "targets": [
+        {
+          "expr": "DCGM_FI_DEV_GPU_UTIL",
+          "legendFormat": "{{Hostname}}/{{gpu}}",
+          "refId": "A"
+        }
+      ],
+      "title": "NVML GPU Utilization",
+      "description": "Classic utilization metric. Shows activity of any engine (SM, copy, encoder). Cluster-wide — DCGM namespace is the exporter's own namespace, not the workload namespace.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 12,
+        "x": 0,
+        "y": 7
+      },
+      "repeatDirection": "h",
+      "options": {
+        "legend": {
+          "displayMode": "table",
+          "placement": "bottom",
+          "showLegend": false,
+          "calcs": [
+            "mean",
+            "max"
+          ]
+        },
+        "tooltip": {
+          "mode": "multi",
+          "sort": ""
+        }
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "percent",
+          "min": 0,
+          "max": 100,
+          "custom": {
+            "drawStyle": "line",
+            "lineInterpolation": "linear",
+            "fillOpacity": 10,
+            "showPoints": "never"
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "timeseries",
+      "id": 12,
+      "targets": [
+        {
+          "expr": "DCGM_FI_PROF_PIPE_TENSOR_ACTIVE * 100",
+          "legendFormat": "{{Hostname}}/{{gpu}}",
+          "refId": "A"
+        }
+      ],
+      "title": "Tensor Pipe Active",
+      "description": "Real tensor core load (HMMA). For AI/LLM inference it should be ≥20%, otherwise the workload is unoptimized. Cluster-wide — DCGM namespace is the exporter's own namespace, not the workload namespace.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 12,
+        "x": 12,
+        "y": 7
+      },
+      "repeatDirection": "h",
+      "options": {
+        "legend": {
+          "displayMode": "table",
+          "placement": "bottom",
+          "showLegend": false,
+          "calcs": [
+            "mean",
+            "max"
+          ]
+        },
+        "tooltip": {
+          "mode": "multi",
+          "sort": ""
+        }
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "percent",
+          "min": 0,
+          "max": 100,
+          "custom": {
+            "drawStyle": "line",
+            "lineInterpolation": "linear",
+            "fillOpacity": 10,
+            "showPoints": "never"
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "row",
+      "collapsed": false,
+      "title": "Per-GPU ranking",
+      "gridPos": {
+        "h": 1,
+        "w": 24,
+        "x": 0,
+        "y": 15
+      },
+      "id": 20,
+      "panels": []
+    },
+    {
+      "type": "table",
+      "id": 21,
+      "targets": [
+        {
+          "expr": "topk(20, gpu:tensor_saturation:avg5m * 100)",
+          "instant": true,
+          "range": false,
+          "format": "table",
+          "refId": "A"
+        }
+      ],
+      "title": "Tensor Saturation per GPU (5m avg)",
+      "description": "Which GPUs are exercising tensor cores and which are not. Sorted descending. Grouped by Hostname/gpu/UUID — DCGM metrics cannot be attributed to workload namespaces.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 12,
+        "x": 0,
+        "y": 16
+      },
+      "repeatDirection": "h",
+      "transformations": [
+        {
+          "id": "organize",
+          "options": {
+            "excludeByName": {
+              "DCGM_FI_DRIVER_VERSION": true,
+              "Time": true,
+              "__name__": true,
+              "cluster": true,
+              "container": true,
+              "device": true,
+              "endpoint": true,
+              "gpu_driver_version": true,
+              "instance": true,
+              "job": true,
+              "modelName": true,
+              "namespace": true,
+              "pci_bus_id": true,
+              "pod": true,
+              "prometheus": true,
+              "service": true,
+              "tenant": true,
+              "tier": true,
+              "uid": true,
+              "unit": true
+            },
+            "indexByName": {
+              "Hostname": 0,
+              "UUID": 2,
+              "Value": 3,
+              "gpu": 1
+            },
+            "renameByName": {
+              "Hostname": "Node",
+              "UUID": "UUID",
+              "Value": "Saturation",
+              "gpu": "GPU"
+            }
+          }
+        }
+      ],
+      "options": {
+        "frameIndex": 0,
+        "showHeader": true,
+        "showTypeIcons": false,
+        "sortBy": [
+          {
+            "displayName": "Saturation",
+            "desc": true
+          }
+        ],
+        "footer": {
+          "show": false,
+          "reducer": []
+        },
+        "cellHeight": "sm"
+      },
+      "fieldConfig": {
+        "defaults": {},
+        "overrides": [
+          {
+            "matcher": {
+              "id": "byName",
+              "options": "Saturation"
+            },
+            "properties": [
+              {
+                "id": "unit",
+                "value": "percent"
+              },
+              {
+                "id": "min",
+                "value": 0
+              },
+              {
+                "id": "max",
+                "value": 100
+              },
+              {
+                "id": "custom.cellOptions",
+                "value": {
+                  "mode": "gradient",
+                  "type": "gauge"
+                }
+              },
+              {
+                "id": "thresholds",
+                "value": {
+                  "mode": "absolute",
+                  "steps": [
+                    {
+                      "color": "red"
+                    },
+                    {
+                      "color": "orange",
+                      "value": 10
+                    },
+                    {
+                      "color": "yellow",
+                      "value": 30
+                    },
+                    {
+                      "color": "green",
+                      "value": 60
+                    }
+                  ]
+                }
+              }
+            ]
+          }
+        ]
+      }
+    },
+    {
+      "type": "table",
+      "id": 22,
+      "targets": [
+        {
+          "expr": "topk(20, gpu:util_per_watt:avg5m)",
+          "instant": true,
+          "range": false,
+          "format": "table",
+          "refId": "A"
+        }
+      ],
+      "title": "Utilization / Watt per GPU (5m avg)",
+      "description": "How efficiently each GPU spends watts. Low value = poor optimization. Grouped by Hostname/gpu/UUID — DCGM metrics cannot be attributed to workload namespaces.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 12,
+        "x": 12,
+        "y": 16
+      },
+      "repeatDirection": "h",
+      "transformations": [
+        {
+          "id": "organize",
+          "options": {
+            "excludeByName": {
+              "DCGM_FI_DRIVER_VERSION": true,
+              "Time": true,
+              "__name__": true,
+              "cluster": true,
+              "container": true,
+              "device": true,
+              "endpoint": true,
+              "gpu_driver_version": true,
+              "instance": true,
+              "job": true,
+              "modelName": true,
+              "namespace": true,
+              "pci_bus_id": true,
+              "pod": true,
+              "prometheus": true,
+              "service": true,
+              "tenant": true,
+              "tier": true,
+              "uid": true,
+              "unit": true
+            },
+            "indexByName": {
+              "Hostname": 0,
+              "UUID": 2,
+              "Value": 3,
+              "gpu": 1
+            },
+            "renameByName": {
+              "Hostname": "Node",
+              "UUID": "UUID",
+              "Value": "Util/Watt",
+              "gpu": "GPU"
+            }
+          }
+        }
+      ],
+      "options": {
+        "frameIndex": 0,
+        "showHeader": true,
+        "showTypeIcons": false,
+        "sortBy": [
+          {
+            "displayName": "Util/Watt",
+            "desc": true
+          }
+        ],
+        "footer": {
+          "show": false,
+          "reducer": []
+        },
+        "cellHeight": "sm"
+      },
+      "fieldConfig": {
+        "defaults": {},
+        "overrides": [
+          {
+            "matcher": {
+              "id": "byName",
+              "options": "Util/Watt"
+            },
+            "properties": [
+              {
+                "id": "unit",
+                "value": "none"
+              },
+              {
+                "id": "decimals",
+                "value": 3
+              },
+              {
+                "id": "min",
+                "value": 0
+              },
+              {
+                "id": "max",
+                "value": 1.5
+              },
+              {
+                "id": "custom.cellOptions",
+                "value": {
+                  "mode": "gradient",
+                  "type": "gauge"
+                }
+              },
+              {
+                "id": "thresholds",
+                "value": {
+                  "mode": "absolute",
+                  "steps": [
+                    {
+                      "color": "red"
+                    },
+                    {
+                      "color": "yellow",
+                      "value": 0.5
+                    },
+                    {
+                      "color": "green",
+                      "value": 1
+                    }
+                  ]
+                }
+              }
+            ]
+          }
+        ]
+      }
+    },
+    {
+      "type": "row",
+      "collapsed": false,
+      "title": "Throttling",
+      "gridPos": {
+        "h": 1,
+        "w": 24,
+        "x": 0,
+        "y": 24
+      },
+      "id": 30,
+      "panels": []
+    },
+    {
+      "type": "timeseries",
+      "id": 31,
+      "targets": [
+        {
+          "expr": "gpu:power_throttle_fraction:rate5m",
+          "legendFormat": "{{Hostname}}/{{gpu}}",
+          "refId": "A"
+        }
+      ],
+      "title": "Power throttle fraction per GPU",
+      "description": "Fraction of time the GPU was power-throttled. 1.0 = always throttled. Cluster-wide — rule aggregates by (Hostname, gpu, UUID) and drops namespace label.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 12,
+        "x": 0,
+        "y": 25
+      },
+      "repeatDirection": "h",
+      "options": {
+        "legend": {
+          "displayMode": "table",
+          "placement": "bottom",
+          "showLegend": false,
+          "calcs": [
+            "mean",
+            "max"
+          ]
+        },
+        "tooltip": {
+          "mode": "multi",
+          "sort": ""
+        }
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "percentunit",
+          "min": 0,
+          "max": 1,
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "value": null,
+                "color": "green"
+              },
+              {
+                "value": 0.05,
+                "color": "yellow"
+              },
+              {
+                "value": 0.2,
+                "color": "red"
+              }
+            ]
+          },
+          "custom": {
+            "drawStyle": "line",
+            "lineInterpolation": "linear",
+            "fillOpacity": 25,
+            "showPoints": "never"
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "timeseries",
+      "id": 32,
+      "targets": [
+        {
+          "expr": "gpu:thermal_throttle_fraction:rate5m",
+          "legendFormat": "{{Hostname}}/{{gpu}}",
+          "refId": "A"
+        }
+      ],
+      "title": "Thermal throttle fraction per GPU",
+      "description": "Fraction of time the GPU was thermal-throttled. Cluster-wide — rule aggregates by (Hostname, gpu, UUID) and drops namespace label.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 12,
+        "x": 12,
+        "y": 25
+      },
+      "repeatDirection": "h",
+      "options": {
+        "legend": {
+          "displayMode": "table",
+          "placement": "bottom",
+          "showLegend": false,
+          "calcs": [
+            "mean",
+            "max"
+          ]
+        },
+        "tooltip": {
+          "mode": "multi",
+          "sort": ""
+        }
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "percentunit",
+          "min": 0,
+          "max": 1,
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "value": null,
+                "color": "green"
+              },
+              {
+                "value": 0.05,
+                "color": "yellow"
+              },
+              {
+                "value": 0.2,
+                "color": "red"
+              }
+            ]
+          },
+          "custom": {
+            "drawStyle": "line",
+            "lineInterpolation": "linear",
+            "fillOpacity": 25,
+            "showPoints": "never"
+          }
+        },
+        "overrides": []
+      }
+    }
+  ],
+  "templating": {
+    "list": [
+      {
+        "type": "datasource",
+        "name": "ds_prometheus",
+        "label": "Prometheus",
+        "skipUrlSync": false,
+        "query": "prometheus",
+        "current": {
+          "selected": false,
+          "text": "default",
+          "value": "default"
+        },
+        "multi": false,
+        "allowCustomValue": true,
+        "includeAll": false,
+        "regex": "",
+        "auto": false,
+        "auto_min": "10s",
+        "auto_count": 30
+      }
+    ]
+  },
+  "annotations": {}
+}
diff --git a/dashboards/gpu/gpu-fleet.json b/dashboards/gpu/gpu-fleet.json
new file mode 100644
index 00000000..60c12bfb
--- /dev/null
+++ b/dashboards/gpu/gpu-fleet.json
@@ -0,0 +1,1157 @@
+{
+  "uid": "gpu-fleet",
+  "title": "GPU Fleet Overview",
+  "description": "Cluster-wide GPU inventory, capacity, utilization and health — admin view across the whole fleet.",
+  "tags": [
+    "gpu",
+    "fleet",
+    "admin"
+  ],
+  "timezone": "browser",
+  "editable": true,
+  "graphTooltip": 1,
+  "time": {
+    "from": "now-6h",
+    "to": "now"
+  },
+  "fiscalYearStartMonth": 0,
+  "schemaVersion": 42,
+  "panels": [
+    {
+      "type": "row",
+      "collapsed": false,
+      "title": "Cluster snapshot",
+      "gridPos": {
+        "h": 1,
+        "w": 24,
+        "x": 0,
+        "y": 0
+      },
+      "id": 1,
+      "panels": []
+    },
+    {
+      "type": "stat",
+      "id": 2,
+      "targets": [
+        {
+          "expr": "cluster:gpu_count:total",
+          "refId": "A"
+        }
+      ],
+      "title": "Total GPUs",
+      "description": "Physical GPUs discovered by DCGM across all nodes.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 4,
+        "w": 4,
+        "x": 0,
+        "y": 1
+      },
+      "repeatDirection": "h",
+      "options": {
+        "graphMode": "none",
+        "colorMode": "value",
+        "justifyMode": "auto",
+        "textMode": "value",
+        "wideLayout": true,
+        "showPercentChange": false,
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ]
+        },
+        "percentChangeColorMode": "standard",
+        "orientation": ""
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "short",
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "value": null,
+                "color": "blue"
+              }
+            ]
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "stat",
+      "id": 3,
+      "targets": [
+        {
+          "expr": "cluster:gpu_count:allocated",
+          "refId": "A"
+        }
+      ],
+      "title": "Allocated",
+      "description": "Sum of kube GPU requests across all pods. Billable view — includes Pending pods.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 4,
+        "w": 4,
+        "x": 4,
+        "y": 1
+      },
+      "repeatDirection": "h",
+      "options": {
+        "graphMode": "none",
+        "colorMode": "background",
+        "justifyMode": "auto",
+        "textMode": "value",
+        "wideLayout": true,
+        "showPercentChange": false,
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ]
+        },
+        "percentChangeColorMode": "standard",
+        "orientation": ""
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "short",
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "value": null,
+                "color": "blue"
+              },
+              {
+                "value": 1,
+                "color": "green"
+              }
+            ]
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "stat",
+      "id": 4,
+      "targets": [
+        {
+          "expr": "cluster:gpu_count:free",
+          "refId": "A"
+        }
+      ],
+      "title": "Free",
+      "description": "Unallocated capacity available for new tenants.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 4,
+        "w": 4,
+        "x": 8,
+        "y": 1
+      },
+      "repeatDirection": "h",
+      "options": {
+        "graphMode": "none",
+        "colorMode": "background",
+        "justifyMode": "auto",
+        "textMode": "value",
+        "wideLayout": true,
+        "showPercentChange": false,
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ]
+        },
+        "percentChangeColorMode": "standard",
+        "orientation": ""
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "short",
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "value": null,
+                "color": "red"
+              },
+              {
+                "value": 1,
+                "color": "yellow"
+              },
+              {
+                "value": 2,
+                "color": "green"
+              }
+            ]
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "stat",
+      "id": 5,
+      "targets": [
+        {
+          "expr": "count(count by (Hostname) (DCGM_FI_DEV_GPU_UTIL))",
+          "refId": "A"
+        }
+      ],
+      "title": "Nodes with GPU",
+      "description": "Nodes reporting at least one GPU to DCGM. Counted via DCGM_FI_DEV_GPU_UTIL because kube-state-metrics does not expose node labels by default.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 4,
+        "w": 4,
+        "x": 12,
+        "y": 1
+      },
+      "repeatDirection": "h",
+      "options": {
+        "graphMode": "none",
+        "colorMode": "value",
+        "justifyMode": "auto",
+        "textMode": "value",
+        "wideLayout": true,
+        "showPercentChange": false,
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ]
+        },
+        "percentChangeColorMode": "standard",
+        "orientation": ""
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "short",
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "value": null,
+                "color": "blue"
+              }
+            ]
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "stat",
+      "id": 6,
+      "targets": [
+        {
+          "expr": "cluster:gpu_util:avg",
+          "refId": "A"
+        }
+      ],
+      "title": "Average utilization",
+      "description": "Legacy NVML utilization across all tenant GPUs.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 4,
+        "w": 4,
+        "x": 16,
+        "y": 1
+      },
+      "repeatDirection": "h",
+      "options": {
+        "graphMode": "area",
+        "colorMode": "value",
+        "justifyMode": "auto",
+        "textMode": "value",
+        "wideLayout": true,
+        "showPercentChange": false,
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ]
+        },
+        "percentChangeColorMode": "standard",
+        "orientation": ""
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "percent",
+          "min": 0,
+          "max": 100,
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "value": null,
+                "color": "blue"
+              },
+              {
+                "value": 30,
+                "color": "green"
+              },
+              {
+                "value": 80,
+                "color": "orange"
+              }
+            ]
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "stat",
+      "id": 7,
+      "targets": [
+        {
+          "expr": "cluster:gpu_power_watts:sum",
+          "refId": "A"
+        }
+      ],
+      "title": "Total power",
+      "description": "Live power draw summed across the fleet.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 4,
+        "w": 4,
+        "x": 20,
+        "y": 1
+      },
+      "repeatDirection": "h",
+      "options": {
+        "graphMode": "area",
+        "colorMode": "value",
+        "justifyMode": "auto",
+        "textMode": "value",
+        "wideLayout": true,
+        "showPercentChange": false,
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ]
+        },
+        "percentChangeColorMode": "standard",
+        "orientation": ""
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "watt",
+          "decimals": 0,
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "value": null,
+                "color": "green"
+              }
+            ]
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "row",
+      "collapsed": false,
+      "title": "Per-node GPU activity",
+      "gridPos": {
+        "h": 1,
+        "w": 24,
+        "x": 0,
+        "y": 5
+      },
+      "id": 10,
+      "panels": []
+    },
+    {
+      "type": "bargauge",
+      "id": 11,
+      "targets": [
+        {
+          "expr": "avg by (Hostname) (DCGM_FI_DEV_GPU_UTIL)",
+          "legendFormat": "{{Hostname}}",
+          "refId": "A"
+        }
+      ],
+      "title": "GPU utilization per node",
+      "description": "Average NVML utilization per node — spot idle or saturated hosts at a glance.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 12,
+        "x": 0,
+        "y": 6
+      },
+      "repeatDirection": "h",
+      "options": {
+        "displayMode": "gradient",
+        "valueMode": "color",
+        "namePlacement": "auto",
+        "showUnfilled": true,
+        "sizing": "auto",
+        "minVizWidth": 8,
+        "minVizHeight": 16,
+        "legend": {
+          "displayMode": "list",
+          "placement": "bottom",
+          "showLegend": false,
+          "calcs": []
+        },
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ]
+        },
+        "maxVizHeight": 300,
+        "orientation": "horizontal"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "percent",
+          "min": 0,
+          "max": 100,
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "value": null,
+                "color": "blue"
+              },
+              {
+                "value": 30,
+                "color": "green"
+              },
+              {
+                "value": 80,
+                "color": "orange"
+              }
+            ]
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "bargauge",
+      "id": 12,
+      "targets": [
+        {
+          "expr": "sum by (Hostname) (DCGM_FI_DEV_POWER_USAGE) / clamp_min(sum by (Hostname) (DCGM_FI_DEV_POWER_MGMT_LIMIT), 1) * 100",
+          "legendFormat": "{{Hostname}}",
+          "refId": "A"
+        }
+      ],
+      "title": "Power draw per node (% of TDP)",
+      "description": "GPU power usage as a fraction of the node's combined TDP cap. Works across GPU SKUs without per-model thresholds. Requires DCGM_FI_DEV_POWER_MGMT_LIMIT from custom DCGM CSV (see packages/system/gpu-operator/examples/dcgm-custom-metrics.yaml).",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 12,
+        "x": 12,
+        "y": 6
+      },
+      "repeatDirection": "h",
+      "options": {
+        "displayMode": "gradient",
+        "valueMode": "color",
+        "namePlacement": "auto",
+        "showUnfilled": true,
+        "sizing": "auto",
+        "minVizWidth": 8,
+        "minVizHeight": 16,
+        "legend": {
+          "displayMode": "list",
+          "placement": "bottom",
+          "showLegend": false,
+          "calcs": []
+        },
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ]
+        },
+        "maxVizHeight": 300,
+        "orientation": "horizontal"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "percent",
+          "min": 0,
+          "max": 100,
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "value": null,
+                "color": "green"
+              },
+              {
+                "value": 60,
+                "color": "yellow"
+              },
+              {
+                "value": 80,
+                "color": "orange"
+              }
+            ]
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "row",
+      "collapsed": false,
+      "title": "Usage trends",
+      "gridPos": {
+        "h": 1,
+        "w": 24,
+        "x": 0,
+        "y": 14
+      },
+      "id": 20,
+      "panels": []
+    },
+    {
+      "type": "timeseries",
+      "id": 21,
+      "targets": [
+        {
+          "expr": "avg by (Hostname) (DCGM_FI_DEV_GPU_UTIL)",
+          "legendFormat": "{{Hostname}}",
+          "refId": "A"
+        }
+      ],
+      "title": "GPU utilization per node (trend)",
+      "description": "Per-node NVML utilization over time — shows shifting workload across hosts.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 12,
+        "x": 0,
+        "y": 15
+      },
+      "repeatDirection": "h",
+      "options": {
+        "legend": {
+          "displayMode": "table",
+          "placement": "bottom",
+          "showLegend": false,
+          "calcs": [
+            "mean",
+            "max"
+          ]
+        },
+        "tooltip": {
+          "mode": "multi",
+          "sort": ""
+        }
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "percent",
+          "min": 0,
+          "max": 100,
+          "custom": {
+            "drawStyle": "line",
+            "lineInterpolation": "linear",
+            "fillOpacity": 20,
+            "showPoints": "never"
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "timeseries",
+      "id": 22,
+      "targets": [
+        {
+          "expr": "sum by (Hostname) (DCGM_FI_DEV_POWER_USAGE)",
+          "legendFormat": "{{Hostname}}",
+          "refId": "A"
+        }
+      ],
+      "title": "Power draw per node (trend)",
+      "description": "Per-node GPU power consumption over time.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 12,
+        "x": 12,
+        "y": 15
+      },
+      "repeatDirection": "h",
+      "options": {
+        "legend": {
+          "displayMode": "table",
+          "placement": "bottom",
+          "showLegend": false,
+          "calcs": [
+            "mean",
+            "max"
+          ]
+        },
+        "tooltip": {
+          "mode": "multi",
+          "sort": ""
+        }
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "watt",
+          "custom": {
+            "drawStyle": "line",
+            "lineInterpolation": "linear",
+            "fillOpacity": 20,
+            "showPoints": "never"
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "row",
+      "collapsed": false,
+      "title": "Health",
+      "gridPos": {
+        "h": 1,
+        "w": 24,
+        "x": 0,
+        "y": 23
+      },
+      "id": 30,
+      "panels": []
+    },
+    {
+      "type": "stat",
+      "id": 31,
+      "targets": [
+        {
+          "expr": "count((max by (Hostname, gpu) (DCGM_FI_DEV_XID_ERRORS)) \u003e 0) or vector(0)",
+          "refId": "A"
+        }
+      ],
+      "title": "GPUs with XID errors",
+      "description": "Count of GPUs reporting a non-zero XID code. XID = NVIDIA hardware/driver error — any non-zero value needs investigation.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 5,
+        "w": 6,
+        "x": 0,
+        "y": 24
+      },
+      "repeatDirection": "h",
+      "options": {
+        "graphMode": "none",
+        "colorMode": "background",
+        "justifyMode": "auto",
+        "textMode": "value",
+        "wideLayout": true,
+        "showPercentChange": false,
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ]
+        },
+        "percentChangeColorMode": "standard",
+        "orientation": ""
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "short",
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "value": null,
+                "color": "green"
+              },
+              {
+                "value": 1,
+                "color": "red"
+              }
+            ]
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "stat",
+      "id": 32,
+      "targets": [
+        {
+          "expr": "avg(gpu:power_throttle_fraction:rate5m)",
+          "refId": "A"
+        }
+      ],
+      "title": "Power throttling (cluster avg)",
+      "description": "Fraction of time the fleet spent power-throttled. \u003e5% = billed FLOPS are being underdelivered.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 5,
+        "w": 6,
+        "x": 6,
+        "y": 24
+      },
+      "repeatDirection": "h",
+      "options": {
+        "graphMode": "none",
+        "colorMode": "background",
+        "justifyMode": "auto",
+        "textMode": "value",
+        "wideLayout": true,
+        "showPercentChange": false,
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ]
+        },
+        "percentChangeColorMode": "standard",
+        "orientation": ""
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "percentunit",
+          "decimals": 2,
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "value": null,
+                "color": "green"
+              },
+              {
+                "value": 0.05,
+                "color": "yellow"
+              },
+              {
+                "value": 0.2,
+                "color": "red"
+              }
+            ]
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "stat",
+      "id": 33,
+      "targets": [
+        {
+          "expr": "avg(gpu:thermal_throttle_fraction:rate5m)",
+          "refId": "A"
+        }
+      ],
+      "title": "Thermal throttling (cluster avg)",
+      "description": "Fraction of time the fleet spent thermal-throttled. Indicates cooling/datacenter issues.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 5,
+        "w": 6,
+        "x": 12,
+        "y": 24
+      },
+      "repeatDirection": "h",
+      "options": {
+        "graphMode": "none",
+        "colorMode": "background",
+        "justifyMode": "auto",
+        "textMode": "value",
+        "wideLayout": true,
+        "showPercentChange": false,
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ]
+        },
+        "percentChangeColorMode": "standard",
+        "orientation": ""
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "percentunit",
+          "decimals": 2,
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "value": null,
+                "color": "green"
+              },
+              {
+                "value": 0.05,
+                "color": "yellow"
+              },
+              {
+                "value": 0.2,
+                "color": "red"
+              }
+            ]
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "stat",
+      "id": 34,
+      "targets": [
+        {
+          "expr": "max(DCGM_FI_DEV_GPU_TEMP)",
+          "refId": "A"
+        }
+      ],
+      "title": "Max GPU temperature",
+      "description": "Hottest GPU in the fleet right now. Adjust thresholds to match your GPU model specifications.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 5,
+        "w": 6,
+        "x": 18,
+        "y": 24
+      },
+      "repeatDirection": "h",
+      "options": {
+        "graphMode": "none",
+        "colorMode": "background",
+        "justifyMode": "auto",
+        "textMode": "value",
+        "wideLayout": true,
+        "showPercentChange": false,
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ]
+        },
+        "percentChangeColorMode": "standard",
+        "orientation": ""
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "celsius",
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "value": null,
+                "color": "green"
+              },
+              {
+                "value": 75,
+                "color": "yellow"
+              },
+              {
+                "value": 85,
+                "color": "red"
+              }
+            ]
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "timeseries",
+      "id": 35,
+      "targets": [
+        {
+          "expr": "DCGM_FI_DEV_GPU_TEMP",
+          "legendFormat": "{{Hostname}}/{{gpu}}",
+          "refId": "A"
+        }
+      ],
+      "title": "Temperature per GPU",
+      "description": "Per-GPU temperature trace. Sustained red zone = cooling issue or aggressive workload.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 12,
+        "x": 0,
+        "y": 29
+      },
+      "repeatDirection": "h",
+      "options": {
+        "legend": {
+          "displayMode": "table",
+          "placement": "bottom",
+          "showLegend": false,
+          "calcs": [
+            "mean",
+            "max"
+          ]
+        },
+        "tooltip": {
+          "mode": "multi",
+          "sort": ""
+        }
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "celsius",
+          "min": 20,
+          "max": 100,
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "value": null,
+                "color": "green"
+              },
+              {
+                "value": 75,
+                "color": "yellow"
+              },
+              {
+                "value": 85,
+                "color": "red"
+              }
+            ]
+          },
+          "custom": {
+            "drawStyle": "line",
+            "lineInterpolation": "linear",
+            "fillOpacity": 10,
+            "showPoints": "never"
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "timeseries",
+      "id": 36,
+      "targets": [
+        {
+          "expr": "gpu:power_throttle_fraction:rate5m",
+          "legendFormat": "power {{Hostname}}/{{gpu}}",
+          "refId": "A"
+        },
+        {
+          "expr": "gpu:thermal_throttle_fraction:rate5m",
+          "legendFormat": "thermal {{Hostname}}/{{gpu}}",
+          "refId": "B"
+        }
+      ],
+      "title": "Throttle fraction per GPU",
+      "description": "Red = power throttle, blue = thermal throttle. Both on same chart to compare causes.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 12,
+        "x": 12,
+        "y": 29
+      },
+      "repeatDirection": "h",
+      "options": {
+        "legend": {
+          "displayMode": "table",
+          "placement": "bottom",
+          "showLegend": false,
+          "calcs": [
+            "mean",
+            "max"
+          ]
+        },
+        "tooltip": {
+          "mode": "multi",
+          "sort": ""
+        }
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "percentunit",
+          "min": 0,
+          "max": 1,
+          "custom": {
+            "drawStyle": "line",
+            "lineInterpolation": "linear",
+            "fillOpacity": 25,
+            "showPoints": "never"
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "row",
+      "collapsed": false,
+      "title": "Hardware inventory",
+      "gridPos": {
+        "h": 1,
+        "w": 24,
+        "x": 0,
+        "y": 37
+      },
+      "id": 40,
+      "panels": []
+    },
+    {
+      "type": "table",
+      "id": 41,
+      "targets": [
+        {
+          "expr": "group by (Hostname, modelName, gpu, UUID) (DCGM_FI_DEV_GPU_UTIL)",
+          "instant": true,
+          "range": false,
+          "format": "table",
+          "refId": "A"
+        }
+      ],
+      "title": "GPU inventory",
+      "description": "One row per physical GPU. Model / index / UUID sourced from DCGM labels.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 6,
+        "w": 24,
+        "x": 0,
+        "y": 38
+      },
+      "repeatDirection": "h",
+      "transformations": [
+        {
+          "id": "organize",
+          "options": {
+            "excludeByName": {
+              "Time": true,
+              "Value": true,
+              "__name__": true,
+              "cluster": true,
+              "container": true,
+              "device": true,
+              "endpoint": true,
+              "instance": true,
+              "job": true,
+              "pci_bus_id": true,
+              "prometheus": true,
+              "service": true,
+              "tenant": true,
+              "tier": true,
+              "uid": true,
+              "unit": true
+            },
+            "indexByName": {
+              "Hostname": 0,
+              "UUID": 3,
+              "gpu": 2,
+              "modelName": 1
+            },
+            "renameByName": {
+              "Hostname": "Node",
+              "UUID": "UUID",
+              "gpu": "Index",
+              "modelName": "GPU Model"
+            }
+          }
+        }
+      ],
+      "options": {
+        "frameIndex": 0,
+        "showHeader": true,
+        "showTypeIcons": false,
+        "footer": {
+          "show": false,
+          "reducer": []
+        },
+        "cellHeight": "sm"
+      }
+    }
+  ],
+  "templating": {
+    "list": [
+      {
+        "type": "datasource",
+        "name": "ds_prometheus",
+        "label": "Prometheus",
+        "skipUrlSync": false,
+        "query": "prometheus",
+        "current": {
+          "selected": false,
+          "text": "default",
+          "value": "default"
+        },
+        "multi": false,
+        "allowCustomValue": true,
+        "includeAll": false,
+        "regex": "",
+        "auto": false,
+        "auto_min": "10s",
+        "auto_count": 30
+      }
+    ]
+  },
+  "annotations": {}
+}
diff --git a/dashboards/gpu/gpu-performance.json b/dashboards/gpu/gpu-performance.json
new file mode 100644
index 00000000..aed9c69a
--- /dev/null
+++ b/dashboards/gpu/gpu-performance.json
@@ -0,0 +1,957 @@
+{
+  "uid": "gpu-performance",
+  "title": "GPU Performance",
+  "tags": [
+    "gpu",
+    "dcgm"
+  ],
+  "timezone": "browser",
+  "editable": true,
+  "graphTooltip": 1,
+  "time": {
+    "from": "now-1h",
+    "to": "now"
+  },
+  "fiscalYearStartMonth": 0,
+  "schemaVersion": 42,
+  "panels": [
+    {
+      "type": "row",
+      "collapsed": false,
+      "title": "Overview",
+      "gridPos": {
+        "h": 1,
+        "w": 24,
+        "x": 0,
+        "y": 0
+      },
+      "id": 1,
+      "panels": []
+    },
+    {
+      "type": "stat",
+      "id": 2,
+      "targets": [
+        {
+          "expr": "cluster:gpu_count:total",
+          "refId": "A"
+        }
+      ],
+      "title": "Total GPUs",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 4,
+        "w": 6,
+        "x": 0,
+        "y": 1
+      },
+      "repeatDirection": "h",
+      "options": {
+        "graphMode": "none",
+        "colorMode": "value",
+        "justifyMode": "auto",
+        "textMode": "value",
+        "wideLayout": true,
+        "showPercentChange": false,
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ]
+        },
+        "percentChangeColorMode": "standard",
+        "orientation": ""
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "short",
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "value": null,
+                "color": "blue"
+              }
+            ]
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "stat",
+      "id": 3,
+      "targets": [
+        {
+          "expr": "cluster:gpu_count:allocated",
+          "refId": "A"
+        }
+      ],
+      "title": "Allocated",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 4,
+        "w": 6,
+        "x": 6,
+        "y": 1
+      },
+      "repeatDirection": "h",
+      "options": {
+        "graphMode": "none",
+        "colorMode": "value",
+        "justifyMode": "auto",
+        "textMode": "value",
+        "wideLayout": true,
+        "showPercentChange": false,
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ]
+        },
+        "percentChangeColorMode": "standard",
+        "orientation": ""
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "short",
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "value": null,
+                "color": "green"
+              },
+              {
+                "value": 1,
+                "color": "yellow"
+              }
+            ]
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "stat",
+      "id": 4,
+      "targets": [
+        {
+          "expr": "cluster:gpu_util:avg",
+          "refId": "A"
+        }
+      ],
+      "title": "Average utilization",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 4,
+        "w": 6,
+        "x": 12,
+        "y": 1
+      },
+      "repeatDirection": "h",
+      "options": {
+        "graphMode": "area",
+        "colorMode": "value",
+        "justifyMode": "auto",
+        "textMode": "value",
+        "wideLayout": true,
+        "showPercentChange": false,
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ]
+        },
+        "percentChangeColorMode": "standard",
+        "orientation": ""
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "percent",
+          "min": 0,
+          "max": 100,
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "value": null,
+                "color": "blue"
+              },
+              {
+                "value": 30,
+                "color": "green"
+              },
+              {
+                "value": 80,
+                "color": "orange"
+              }
+            ]
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "stat",
+      "id": 5,
+      "targets": [
+        {
+          "expr": "cluster:gpu_power_watts:sum",
+          "refId": "A"
+        }
+      ],
+      "title": "Power draw",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 4,
+        "w": 6,
+        "x": 18,
+        "y": 1
+      },
+      "repeatDirection": "h",
+      "options": {
+        "graphMode": "area",
+        "colorMode": "value",
+        "justifyMode": "auto",
+        "textMode": "value",
+        "wideLayout": true,
+        "showPercentChange": false,
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ]
+        },
+        "percentChangeColorMode": "standard",
+        "orientation": ""
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "watt",
+          "decimals": 0,
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "value": null,
+                "color": "green"
+              }
+            ]
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "row",
+      "collapsed": false,
+      "title": "Utilization",
+      "gridPos": {
+        "h": 1,
+        "w": 24,
+        "x": 0,
+        "y": 5
+      },
+      "id": 10,
+      "panels": []
+    },
+    {
+      "type": "timeseries",
+      "id": 11,
+      "targets": [
+        {
+          "expr": "DCGM_FI_DEV_GPU_UTIL{Hostname=~\"$Hostname\"}",
+          "legendFormat": "{{Hostname}}/{{gpu}} {{pod}}",
+          "refId": "A"
+        }
+      ],
+      "title": "GPU Utilization (NVML)",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 12,
+        "x": 0,
+        "y": 6
+      },
+      "repeatDirection": "h",
+      "options": {
+        "legend": {
+          "displayMode": "table",
+          "placement": "bottom",
+          "showLegend": false,
+          "calcs": [
+            "mean",
+            "max"
+          ]
+        },
+        "tooltip": {
+          "mode": "multi",
+          "sort": ""
+        }
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "percent",
+          "min": 0,
+          "max": 100,
+          "custom": {
+            "drawStyle": "line",
+            "lineInterpolation": "linear",
+            "fillOpacity": 10,
+            "showPoints": "never",
+            "spanNulls": false
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "timeseries",
+      "id": 12,
+      "targets": [
+        {
+          "expr": "DCGM_FI_PROF_PIPE_TENSOR_ACTIVE{Hostname=~\"$Hostname\"} * 100",
+          "legendFormat": "{{Hostname}}/{{gpu}} {{pod}}",
+          "refId": "A"
+        }
+      ],
+      "title": "Tensor Pipe Active (realistic load for LLM/AI)",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 12,
+        "x": 12,
+        "y": 6
+      },
+      "repeatDirection": "h",
+      "options": {
+        "legend": {
+          "displayMode": "table",
+          "placement": "bottom",
+          "showLegend": false,
+          "calcs": [
+            "mean",
+            "max"
+          ]
+        },
+        "tooltip": {
+          "mode": "multi",
+          "sort": ""
+        }
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "percent",
+          "min": 0,
+          "max": 100,
+          "custom": {
+            "drawStyle": "line",
+            "lineInterpolation": "linear",
+            "fillOpacity": 10,
+            "showPoints": "never",
+            "spanNulls": false
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "timeseries",
+      "id": 13,
+      "targets": [
+        {
+          "expr": "DCGM_FI_PROF_GR_ENGINE_ACTIVE{Hostname=~\"$Hostname\"} * 100",
+          "legendFormat": "{{Hostname}}/{{gpu}} {{pod}}",
+          "refId": "A"
+        }
+      ],
+      "title": "Graphics Engine Active",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 12,
+        "x": 0,
+        "y": 14
+      },
+      "repeatDirection": "h",
+      "options": {
+        "legend": {
+          "displayMode": "table",
+          "placement": "bottom",
+          "showLegend": false,
+          "calcs": [
+            "mean",
+            "max"
+          ]
+        },
+        "tooltip": {
+          "mode": "multi",
+          "sort": ""
+        }
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "percent",
+          "min": 0,
+          "max": 100,
+          "custom": {
+            "drawStyle": "line",
+            "lineInterpolation": "linear",
+            "fillOpacity": 10,
+            "showPoints": "never",
+            "spanNulls": false
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "timeseries",
+      "id": 14,
+      "targets": [
+        {
+          "expr": "DCGM_FI_DEV_MEM_COPY_UTIL{Hostname=~\"$Hostname\"}",
+          "legendFormat": "{{Hostname}}/{{gpu}} {{pod}}",
+          "refId": "A"
+        }
+      ],
+      "title": "Memory Copy Utilization",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 12,
+        "x": 12,
+        "y": 14
+      },
+      "repeatDirection": "h",
+      "options": {
+        "legend": {
+          "displayMode": "table",
+          "placement": "bottom",
+          "showLegend": false,
+          "calcs": [
+            "mean",
+            "max"
+          ]
+        },
+        "tooltip": {
+          "mode": "multi",
+          "sort": ""
+        }
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "percent",
+          "min": 0,
+          "max": 100,
+          "custom": {
+            "drawStyle": "line",
+            "lineInterpolation": "linear",
+            "fillOpacity": 10,
+            "showPoints": "never",
+            "spanNulls": false
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "row",
+      "collapsed": false,
+      "title": "Memory",
+      "gridPos": {
+        "h": 1,
+        "w": 24,
+        "x": 0,
+        "y": 22
+      },
+      "id": 20,
+      "panels": []
+    },
+    {
+      "type": "timeseries",
+      "id": 21,
+      "targets": [
+        {
+          "expr": "DCGM_FI_DEV_FB_USED{Hostname=~\"$Hostname\"} * 1048576",
+          "legendFormat": "{{Hostname}}/{{gpu}} {{pod}}",
+          "refId": "A"
+        }
+      ],
+      "title": "VRAM Used",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 12,
+        "x": 0,
+        "y": 23
+      },
+      "repeatDirection": "h",
+      "options": {
+        "legend": {
+          "displayMode": "table",
+          "placement": "bottom",
+          "showLegend": false,
+          "calcs": [
+            "mean",
+            "max"
+          ]
+        },
+        "tooltip": {
+          "mode": "multi",
+          "sort": ""
+        }
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "bytes",
+          "custom": {
+            "drawStyle": "line",
+            "lineInterpolation": "linear",
+            "fillOpacity": 25,
+            "showPoints": "never"
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "timeseries",
+      "id": 22,
+      "targets": [
+        {
+          "expr": "DCGM_FI_DEV_FB_FREE{Hostname=~\"$Hostname\"} * 1048576",
+          "legendFormat": "{{Hostname}}/{{gpu}}",
+          "refId": "A"
+        }
+      ],
+      "title": "VRAM Free",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 12,
+        "x": 12,
+        "y": 23
+      },
+      "repeatDirection": "h",
+      "options": {
+        "legend": {
+          "displayMode": "table",
+          "placement": "bottom",
+          "showLegend": false,
+          "calcs": [
+            "mean",
+            "min"
+          ]
+        },
+        "tooltip": {
+          "mode": "multi",
+          "sort": ""
+        }
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "bytes",
+          "custom": {
+            "drawStyle": "line",
+            "lineInterpolation": "linear",
+            "fillOpacity": 10,
+            "showPoints": "never"
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "row",
+      "collapsed": false,
+      "title": "Power \u0026 Temperature",
+      "gridPos": {
+        "h": 1,
+        "w": 24,
+        "x": 0,
+        "y": 31
+      },
+      "id": 30,
+      "panels": []
+    },
+    {
+      "type": "timeseries",
+      "id": 31,
+      "targets": [
+        {
+          "expr": "DCGM_FI_DEV_POWER_USAGE{Hostname=~\"$Hostname\"}",
+          "legendFormat": "{{Hostname}}/{{gpu}}",
+          "refId": "A"
+        }
+      ],
+      "title": "Power Usage per GPU",
+      "description": "Hardware-level metric — shown cluster-wide regardless of namespace filter.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 12,
+        "x": 0,
+        "y": 32
+      },
+      "repeatDirection": "h",
+      "options": {
+        "legend": {
+          "displayMode": "table",
+          "placement": "bottom",
+          "showLegend": false,
+          "calcs": [
+            "mean",
+            "max"
+          ]
+        },
+        "tooltip": {
+          "mode": "multi",
+          "sort": ""
+        }
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "watt",
+          "custom": {
+            "drawStyle": "line",
+            "lineInterpolation": "linear",
+            "fillOpacity": 10,
+            "showPoints": "never"
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "timeseries",
+      "id": 32,
+      "targets": [
+        {
+          "expr": "DCGM_FI_DEV_GPU_TEMP{Hostname=~\"$Hostname\"}",
+          "legendFormat": "{{Hostname}}/{{gpu}}",
+          "refId": "A"
+        }
+      ],
+      "title": "GPU Temperature",
+      "description": "Hardware-level metric — shown cluster-wide regardless of namespace filter.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 12,
+        "x": 12,
+        "y": 32
+      },
+      "repeatDirection": "h",
+      "options": {
+        "legend": {
+          "displayMode": "table",
+          "placement": "bottom",
+          "showLegend": false,
+          "calcs": [
+            "mean",
+            "max"
+          ]
+        },
+        "tooltip": {
+          "mode": "multi",
+          "sort": ""
+        }
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "celsius",
+          "min": 20,
+          "max": 100,
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "value": null,
+                "color": "green"
+              },
+              {
+                "value": 75,
+                "color": "yellow"
+              },
+              {
+                "value": 85,
+                "color": "red"
+              }
+            ]
+          },
+          "custom": {
+            "drawStyle": "line",
+            "lineInterpolation": "linear",
+            "fillOpacity": 10,
+            "showPoints": "never"
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "row",
+      "collapsed": false,
+      "title": "Health",
+      "gridPos": {
+        "h": 1,
+        "w": 24,
+        "x": 0,
+        "y": 40
+      },
+      "id": 40,
+      "panels": []
+    },
+    {
+      "type": "stat",
+      "id": 41,
+      "targets": [
+        {
+          "expr": "max by (Hostname, gpu) (DCGM_FI_DEV_XID_ERRORS{Hostname=~\"$Hostname\"})",
+          "legendFormat": "{{Hostname}}/{{gpu}}",
+          "refId": "A"
+        }
+      ],
+      "title": "XID errors (latest)",
+      "description": "Hardware-level metric — shown cluster-wide regardless of namespace filter.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 5,
+        "w": 8,
+        "x": 0,
+        "y": 41
+      },
+      "repeatDirection": "h",
+      "options": {
+        "graphMode": "none",
+        "colorMode": "background",
+        "justifyMode": "auto",
+        "textMode": "value_and_name",
+        "wideLayout": true,
+        "showPercentChange": false,
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ]
+        },
+        "percentChangeColorMode": "standard",
+        "orientation": ""
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "short",
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "value": null,
+                "color": "green"
+              },
+              {
+                "value": 1,
+                "color": "red"
+              }
+            ]
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "timeseries",
+      "id": 42,
+      "targets": [
+        {
+          "expr": "rate(DCGM_FI_DEV_POWER_VIOLATION{Hostname=~\"$Hostname\"}[5m])",
+          "legendFormat": "{{Hostname}}/{{gpu}}",
+          "refId": "A"
+        }
+      ],
+      "title": "Power Violation (µs/s throttled due to power)",
+      "description": "Hardware-level metric — shown cluster-wide regardless of namespace filter.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 5,
+        "w": 8,
+        "x": 8,
+        "y": 41
+      },
+      "repeatDirection": "h",
+      "options": {
+        "legend": {
+          "displayMode": "list",
+          "placement": "bottom",
+          "showLegend": false,
+          "calcs": []
+        },
+        "tooltip": {
+          "mode": "multi",
+          "sort": ""
+        }
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "µs",
+          "custom": {
+            "drawStyle": "line",
+            "lineInterpolation": "linear",
+            "fillOpacity": 10,
+            "showPoints": "never"
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "timeseries",
+      "id": 43,
+      "targets": [
+        {
+          "expr": "rate(DCGM_FI_DEV_THERMAL_VIOLATION{Hostname=~\"$Hostname\"}[5m])",
+          "legendFormat": "{{Hostname}}/{{gpu}}",
+          "refId": "A"
+        }
+      ],
+      "title": "Thermal Violation (µs/s throttled due to thermals)",
+      "description": "Hardware-level metric — shown cluster-wide regardless of namespace filter.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 5,
+        "w": 8,
+        "x": 16,
+        "y": 41
+      },
+      "repeatDirection": "h",
+      "options": {
+        "legend": {
+          "displayMode": "list",
+          "placement": "bottom",
+          "showLegend": false,
+          "calcs": []
+        },
+        "tooltip": {
+          "mode": "multi",
+          "sort": ""
+        }
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "µs",
+          "custom": {
+            "drawStyle": "line",
+            "lineInterpolation": "linear",
+            "fillOpacity": 10,
+            "showPoints": "never"
+          }
+        },
+        "overrides": []
+      }
+    }
+  ],
+  "templating": {
+    "list": [
+      {
+        "type": "datasource",
+        "name": "ds_prometheus",
+        "label": "Prometheus",
+        "skipUrlSync": false,
+        "query": "prometheus",
+        "current": {
+          "selected": false,
+          "text": "default",
+          "value": "default"
+        },
+        "multi": false,
+        "allowCustomValue": true,
+        "includeAll": false,
+        "regex": "",
+        "auto": false,
+        "auto_min": "10s",
+        "auto_count": 30
+      },
+      {
+        "type": "query",
+        "name": "Hostname",
+        "label": "Host",
+        "skipUrlSync": false,
+        "query": "label_values(DCGM_FI_DEV_GPU_UTIL, Hostname)",
+        "datasource": {
+          "type": "prometheus",
+          "uid": "$ds_prometheus"
+        },
+        "current": {
+          "selected": true,
+          "text": "All",
+          "value": "$__all"
+        },
+        "multi": true,
+        "allowCustomValue": true,
+        "refresh": 2,
+        "sort": 1,
+        "includeAll": true,
+        "auto": false,
+        "auto_min": "10s",
+        "auto_count": 30
+      }
+    ]
+  },
+  "annotations": {}
+}
diff --git a/dashboards/gpu/gpu-quotas.json b/dashboards/gpu/gpu-quotas.json
new file mode 100644
index 00000000..865fb95c
--- /dev/null
+++ b/dashboards/gpu/gpu-quotas.json
@@ -0,0 +1,637 @@
+{
+  "uid": "gpu-quotas",
+  "title": "GPU Quotas \u0026 Allocation",
+  "tags": [
+    "gpu",
+    "quotas"
+  ],
+  "timezone": "browser",
+  "editable": true,
+  "graphTooltip": 1,
+  "time": {
+    "from": "now-6h",
+    "to": "now"
+  },
+  "fiscalYearStartMonth": 0,
+  "schemaVersion": 42,
+  "panels": [
+    {
+      "type": "row",
+      "collapsed": false,
+      "title": "Allocation overview",
+      "gridPos": {
+        "h": 1,
+        "w": 24,
+        "x": 0,
+        "y": 0
+      },
+      "id": 1,
+      "panels": []
+    },
+    {
+      "type": "stat",
+      "id": 2,
+      "targets": [
+        {
+          "expr": "sum(kube_node_status_allocatable{resource=\"nvidia_com_gpu\"})",
+          "refId": "A"
+        }
+      ],
+      "title": "GPU allocatable",
+      "description": "Total GPU capacity the cluster can schedule to pods.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 4,
+        "w": 6,
+        "x": 0,
+        "y": 1
+      },
+      "repeatDirection": "h",
+      "options": {
+        "graphMode": "none",
+        "colorMode": "value",
+        "justifyMode": "auto",
+        "textMode": "value",
+        "wideLayout": true,
+        "showPercentChange": false,
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ]
+        },
+        "percentChangeColorMode": "standard",
+        "orientation": ""
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "short",
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "value": null,
+                "color": "blue"
+              }
+            ]
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "stat",
+      "id": 3,
+      "targets": [
+        {
+          "expr": "cluster:gpu_count:allocated",
+          "refId": "A"
+        }
+      ],
+      "title": "GPU requested",
+      "description": "Sum of GPU requests across all pods cluster-wide, including system namespaces.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 4,
+        "w": 6,
+        "x": 6,
+        "y": 1
+      },
+      "repeatDirection": "h",
+      "options": {
+        "graphMode": "none",
+        "colorMode": "value",
+        "justifyMode": "auto",
+        "textMode": "value",
+        "wideLayout": true,
+        "showPercentChange": false,
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ]
+        },
+        "percentChangeColorMode": "standard",
+        "orientation": ""
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "short",
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "value": null,
+                "color": "green"
+              },
+              {
+                "value": 1,
+                "color": "yellow"
+              }
+            ]
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "gauge",
+      "id": 4,
+      "targets": [
+        {
+          "expr": "cluster:gpu_count:allocated / sum(kube_node_status_allocatable{resource=\"nvidia_com_gpu\"}) * 100",
+          "refId": "A"
+        }
+      ],
+      "title": "Allocation ratio",
+      "description": "Percentage of allocatable GPUs currently requested by pods.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 4,
+        "w": 6,
+        "x": 12,
+        "y": 1
+      },
+      "repeatDirection": "h",
+      "options": {
+        "showThresholdLabels": false,
+        "showThresholdMarkers": true,
+        "sizing": "auto",
+        "minVizWidth": 75,
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ]
+        },
+        "minVizHeight": 75,
+        "orientation": ""
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "percent",
+          "min": 0,
+          "max": 100,
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "value": null,
+                "color": "blue"
+              },
+              {
+                "value": 40,
+                "color": "green"
+              },
+              {
+                "value": 80,
+                "color": "yellow"
+              },
+              {
+                "value": 95,
+                "color": "red"
+              }
+            ]
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "stat",
+      "id": 5,
+      "targets": [
+        {
+          "expr": "count((sum by (namespace, pod) (kube_pod_container_resource_requests{resource=\"nvidia_com_gpu\"}) \u003e 0) * on(namespace, pod) group_left() (kube_pod_status_phase{phase=\"Pending\"} == 1)) or vector(0)",
+          "refId": "A"
+        }
+      ],
+      "title": "Pending pods (GPU)",
+      "description": "Pods requesting GPUs that are stuck in Pending state — indicates capacity shortage.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 4,
+        "w": 6,
+        "x": 18,
+        "y": 1
+      },
+      "repeatDirection": "h",
+      "options": {
+        "graphMode": "none",
+        "colorMode": "background",
+        "justifyMode": "auto",
+        "textMode": "value",
+        "wideLayout": true,
+        "showPercentChange": false,
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ]
+        },
+        "percentChangeColorMode": "standard",
+        "orientation": ""
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "short",
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "value": null,
+                "color": "green"
+              },
+              {
+                "value": 1,
+                "color": "red"
+              }
+            ]
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "row",
+      "collapsed": false,
+      "title": "Per namespace",
+      "gridPos": {
+        "h": 1,
+        "w": 24,
+        "x": 0,
+        "y": 5
+      },
+      "id": 10,
+      "panels": []
+    },
+    {
+      "type": "bargauge",
+      "id": 11,
+      "targets": [
+        {
+          "expr": "sum by (namespace) (namespace:gpu_count:allocated{namespace=~\"$namespace\"})",
+          "legendFormat": "{{namespace}}",
+          "refId": "A"
+        }
+      ],
+      "title": "GPU requested per namespace",
+      "description": "GPU allocation breakdown by namespace — spot top consumers at a glance.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 12,
+        "x": 0,
+        "y": 6
+      },
+      "repeatDirection": "h",
+      "options": {
+        "displayMode": "gradient",
+        "valueMode": "color",
+        "namePlacement": "auto",
+        "showUnfilled": true,
+        "sizing": "auto",
+        "minVizWidth": 8,
+        "minVizHeight": 16,
+        "legend": {
+          "displayMode": "list",
+          "placement": "bottom",
+          "showLegend": false,
+          "calcs": []
+        },
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ]
+        },
+        "maxVizHeight": 300,
+        "orientation": "horizontal"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "short",
+          "min": 0,
+          "max": 8,
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "value": null,
+                "color": "green"
+              }
+            ]
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "timeseries",
+      "id": 12,
+      "targets": [
+        {
+          "expr": "sum(kube_node_status_allocatable{resource=\"nvidia_com_gpu\"})",
+          "legendFormat": "Allocatable (total)",
+          "refId": "A"
+        },
+        {
+          "expr": "sum(namespace:gpu_count:allocated{namespace=~\"$namespace\"})",
+          "legendFormat": "Requested",
+          "refId": "B"
+        }
+      ],
+      "title": "GPU allocated over time",
+      "description": "Requested vs allocatable GPUs over time — shows allocation pressure trends.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 12,
+        "x": 12,
+        "y": 6
+      },
+      "repeatDirection": "h",
+      "options": {
+        "legend": {
+          "displayMode": "list",
+          "placement": "bottom",
+          "showLegend": false,
+          "calcs": []
+        },
+        "tooltip": {
+          "mode": "multi",
+          "sort": ""
+        }
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "short",
+          "custom": {
+            "drawStyle": "line",
+            "lineInterpolation": "stepAfter",
+            "fillOpacity": 10,
+            "showPoints": "never"
+          }
+        },
+        "overrides": [
+          {
+            "matcher": {
+              "id": "byName",
+              "options": "Allocatable (total)"
+            },
+            "properties": [
+              {
+                "id": "color",
+                "value": {
+                  "fixedColor": "blue",
+                  "mode": "fixed"
+                }
+              }
+            ]
+          }
+        ]
+      }
+    },
+    {
+      "type": "row",
+      "collapsed": false,
+      "title": "Pods with GPU",
+      "gridPos": {
+        "h": 1,
+        "w": 24,
+        "x": 0,
+        "y": 14
+      },
+      "id": 20,
+      "panels": []
+    },
+    {
+      "type": "table",
+      "id": 21,
+      "targets": [
+        {
+          "expr": "kube_pod_container_resource_requests{resource=\"nvidia_com_gpu\", namespace=~\"$namespace\"} * on(namespace, pod) group_left(phase) (kube_pod_status_phase{phase=~\"Running|Pending\"} == 1)",
+          "instant": true,
+          "range": false,
+          "format": "table",
+          "refId": "requested"
+        },
+        {
+          "expr": "kube_pod_container_resource_limits{resource=\"nvidia_com_gpu\", namespace=~\"$namespace\"} * on(namespace, pod) group_left() (kube_pod_status_phase{phase=~\"Running|Pending\"} == 1)",
+          "instant": true,
+          "range": false,
+          "format": "table",
+          "refId": "limits"
+        }
+      ],
+      "title": "Pods requesting GPU",
+      "description": "Per-pod GPU requests and limits with scheduling status — Running or Pending.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 10,
+        "w": 24,
+        "x": 0,
+        "y": 15
+      },
+      "repeatDirection": "h",
+      "transformations": [
+        {
+          "id": "joinByField",
+          "options": {
+            "byField": "pod",
+            "mode": "outer"
+          }
+        },
+        {
+          "id": "organize",
+          "options": {
+            "excludeByName": {
+              "Time": true,
+              "Time 1": true,
+              "Time 2": true,
+              "__name__": true,
+              "__name__ 1": true,
+              "__name__ 2": true,
+              "cluster": true,
+              "cluster 1": true,
+              "cluster 2": true,
+              "container 2": true,
+              "endpoint": true,
+              "endpoint 1": true,
+              "endpoint 2": true,
+              "instance": true,
+              "instance 1": true,
+              "instance 2": true,
+              "job": true,
+              "job 1": true,
+              "job 2": true,
+              "namespace 2": true,
+              "node 2": true,
+              "prometheus": true,
+              "prometheus 1": true,
+              "prometheus 2": true,
+              "resource": true,
+              "resource 1": true,
+              "resource 2": true,
+              "service": true,
+              "service 1": true,
+              "service 2": true,
+              "tenant": true,
+              "tenant 1": true,
+              "tenant 2": true,
+              "tier": true,
+              "tier 1": true,
+              "tier 2": true,
+              "uid": true,
+              "uid 1": true,
+              "uid 2": true,
+              "unit": true,
+              "unit 1": true,
+              "unit 2": true
+            },
+            "indexByName": {
+              "Value #limits": 5,
+              "Value #requested": 4,
+              "container 1": 2,
+              "namespace 1": 0,
+              "node 1": 3,
+              "phase": 6,
+              "pod": 1
+            },
+            "renameByName": {
+              "Value #limits": "Limit",
+              "Value #requested": "Req",
+              "container 1": "Container",
+              "namespace 1": "Namespace",
+              "node 1": "Node",
+              "phase": "Status"
+            }
+          }
+        }
+      ],
+      "options": {
+        "frameIndex": 0,
+        "showHeader": true,
+        "showTypeIcons": false,
+        "footer": {
+          "show": false,
+          "reducer": []
+        },
+        "cellHeight": "sm"
+      },
+      "fieldConfig": {
+        "defaults": {},
+        "overrides": [
+          {
+            "matcher": {
+              "id": "byName",
+              "options": "Status"
+            },
+            "properties": [
+              {
+                "id": "custom.cellOptions",
+                "value": {
+                  "mode": "basic",
+                  "type": "color-background"
+                }
+              },
+              {
+                "id": "mappings",
+                "value": [
+                  {
+                    "options": {
+                      "Failed": {
+                        "color": "red",
+                        "index": 2
+                      },
+                      "Pending": {
+                        "color": "orange",
+                        "index": 1
+                      },
+                      "Running": {
+                        "color": "green",
+                        "index": 0
+                      }
+                    },
+                    "type": "value"
+                  }
+                ]
+              }
+            ]
+          }
+        ]
+      }
+    }
+  ],
+  "templating": {
+    "list": [
+      {
+        "type": "datasource",
+        "name": "ds_prometheus",
+        "label": "Prometheus",
+        "skipUrlSync": false,
+        "query": "prometheus",
+        "current": {
+          "selected": false,
+          "text": "default",
+          "value": "default"
+        },
+        "multi": false,
+        "allowCustomValue": true,
+        "includeAll": false,
+        "regex": "",
+        "auto": false,
+        "auto_min": "10s",
+        "auto_count": 30
+      },
+      {
+        "type": "query",
+        "name": "namespace",
+        "label": "Namespace",
+        "skipUrlSync": false,
+        "query": "label_values(kube_pod_container_resource_requests{resource=\"nvidia_com_gpu\"}, namespace)",
+        "datasource": {
+          "type": "prometheus",
+          "uid": "$ds_prometheus"
+        },
+        "current": {
+          "selected": true,
+          "text": "All",
+          "value": "$__all"
+        },
+        "multi": true,
+        "allowCustomValue": true,
+        "refresh": 2,
+        "sort": 1,
+        "includeAll": true,
+        "auto": false,
+        "auto_min": "10s",
+        "auto_count": 30
+      }
+    ]
+  },
+  "annotations": {}
+}
diff --git a/dashboards/gpu/gpu-tenants.json b/dashboards/gpu/gpu-tenants.json
new file mode 100644
index 00000000..7fef321d
--- /dev/null
+++ b/dashboards/gpu/gpu-tenants.json
@@ -0,0 +1,1098 @@
+{
+  "uid": "gpu-tenants",
+  "title": "GPU Tenant Usage",
+  "description": "Per-tenant GPU consumption: live allocation, utilization, tensor saturation, power, and 24h GPU-hours / kWh — admin view across all namespaces.",
+  "tags": [
+    "gpu",
+    "tenants",
+    "admin"
+  ],
+  "timezone": "browser",
+  "editable": true,
+  "graphTooltip": 1,
+  "time": {
+    "from": "now-24h",
+    "to": "now"
+  },
+  "fiscalYearStartMonth": 0,
+  "schemaVersion": 42,
+  "panels": [
+    {
+      "type": "row",
+      "collapsed": false,
+      "title": "Tenant fleet snapshot",
+      "gridPos": {
+        "h": 1,
+        "w": 24,
+        "x": 0,
+        "y": 0
+      },
+      "id": 1,
+      "panels": []
+    },
+    {
+      "type": "stat",
+      "id": 2,
+      "targets": [
+        {
+          "expr": "count(namespace:gpu_count:allocated{namespace=~\"$namespace\"} \u003e 0)",
+          "refId": "A"
+        }
+      ],
+      "title": "Active tenants",
+      "description": "Namespaces with at least one GPU request (Pending + Running).",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 4,
+        "w": 4,
+        "x": 0,
+        "y": 1
+      },
+      "repeatDirection": "h",
+      "options": {
+        "graphMode": "none",
+        "colorMode": "value",
+        "justifyMode": "auto",
+        "textMode": "value",
+        "wideLayout": true,
+        "showPercentChange": false,
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ]
+        },
+        "percentChangeColorMode": "standard",
+        "orientation": ""
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "short",
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "value": null,
+                "color": "blue"
+              }
+            ]
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "stat",
+      "id": 3,
+      "targets": [
+        {
+          "expr": "sum(namespace:gpu_count:allocated{namespace=~\"$namespace\"})",
+          "refId": "A"
+        }
+      ],
+      "title": "Allocated GPUs",
+      "description": "Sum of kube GPU requests across tenants. Billable view — includes Pending pods.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 4,
+        "w": 4,
+        "x": 4,
+        "y": 1
+      },
+      "repeatDirection": "h",
+      "options": {
+        "graphMode": "none",
+        "colorMode": "background",
+        "justifyMode": "auto",
+        "textMode": "value",
+        "wideLayout": true,
+        "showPercentChange": false,
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ]
+        },
+        "percentChangeColorMode": "standard",
+        "orientation": ""
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "short",
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "value": null,
+                "color": "blue"
+              },
+              {
+                "value": 1,
+                "color": "green"
+              }
+            ]
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "stat",
+      "id": 4,
+      "targets": [
+        {
+          "expr": "cluster:gpu_util:avg",
+          "refId": "A"
+        }
+      ],
+      "title": "Avg cluster GPU util",
+      "description": "Average NVML utilization across all GPUs in the cluster. Cluster-wide — DCGM hardware metrics cannot be attributed to workload namespaces.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 4,
+        "w": 4,
+        "x": 8,
+        "y": 1
+      },
+      "repeatDirection": "h",
+      "options": {
+        "graphMode": "area",
+        "colorMode": "value",
+        "justifyMode": "auto",
+        "textMode": "value",
+        "wideLayout": true,
+        "showPercentChange": false,
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ]
+        },
+        "percentChangeColorMode": "standard",
+        "orientation": ""
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "percent",
+          "min": 0,
+          "max": 100,
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "value": null,
+                "color": "blue"
+              },
+              {
+                "value": 30,
+                "color": "green"
+              },
+              {
+                "value": 80,
+                "color": "orange"
+              }
+            ]
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "stat",
+      "id": 5,
+      "targets": [
+        {
+          "expr": "cluster:gpu_power_watts:sum",
+          "refId": "A"
+        }
+      ],
+      "title": "Total GPU power",
+      "description": "Live power draw summed across all GPUs in the cluster. Cluster-wide — DCGM hardware metrics cannot be attributed to workload namespaces.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 4,
+        "w": 4,
+        "x": 12,
+        "y": 1
+      },
+      "repeatDirection": "h",
+      "options": {
+        "graphMode": "area",
+        "colorMode": "value",
+        "justifyMode": "auto",
+        "textMode": "value",
+        "wideLayout": true,
+        "showPercentChange": false,
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ]
+        },
+        "percentChangeColorMode": "standard",
+        "orientation": ""
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "watt",
+          "decimals": 0,
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "value": null,
+                "color": "green"
+              }
+            ]
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "stat",
+      "id": 6,
+      "targets": [
+        {
+          "expr": "sum(sum_over_time(node:power_watts:sum[24h:1m])) / 60 / 1000",
+          "refId": "A"
+        }
+      ],
+      "title": "Energy (last 24h)",
+      "description": "Integrated cluster-wide GPU power draw over 24h. Cluster-wide — DCGM hardware metrics cannot be attributed to workload namespaces.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 4,
+        "w": 4,
+        "x": 16,
+        "y": 1
+      },
+      "repeatDirection": "h",
+      "options": {
+        "graphMode": "none",
+        "colorMode": "value",
+        "justifyMode": "auto",
+        "textMode": "value",
+        "wideLayout": true,
+        "showPercentChange": false,
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ]
+        },
+        "percentChangeColorMode": "standard",
+        "orientation": ""
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "kwatth",
+          "decimals": 1,
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "value": null,
+                "color": "green"
+              }
+            ]
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "stat",
+      "id": 7,
+      "targets": [
+        {
+          "expr": "sum(sum_over_time(namespace:gpu_count:allocated{namespace=~\"$namespace\"}[24h:1m])) / 60",
+          "refId": "A"
+        }
+      ],
+      "title": "GPU-hours (last 24h)",
+      "description": "Sum of (GPU count × time requested) across all tenants over the last 24h. Integrated from kube requests, so Pending pods count.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 4,
+        "w": 4,
+        "x": 20,
+        "y": 1
+      },
+      "repeatDirection": "h",
+      "options": {
+        "graphMode": "none",
+        "colorMode": "value",
+        "justifyMode": "auto",
+        "textMode": "value",
+        "wideLayout": true,
+        "showPercentChange": false,
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ]
+        },
+        "percentChangeColorMode": "standard",
+        "orientation": ""
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "short",
+          "decimals": 1,
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "value": null,
+                "color": "blue"
+              }
+            ]
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "row",
+      "collapsed": false,
+      "title": "Current allocation",
+      "gridPos": {
+        "h": 1,
+        "w": 24,
+        "x": 0,
+        "y": 5
+      },
+      "id": 10,
+      "panels": []
+    },
+    {
+      "type": "bargauge",
+      "id": 11,
+      "targets": [
+        {
+          "expr": "namespace:gpu_count:allocated{namespace=~\"$namespace\"}",
+          "legendFormat": "{{namespace}}",
+          "refId": "A"
+        }
+      ],
+      "title": "GPUs per namespace",
+      "description": "Kube GPU requests per tenant (billable). Includes Pending pods.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 12,
+        "x": 0,
+        "y": 6
+      },
+      "repeatDirection": "h",
+      "options": {
+        "displayMode": "gradient",
+        "valueMode": "color",
+        "namePlacement": "auto",
+        "showUnfilled": true,
+        "sizing": "auto",
+        "minVizWidth": 8,
+        "minVizHeight": 16,
+        "legend": {
+          "displayMode": "list",
+          "placement": "bottom",
+          "showLegend": false,
+          "calcs": []
+        },
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ]
+        },
+        "maxVizHeight": 300,
+        "orientation": "horizontal"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "short",
+          "min": 0,
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "value": null,
+                "color": "green"
+              }
+            ]
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "bargauge",
+      "id": 12,
+      "targets": [
+        {
+          "expr": "node:fb_used_bytes:sum",
+          "legendFormat": "{{Hostname}}",
+          "refId": "A"
+        }
+      ],
+      "title": "VRAM used per node",
+      "description": "FB memory actively used per node. DCGM hardware metrics cannot be attributed to workload namespaces — shown per-node instead.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 12,
+        "x": 12,
+        "y": 6
+      },
+      "repeatDirection": "h",
+      "options": {
+        "displayMode": "gradient",
+        "valueMode": "color",
+        "namePlacement": "auto",
+        "showUnfilled": true,
+        "sizing": "auto",
+        "minVizWidth": 8,
+        "minVizHeight": 16,
+        "legend": {
+          "displayMode": "list",
+          "placement": "bottom",
+          "showLegend": false,
+          "calcs": []
+        },
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ]
+        },
+        "maxVizHeight": 300,
+        "orientation": "horizontal"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "bytes",
+          "min": 0,
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "value": null,
+                "color": "green"
+              }
+            ]
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "row",
+      "collapsed": false,
+      "title": "Utilization (per-node)",
+      "gridPos": {
+        "h": 1,
+        "w": 24,
+        "x": 0,
+        "y": 14
+      },
+      "id": 20,
+      "panels": []
+    },
+    {
+      "type": "timeseries",
+      "id": 21,
+      "targets": [
+        {
+          "expr": "node:gpu_util:avg",
+          "legendFormat": "{{Hostname}}",
+          "refId": "A"
+        }
+      ],
+      "title": "NVML utilization per node",
+      "description": "Per-node NVML utilization. Stacked to show fleet-wide pressure over time. DCGM hardware metrics cannot be attributed to workload namespaces — shown per-node instead.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 12,
+        "x": 0,
+        "y": 15
+      },
+      "repeatDirection": "h",
+      "options": {
+        "legend": {
+          "displayMode": "table",
+          "placement": "bottom",
+          "showLegend": false,
+          "calcs": [
+            "mean",
+            "max"
+          ]
+        },
+        "tooltip": {
+          "mode": "multi",
+          "sort": ""
+        }
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "percent",
+          "min": 0,
+          "custom": {
+            "drawStyle": "line",
+            "lineInterpolation": "linear",
+            "fillOpacity": 40,
+            "showPoints": "never",
+            "stacking": {
+              "mode": "normal"
+            }
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "timeseries",
+      "id": 22,
+      "targets": [
+        {
+          "expr": "node:tensor_active:avg * 100",
+          "legendFormat": "{{Hostname}}",
+          "refId": "A"
+        }
+      ],
+      "title": "Tensor saturation per node",
+      "description": "Real tensor core load per node. Useful to spot nodes with idle tensor cores. DCGM hardware metrics cannot be attributed to workload namespaces — shown per-node instead.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 12,
+        "x": 12,
+        "y": 15
+      },
+      "repeatDirection": "h",
+      "options": {
+        "legend": {
+          "displayMode": "table",
+          "placement": "bottom",
+          "showLegend": false,
+          "calcs": [
+            "mean",
+            "max"
+          ]
+        },
+        "tooltip": {
+          "mode": "multi",
+          "sort": ""
+        }
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "percent",
+          "min": 0,
+          "custom": {
+            "drawStyle": "line",
+            "lineInterpolation": "linear",
+            "fillOpacity": 40,
+            "showPoints": "never",
+            "stacking": {
+              "mode": "normal"
+            }
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "row",
+      "collapsed": false,
+      "title": "Power \u0026 allocation trend",
+      "gridPos": {
+        "h": 1,
+        "w": 24,
+        "x": 0,
+        "y": 23
+      },
+      "id": 30,
+      "panels": []
+    },
+    {
+      "type": "timeseries",
+      "id": 31,
+      "targets": [
+        {
+          "expr": "node:power_watts:sum",
+          "legendFormat": "{{Hostname}}",
+          "refId": "A"
+        }
+      ],
+      "title": "Power draw per node",
+      "description": "Per-node GPU power draw. Stacked to see cluster-wide energy burn over time. DCGM hardware metrics cannot be attributed to workload namespaces — shown per-node instead.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 12,
+        "x": 0,
+        "y": 24
+      },
+      "repeatDirection": "h",
+      "options": {
+        "legend": {
+          "displayMode": "table",
+          "placement": "bottom",
+          "showLegend": false,
+          "calcs": [
+            "mean",
+            "max"
+          ]
+        },
+        "tooltip": {
+          "mode": "multi",
+          "sort": ""
+        }
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "watt",
+          "min": 0,
+          "custom": {
+            "drawStyle": "line",
+            "lineInterpolation": "linear",
+            "fillOpacity": 40,
+            "showPoints": "never",
+            "stacking": {
+              "mode": "normal"
+            }
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "timeseries",
+      "id": 32,
+      "targets": [
+        {
+          "expr": "namespace:gpu_count:allocated{namespace=~\"$namespace\"}",
+          "legendFormat": "{{namespace}}",
+          "refId": "A"
+        }
+      ],
+      "title": "Allocated GPUs per namespace",
+      "description": "Kube-requested GPU count per tenant over time. Step-after interpolation reflects discrete allocation changes.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 12,
+        "x": 12,
+        "y": 24
+      },
+      "repeatDirection": "h",
+      "options": {
+        "legend": {
+          "displayMode": "table",
+          "placement": "bottom",
+          "showLegend": false,
+          "calcs": [
+            "mean",
+            "max"
+          ]
+        },
+        "tooltip": {
+          "mode": "multi",
+          "sort": ""
+        }
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "short",
+          "min": 0,
+          "custom": {
+            "drawStyle": "line",
+            "lineInterpolation": "stepAfter",
+            "fillOpacity": 40,
+            "showPoints": "never",
+            "stacking": {
+              "mode": "normal"
+            }
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "row",
+      "collapsed": false,
+      "title": "Top consumers (live)",
+      "gridPos": {
+        "h": 1,
+        "w": 24,
+        "x": 0,
+        "y": 32
+      },
+      "id": 40,
+      "panels": []
+    },
+    {
+      "type": "table",
+      "id": 41,
+      "targets": [
+        {
+          "expr": "namespace:gpu_count:allocated{namespace=~\"$namespace\"}",
+          "instant": true,
+          "range": false,
+          "format": "table",
+          "refId": "gpus"
+        }
+      ],
+      "title": "Top consumers (allocation)",
+      "description": "Live ranking of tenants by GPU allocation. Hardware metrics (util, tensor, power, VRAM) cannot be attributed to workload namespaces — see Fleet and Performance dashboards for per-node/per-GPU hardware views.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 10,
+        "w": 24,
+        "x": 0,
+        "y": 33
+      },
+      "repeatDirection": "h",
+      "transformations": [
+        {
+          "id": "organize",
+          "options": {
+            "excludeByName": {
+              "Time": true,
+              "Time 1": true,
+              "Time 2": true,
+              "Time 3": true,
+              "Time 4": true,
+              "Time 5": true,
+              "__name__": true,
+              "__name__ 1": true,
+              "__name__ 2": true,
+              "__name__ 3": true,
+              "__name__ 4": true,
+              "__name__ 5": true,
+              "cluster": true,
+              "cluster 1": true,
+              "cluster 2": true,
+              "cluster 3": true,
+              "cluster 4": true,
+              "cluster 5": true,
+              "endpoint": true,
+              "endpoint 1": true,
+              "endpoint 2": true,
+              "endpoint 3": true,
+              "endpoint 4": true,
+              "endpoint 5": true,
+              "instance": true,
+              "instance 1": true,
+              "instance 2": true,
+              "instance 3": true,
+              "instance 4": true,
+              "instance 5": true,
+              "job": true,
+              "job 1": true,
+              "job 2": true,
+              "job 3": true,
+              "job 4": true,
+              "job 5": true,
+              "prometheus": true,
+              "prometheus 1": true,
+              "prometheus 2": true,
+              "prometheus 3": true,
+              "prometheus 4": true,
+              "prometheus 5": true,
+              "service": true,
+              "service 1": true,
+              "service 2": true,
+              "service 3": true,
+              "service 4": true,
+              "service 5": true,
+              "tenant": true,
+              "tenant 1": true,
+              "tenant 2": true,
+              "tenant 3": true,
+              "tenant 4": true,
+              "tenant 5": true,
+              "tier": true,
+              "tier 1": true,
+              "tier 2": true,
+              "tier 3": true,
+              "tier 4": true,
+              "tier 5": true,
+              "uid": true,
+              "uid 1": true,
+              "uid 2": true,
+              "uid 3": true,
+              "uid 4": true,
+              "uid 5": true,
+              "unit": true,
+              "unit 1": true,
+              "unit 2": true,
+              "unit 3": true,
+              "unit 4": true,
+              "unit 5": true
+            },
+            "indexByName": {
+              "Value #gpus": 1,
+              "namespace": 0
+            },
+            "renameByName": {
+              "Value #gpus": "GPUs",
+              "namespace": "Namespace"
+            }
+          }
+        }
+      ],
+      "options": {
+        "frameIndex": 0,
+        "showHeader": true,
+        "showTypeIcons": false,
+        "sortBy": [
+          {
+            "displayName": "GPUs",
+            "desc": true
+          }
+        ],
+        "footer": {
+          "show": false,
+          "reducer": []
+        },
+        "cellHeight": "sm"
+      }
+    },
+    {
+      "type": "row",
+      "collapsed": false,
+      "title": "Historical usage (last 24h)",
+      "gridPos": {
+        "h": 1,
+        "w": 24,
+        "x": 0,
+        "y": 43
+      },
+      "id": 50,
+      "panels": []
+    },
+    {
+      "type": "bargauge",
+      "id": 51,
+      "targets": [
+        {
+          "expr": "sum_over_time(namespace:gpu_count:allocated{namespace=~\"$namespace\"}[24h:1m]) / 60",
+          "legendFormat": "{{namespace}}",
+          "refId": "A"
+        }
+      ],
+      "title": "GPU-hours per namespace (last 24h)",
+      "description": "Billable GPU-hours per tenant over the last 24h window.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 7,
+        "w": 12,
+        "x": 0,
+        "y": 44
+      },
+      "repeatDirection": "h",
+      "options": {
+        "displayMode": "gradient",
+        "valueMode": "color",
+        "namePlacement": "auto",
+        "showUnfilled": true,
+        "sizing": "auto",
+        "minVizWidth": 8,
+        "minVizHeight": 16,
+        "legend": {
+          "displayMode": "list",
+          "placement": "bottom",
+          "showLegend": false,
+          "calcs": []
+        },
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ]
+        },
+        "maxVizHeight": 300,
+        "orientation": "horizontal"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "short",
+          "decimals": 1,
+          "min": 0,
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "value": null,
+                "color": "blue"
+              }
+            ]
+          }
+        },
+        "overrides": []
+      }
+    },
+    {
+      "type": "bargauge",
+      "id": 52,
+      "targets": [
+        {
+          "expr": "sum_over_time(node:power_watts:sum[24h:1m]) / 60 / 1000",
+          "legendFormat": "{{Hostname}}",
+          "refId": "A"
+        }
+      ],
+      "title": "Energy per node (last 24h)",
+      "description": "kWh consumed per node over the last 24h. Integrated from 1-minute power samples. DCGM hardware metrics cannot be attributed to workload namespaces — shown per-node instead.",
+      "transparent": false,
+      "datasource": {
+        "type": "prometheus",
+        "uid": "$ds_prometheus"
+      },
+      "gridPos": {
+        "h": 7,
+        "w": 12,
+        "x": 12,
+        "y": 44
+      },
+      "repeatDirection": "h",
+      "options": {
+        "displayMode": "gradient",
+        "valueMode": "color",
+        "namePlacement": "auto",
+        "showUnfilled": true,
+        "sizing": "auto",
+        "minVizWidth": 8,
+        "minVizHeight": 16,
+        "legend": {
+          "displayMode": "list",
+          "placement": "bottom",
+          "showLegend": false,
+          "calcs": []
+        },
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ]
+        },
+        "maxVizHeight": 300,
+        "orientation": "horizontal"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "kwatth",
+          "decimals": 2,
+          "min": 0,
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "value": null,
+                "color": "green"
+              }
+            ]
+          }
+        },
+        "overrides": []
+      }
+    }
+  ],
+  "templating": {
+    "list": [
+      {
+        "type": "datasource",
+        "name": "ds_prometheus",
+        "label": "Prometheus",
+        "skipUrlSync": false,
+        "query": "prometheus",
+        "current": {
+          "selected": false,
+          "text": "default",
+          "value": "default"
+        },
+        "multi": false,
+        "allowCustomValue": true,
+        "includeAll": false,
+        "regex": "",
+        "auto": false,
+        "auto_min": "10s",
+        "auto_count": 30
+      },
+      {
+        "type": "query",
+        "name": "namespace",
+        "label": "Namespace",
+        "skipUrlSync": false,
+        "query": "label_values(namespace:gpu_count:allocated, namespace)",
+        "datasource": {
+          "type": "prometheus",
+          "uid": "$ds_prometheus"
+        },
+        "current": {
+          "selected": true,
+          "text": "All",
+          "value": "$__all"
+        },
+        "multi": true,
+        "allowCustomValue": true,
+        "refresh": 2,
+        "sort": 1,
+        "includeAll": true,
+        "auto": false,
+        "auto_min": "10s",
+        "auto_count": 30
+      }
+    ]
+  },
+  "annotations": {}
+}
diff --git a/docs/agents/changelog.md b/docs/agents/changelog.md
index feb70597..2fde1224 100644
--- a/docs/agents/changelog.md
+++ b/docs/agents/changelog.md
@@ -6,6 +6,20 @@ This file contains detailed instructions for AI-powered IDE on how to generate c
 
 Follow these instructions when the user explicitly asks to generate a changelog.
 
+## Scope and boundaries
+
+**Your single deliverable is the file `docs/changelogs/v<version>.md`.** Write the complete, verified changelog to that path. That is the entire task. Exit as soon as the file is written and verified against the checklist in Step 9.
+
+Unless the caller explicitly instructs otherwise:
+
+- **In the cozystack working tree**, do not run `git commit`, `git push`, `git checkout` (to switch branches), `git branch`, `git tag`, `git reset`, `git merge`, or `git rebase`. Do not write to local branches, tags, or HEAD. `git fetch` is expected and fine (see the read-only analysis list below).
+- **Do not** push to any remote, open pull requests, or issue GitHub API write calls (POST / PATCH / DELETE) for any repository.
+- In the cozystack working tree, the **only** file you create or modify is `docs/changelogs/v<version>.md`. Cloning auxiliary repositories under `_repos/` for cross-repo analysis (see Step 6) is fine; local git operations inside those disposable clones (`git checkout`, `git pull`, etc.) are allowed — just never push from them or open PRs against them.
+
+The caller — a GitHub Actions workflow in CI, or a developer running you interactively — owns branching, committing, pushing, and PR creation. They will perform those actions after you exit. Do not pre-empt them even if the working tree looks ready.
+
+Read-only analysis is expected and encouraged: `git log`, `git show`, `git fetch`, `git diff`, `gh pr view`, `gh api` GET requests, and reading any file in the repository.
+
 ## Required Tools
 
 Before generating changelogs, ensure you have access to `gh` (GitHub CLI) tool, which is used to fetch commit and PR author information. The GitHub CLI is used to correctly identify PR authors from commits and pull requests.
@@ -22,7 +36,7 @@ When the user asks to generate a changelog, follow these steps in the specified
 - [ ] Step 5: Get the list of commits for the release period
 - [ ] Step 6: Check additional repositories (website is REQUIRED, optional repos if tags exist)
   - [ ] **MANDATORY**: Check website repository for documentation changes WITH authors and PR links via GitHub CLI
-  - [ ] **MANDATORY**: Check ALL optional repositories (talm, boot-to-talos, cozyhr, cozy-proxy) for tags during release period
+  - [ ] **MANDATORY**: Check ALL optional repositories (talm, boot-to-talos, cozyhr, cozy-proxy, external-apps-example, ansible-cozystack) for tags during release period
   - [ ] **MANDATORY**: For ALL commits from additional repos, get GitHub username via CLI, prioritizing PR author over commit author.
 - [ ] Step 7: Analyze commits (extract PR numbers, authors, user impact)
   - [ ] **MANDATORY**: For EVERY PR in main repo, get PR author via `gh pr view <PR_NUMBER> --json author --jq .author.login` (do NOT skip this step)
@@ -148,6 +162,8 @@ Cozystack release may include changes from related repositories. Check and inclu
 - [https://github.com/cozystack/boot-to-talos](https://github.com/cozystack/boot-to-talos)
 - [https://github.com/cozystack/cozyhr](https://github.com/cozystack/cozyhr)
 - [https://github.com/cozystack/cozy-proxy](https://github.com/cozystack/cozy-proxy)
+- [https://github.com/cozystack/external-apps-example](https://github.com/cozystack/external-apps-example)
+- [https://github.com/cozystack/ansible-cozystack](https://github.com/cozystack/ansible-cozystack)
 
 **⚠️ IMPORTANT**: You MUST check ALL optional repositories for tags created during the release period. Do NOT skip this step even if you think there might not be any tags. Use the process below to verify.
 
@@ -195,7 +211,7 @@ Cozystack release may include changes from related repositories. Check and inclu
 
 3. **For optional repositories, check if tags exist during release period:**
 
-   **⚠️ MANDATORY: You MUST check ALL optional repositories (talm, boot-to-talos, cozyhr, cozy-proxy). Do NOT skip any repository!**
+   **⚠️ MANDATORY: You MUST check ALL optional repositories (talm, boot-to-talos, cozyhr, cozy-proxy, external-apps-example, ansible-cozystack). Do NOT skip any repository!**
 
    **Use the helper script:**
    ```bash
@@ -208,7 +224,7 @@ Cozystack release may include changes from related repositories. Check and inclu
    ```
    
    The script will:
-   - Check ALL optional repositories (talm, boot-to-talos, cozyhr, cozy-proxy)
+   - Check ALL optional repositories (talm, boot-to-talos, cozyhr, cozy-proxy, external-apps-example, ansible-cozystack)
    - Look for tags created during the release period
    - Get commits between tags (if tags exist) or by date range (if no tags)
    - Extract PR numbers from commit messages
@@ -569,7 +585,7 @@ Create a new changelog file in the format matching previous versions:
 - [ ] Step 5 completed: **ALL commits included** (including merge commits and backports) - do not skip any commits
 - [ ] Step 5 completed: **Backports identified and handled correctly** - original PR author used, both original and backport PR numbers included
 - [ ] Step 6 completed: Website repository checked for documentation changes WITH authors and PR links via GitHub CLI
-- [ ] Step 6 completed: **ALL** optional repositories (talm, boot-to-talos, cozyhr, cozy-proxy) checked for tags during release period
+- [ ] Step 6 completed: **ALL** optional repositories (talm, boot-to-talos, cozyhr, cozy-proxy, external-apps-example, ansible-cozystack) checked for tags during release period
 - [ ] Step 6 completed: For ALL commits from additional repos, GitHub username obtained via GitHub CLI (not skipped). For commits with PR numbers, PR author used via `gh pr view` (not commit author)
 - [ ] Step 7 completed: For EVERY PR in main repo (including backports), PR author obtained via `gh pr view <PR_NUMBER> --json author --jq .author.login` (not skipped or assumed). Commit author NOT used - always use PR author
 - [ ] Step 7 completed: **Backports verified** - for each backport PR, original PR found and original PR author used in changelog
@@ -606,6 +622,8 @@ Create a new changelog file in the format matching previous versions:
 **Save the changelog:**
 Save the changelog to file `docs/changelogs/v<version>.md` according to the version for which the changelog is being generated.
 
+**Then exit.** Do not commit, push, create a branch, or open a pull request — the caller handles all git and GitHub operations after you return. See the "Scope and boundaries" section at the top of this document.
+
 ### Important notes
 
 - **After fetch with --force** local tags are up-to-date, use them for work
@@ -628,7 +646,7 @@ Save the changelog to file `docs/changelogs/v<version>.md` according to the vers
 
 - **Additional repositories (Step 6) - MANDATORY**: 
   - **⚠️ CRITICAL**: Always check the **website** repository for documentation changes during the release period. This is a required step and MUST NOT be skipped.
-  - **⚠️ CRITICAL**: You MUST check ALL optional repositories (talm, boot-to-talos, cozyhr, cozy-proxy) for tags during the release period. Do NOT skip any repository even if you think there might not be tags.
+  - **⚠️ CRITICAL**: You MUST check ALL optional repositories (talm, boot-to-talos, cozyhr, cozy-proxy, external-apps-example, ansible-cozystack) for tags during the release period. Do NOT skip any repository even if you think there might not be tags.
   - **CRITICAL**: For ALL entries from additional repositories (website and optional), you MUST:
     - **MANDATORY**: Extract PR number from commit message first
     - **MANDATORY**: For commits with PR numbers, ALWAYS use `gh pr view <PR_NUMBER> --repo cozystack/<repo> --json author --jq .author.login` to get PR author (not commit author)
@@ -637,7 +655,7 @@ Save the changelog to file `docs/changelogs/v<version>.md` according to the vers
     - **MANDATORY**: Do NOT use commit author for PRs - always use PR author
     - Include PR link or commit hash reference
     - Format: `* **[repo] Description**: details ([**@username**](https://github.com/username) in cozystack/repo#123)`
-  - For **optional repositories** (talm, boot-to-talos, cozyhr, cozy-proxy), you MUST check ALL of them for tags during the release period. Use the loop provided in Step 6 to check each repository systematically.
+  - For **optional repositories** (talm, boot-to-talos, cozyhr, cozy-proxy, external-apps-example, ansible-cozystack), you MUST check ALL of them for tags during the release period. Use the loop provided in Step 6 to check each repository systematically.
   - When including changes from additional repositories, use the format: `[repo-name] Description` and link to the repository's PR/issue if available
   - **Prefer PR numbers over commit hashes**: For commits from additional repositories, extract PR number from commit message using GitHub API. Use PR format (`cozystack/website#123`) instead of commit hash (`cozystack/website@abc1234`) when available
   - **Never add entries without author and PR/commit reference**: Every entry from additional repositories must have both author and link
diff --git a/docs/agents/contributing.md b/docs/agents/contributing.md
index d5a4366e..88b062ea 100644
--- a/docs/agents/contributing.md
+++ b/docs/agents/contributing.md
@@ -1,167 +1,140 @@
-# Instructions for AI Agents
+# Contributing Conventions for AI Agents
 
-Guidelines for AI agents contributing to Cozystack.
+Project-side conventions for commits, branches, and pull requests in Cozystack.
 
 ## Checklist for Creating a Pull Request
 
-- [ ] Changes are made and tested
-- [ ] Commit message uses correct `[component]` prefix
+- [ ] Commit message follows Conventional Commits format
 - [ ] Commit is signed off with `--signoff`
 - [ ] Branch is rebased on `upstream/main` (no extra commits)
 - [ ] PR body includes description and release note
-- [ ] PR is pushed and created with `gh pr create`
+- [ ] Ran `make generate` in every package whose `values.yaml`, `values.schema.json`, `Chart.yaml`, or `README.md` was touched, and committed the regenerated files
 
-## How to Commit and Create Pull Requests
+## Regenerate Artifacts Before Committing
 
-### 1. Make Your Changes
+Several files in each package are produced by `make generate` from `values.yaml` + `values.schema.json` and must stay in sync with the hand-edited sources:
 
-Edit the necessary files in the codebase.
+- `packages/(apps|extra)/<name>/README.md` — regenerated by `cozyvalues-gen` (parameter table, formatting).
+- `packages/(apps|extra)/<name>/values.schema.json` — `cozyvalues-gen` rewrites ordering and derived fields.
+- `packages/system/<name>-rd/cozyrds/<name>.yaml` — produced by `hack/update-crd.sh`, which `make generate` invokes.
 
-### 2. Commit with Proper Format
-
-Use the `[component]` prefix and `--signoff` flag:
+**Before committing edits to any of those sources**, run `make generate` inside the package and stage the full diff:
 
 ```bash
-git commit --signoff -m "[component] Brief description of changes"
+make -C packages/<apps-or-extra>/<name> generate
+git add packages/<apps-or-extra>/<name>/ packages/system/<name>-rd/
 ```
 
-**Component prefixes:**
-- System: `[dashboard]`, `[platform]`, `[cilium]`, `[kube-ovn]`, `[linstor]`, `[fluxcd]`, `[cluster-api]`
-- Apps: `[postgres]`, `[mariadb]`, `[redis]`, `[kafka]`, `[clickhouse]`, `[virtual-machine]`, `[kubernetes]`
-- Other: `[tests]`, `[ci]`, `[docs]`, `[maintenance]`
+The repo's pre-commit CI job runs `make generate` in every package and then `git diff --exit-code`. Any unstaged generator output fails the job with exit code 123 and blocks the PR. Also rerun `make generate` after a `git commit --amend` if the amended change touched any of the sources above.
+
+To locate packages a WIP branch likely needs to be regenerated:
+
+```bash
+git diff --name-only | xargs -n1 dirname | sort -u | grep ^packages/
+```
+
+## Commit Format
+
+Follow [Conventional Commits](https://www.conventionalcommits.org/) with `--signoff`:
+
+```bash
+git commit --signoff -m "type(scope): brief description"
+```
+
+**Types:** `feat`, `fix`, `docs`, `style`, `refactor`, `perf`, `test`, `build`, `ci`, `chore`
+
+**Scopes** (examples — not an exhaustive list; pick the most specific scope that describes the change, and introduce a new one if a genuinely new area needs its own):
+
+- System, e.g.: `dashboard`, `platform`, `operator`, `cilium`, `kube-ovn`, `linstor`, `fluxcd`, `cluster-api`
+- Apps, e.g.: `postgres`, `mariadb`, `redis`, `kafka`, `clickhouse`, `virtual-machine`, `kubernetes`
+- Other, e.g.: `api`, `hack`, `tests`, `ci`, `docs`, `agents`, `maintenance`
+
+Breaking changes: append `!` after type/scope (`feat(api)!: ...`) or add a `BREAKING CHANGE:` footer.
 
 **Examples:**
 ```bash
-git commit --signoff -m "[dashboard] Add config hash annotations to restart pods on config changes"
-git commit --signoff -m "[postgres] Update operator to version 1.2.3"
-git commit --signoff -m "[docs] Add installation guide"
+git commit --signoff -m "feat(dashboard): add config hash annotations to restart pods on config changes"
+git commit --signoff -m "fix(postgres): update operator to version 1.2.3"
+git commit --signoff -m "docs(contributing): add installation guide"
 ```
 
-### 3. Rebase on upstream/main (if needed)
+## PR Title Auto-Labeling
 
-If your branch has extra commits, clean it up:
+`.github/workflows/pr-labeler.yaml` parses the PR title on `opened`, `edited`, `reopened`, and `synchronize` events and applies labels additively (never removes). The title is expected to follow Conventional Commits — same format as commit messages above.
+
+**Type → `kind/*`:**
+
+| type      | label              |
+| --------- | ------------------ |
+| feat      | kind/feature       |
+| fix       | kind/bug           |
+| docs      | kind/documentation |
+| chore     | kind/cleanup       |
+| refactor  | kind/cleanup       |
+| style, perf, test, build, ci, revert | (no kind label) |
+
+**Scope → `area/*`** (full mapping in `.github/workflows/pr-labeler.yaml`):
+
+| scope (examples) | label |
+| --- | --- |
+| agents, ai | area/ai |
+| api, cozystack-api | area/api |
+| build | area/build |
+| ci | area/ci |
+| dashboard | area/dashboard |
+| postgres, mariadb, redis, etcd, kafka, clickhouse, postgres-operator, mariadb-operator | area/database |
+| extra | area/extra |
+| kubernetes | area/kubernetes |
+| monitoring, vlogs, vmstack, grafana, workloadmonitor | area/monitoring |
+| ingress, gateway, vpn, metallb, cilium, kube-ovn, cozy-proxy, ... | area/networking |
+| platform, bundle, flux, fluxcd, cluster-api, talos, installer, cozyctl, cozystack-engine, cozy-lib | area/platform |
+| backport, release | area/release |
+| seaweedfs, bucket, linstor, velero, harbor, backups | area/storage |
+| tests, e2e | area/testing |
+| kubevirt, cdi, vmi, vm-import, virtual-machine, hami, gpu-operator | area/virtualization |
+
+**Special handling:**
+
+- `[Backport release-1.x]` prefix is stripped before parsing; `area/release` and `backport` labels are added.
+- Composite scope (`feat(platform, system, apps): ...`) — each comma-separated part is mapped independently.
+- `!` after type or `BREAKING CHANGE:` footer in the body → `kind/breaking-change`.
+- Unmapped scope or non-conventional title → `area/uncategorized` (signals the PR needs manual area selection).
+- Bracket-style fallback (`[scope] description`) maps `scope` → `area/*` but cannot infer `kind/*`.
+
+### AI Agent Attribution
+
+When an AI agent authors or materially assists with a commit, add an `Assisted-By:` trailer naming the model:
+
+```text
+Assisted-By: Claude <noreply@anthropic.com>
+Assisted-By: GPT-5 <noreply@openai.com>
+Assisted-By: Gemini <noreply@google.com>
+```
+
+This sits alongside the `Signed-off-by:` trailer produced by `--signoff`. Use one trailer per model if multiple contributed.
+
+## Rebasing on upstream/main
+
+If the branch has extra commits, clean it up:
 
 ```bash
-# Fetch latest
 git fetch upstream
-
-# Create clean branch from upstream/main
 git checkout -b my-feature upstream/main
-
-# Cherry-pick only your commit
 git cherry-pick <your-commit-hash>
-
-# Force push to your branch
-git push -f origin my-feature:my-branch-name
+git push -f origin my-feature
 ```
 
-### 4. Push Your Branch
+## Pull Request Body
 
-```bash
-git push origin <branch-name>
-```
+Fill in the template at [`.github/PULL_REQUEST_TEMPLATE.md`](../../.github/PULL_REQUEST_TEMPLATE.md). It includes the required `release-note` block.
 
-### 5. Create Pull Request
+Create the PR with `gh pr create --title "type(scope): brief description" --body-file <file>`.
 
-Write the PR body to a temporary file:
+## Fetching Unresolved Review Comments
 
-```bash
-cat > /tmp/pr_body.md << 'EOF'
-## What this PR does
+Cozystack uses GitHub review threads with resolution status. Only unresolved threads are actionable — resolved threads are already handled.
 
-Brief description of the changes.
-
-Changes:
-- Change 1
-- Change 2
-
-### Release note
-
-```release-note
-[component] Description for changelog
-```
-EOF
-```
-
-Create the PR:
-
-```bash
-gh pr create --title "[component] Brief description" --body-file /tmp/pr_body.md
-```
-
-Clean up:
-
-```bash
-rm /tmp/pr_body.md
-```
-
-## Addressing AI Bot Reviewer Comments
-
-When the user asks to fix comments from AI bot reviewers (like Qodo, Copilot, etc.):
-
-### 1. Get PR Comments
-
-View all comments on the pull request:
-
-```bash
-gh pr view <PR-number> --comments
-```
-
-Or for the current branch:
-
-```bash
-gh pr view --comments
-```
-
-### 2. Review Each Comment Carefully
-
-**Important**: Do NOT blindly apply all suggestions. Each comment should be evaluated:
-
-- **Consider context** - Does the suggestion make sense for this specific case?
-- **Check project conventions** - Does it align with Cozystack patterns?
-- **Evaluate impact** - Will this improve code quality or introduce issues?
-- **Question validity** - AI bots can be wrong or miss context
-
-**When to apply:**
-- ✅ Legitimate bugs or security issues
-- ✅ Clear improvements to code quality
-- ✅ Better error handling or edge cases
-- ✅ Conformance to project conventions
-
-**When to skip:**
-- ❌ Stylistic preferences that don't match project style
-- ❌ Over-engineering simple code
-- ❌ Changes that break existing patterns
-- ❌ Suggestions that show misunderstanding of the code
-
-### 3. Apply Valid Fixes
-
-Make changes addressing the valid comments. Use your judgment.
-
-### 4. Leave Changes Uncommitted
-
-**Critical**: Do NOT commit or push the changes automatically.
-
-Leave the changes in the working directory so the user can:
-- Review the fixes
-- Decide whether to commit them
-- Make additional adjustments if needed
-
-```bash
-# After making changes, show status but DON'T commit
-git status
-git diff
-```
-
-The user will commit and push when ready.
-
-## Code Review Comments
-
-When asked to fix code review comments, **always work only with unresolved (open) comments**. Resolved comments should be ignored as they have already been addressed.
-
-### Getting Unresolved Review Comments
-
-Use GitHub GraphQL API to fetch only unresolved review comments from a pull request:
+The REST endpoint `/pulls/{pr}/reviews` returns review summaries, not individual review comments. Use the GraphQL API to access `reviewThreads` with `isResolved` status:
 
 ```bash
 gh api graphql -F owner=cozystack -F repo=cozystack -F pr=<PR_NUMBER> -f query='
@@ -189,27 +162,7 @@ query($owner: String!, $repo: String!, $pr: Int!) {
 }' --jq '.data.repository.pullRequest.reviewThreads.nodes[] | select(.isResolved == false) | .comments.nodes[]'
 ```
 
-### Filtering for Unresolved Comments
-
-The key filter is `select(.isResolved == false)` which ensures only unresolved review threads are processed. Each thread can contain multiple comments, but if the thread is resolved, all its comments should be ignored.
-
-### Working with Review Comments
-
-1. **Fetch unresolved comments** using the GraphQL query above
-2. **Parse the results** to identify:
-   - File path (`path`)
-   - Line number (`line` or `originalLine`)
-   - Comment text (`bodyText`)
-   - Author (`author.login`)
-3. **Address each unresolved comment** by:
-   - Locating the relevant code section
-   - Making the requested changes
-   - Ensuring the fix addresses the concern raised
-4. **Do NOT process resolved comments** - they have already been handled
-
-### Example: Compact List of Unresolved Comments
-
-For a quick overview of unresolved comments:
+Compact one-line variant:
 
 ```bash
 gh api graphql -F owner=cozystack -F repo=cozystack -F pr=<PR_NUMBER> -f query='
@@ -233,43 +186,3 @@ query($owner: String!, $repo: String!, $pr: Int!) {
   }
 }' --jq '.data.repository.pullRequest.reviewThreads.nodes[] | select(.isResolved == false) | .comments.nodes[] | "\(.path):\(.line // "N/A") - \(.author.login): \(.bodyText[:150])"'
 ```
-
-### Important Notes
-
-- **REST API limitation**: The REST endpoint `/pulls/{pr}/reviews` returns review summaries, not individual review comments. Use GraphQL API for accessing `reviewThreads` with `isResolved` status.
-- **Thread-based resolution**: Comments are organized in threads. If a thread is resolved (`isResolved: true`), ignore all comments in that thread.
-- **Always filter**: Never process comments from resolved threads, even if they appear in the results.
-
-### Example Workflow
-
-```bash
-# Get PR comments
-gh pr view 1234 --comments
-
-# Review comments and identify valid ones
-# Make necessary changes to address valid comments
-# ... edit files ...
-
-# Show what was changed (but don't commit)
-git status
-git diff
-
-# Tell the user what was fixed and what was skipped
-```
-
-## Git Permissions
-
-Request these permissions when needed:
-- `git_write` - For commit, rebase, cherry-pick, branch operations
-- `network` - For push, fetch, pull operations
-
-## Common Issues
-
-**PR has extra commits?**  
-→ Rebase on `upstream/main` and cherry-pick only your commits
-
-**Wrong commit message?**  
-→ `git commit --amend --signoff -m "[correct] message"` then `git push -f`
-
-**Need to update PR?**  
-→ `gh pr edit <number> --body "new description"`
diff --git a/docs/agents/overview.md b/docs/agents/overview.md
index b6d69e68..ae0fd1c4 100644
--- a/docs/agents/overview.md
+++ b/docs/agents/overview.md
@@ -78,11 +78,18 @@ packages/<category>/<package-name>/
 - Add proper error handling and structured logging
 
 ### Git Commits
-- Use format: `[component] Description`
+- Follow [Conventional Commits](https://www.conventionalcommits.org/) format: `type(scope): description`
 - Always use `--signoff` flag
 - Reference PR numbers when available
 - Keep commits atomic and focused
-- Follow conventional commit format for changelogs
+
+### PackageSource CRD upgrade policy
+
+Each component in a `PackageSource` may set `install.upgradeCRDs` to control how CRDs from the chart's `crds/` directory are handled on `HelmRelease` upgrades. Allowed values: `Skip` (default — helm-controller does not touch CRDs on upgrade), `Create` (create new CRDs only), `CreateReplace` (create new and overwrite existing).
+
+Set `upgradeCRDs: CreateReplace` for operators whose upstream regularly adds new CRDs between versions (etcd-operator, cnpg, kubevirt, kamaji). Without it, new CRDs from a chart bump do not land on existing clusters — only fresh installs get them.
+
+Do **not** set `CreateReplace` blindly: it overwrites every CRD in `crds/` and can cause silent data loss if upstream drops a field from a CRD that has live objects. Only enable it for operators whose schema evolution is additive-only. When in doubt, leave it unset and apply new CRDs manually.
 
 ### Documentation
 
diff --git a/docs/changelogs/v1.1.6.md b/docs/changelogs/v1.1.6.md
new file mode 100644
index 00000000..77c49475
--- /dev/null
+++ b/docs/changelogs/v1.1.6.md
@@ -0,0 +1,19 @@
+<!--
+https://github.com/cozystack/cozystack/releases/tag/v1.1.6
+-->
+
+## Fixes
+
+* **[build] Filter git describe to match only v* tags**: Adds `--match 'v*'` to all `git describe` calls in `hack/common-envs.mk`. The `api/apps/v1alpha1/*` subtags share the same commit as release tags, causing `git describe --exact-match` to pick `api/apps/v1alpha1/vX.Y.Z` instead of `vX.Y.Z`, producing invalid Docker image tags ([**@kvaps**](https://github.com/kvaps) in #2386, #2388).
+
+## Development, Testing, and CI/CD
+
+* **[ci] Replace cozystack-bot PAT with cozystack-ci GitHub App**: Replaces the long-lived `cozystack-bot` personal access token with short-lived, scoped tokens from the `cozystack-ci` GitHub App across all release workflows. Improves security and auditability of CI operations ([**@tym83**](https://github.com/tym83) in #2351).
+
+* **[ci] Replace GH_PAT with cozystack-ci GitHub App token in pull-requests workflow**: Switches the pull-requests release workflow to use the cozystack-ci GitHub App token instead of the personal access token ([**@kvaps**](https://github.com/kvaps) in #2383).
+
+* **[ci] Use cozystack org noreply email for bot commits**: Updates CI workflows to use the cozystack organization noreply email for bot commits ([**@kvaps**](https://github.com/kvaps) in #2392).
+
+---
+
+**Full Changelog**: https://github.com/cozystack/cozystack/compare/v1.1.5...v1.1.6
diff --git a/docs/changelogs/v1.2.2.md b/docs/changelogs/v1.2.2.md
new file mode 100644
index 00000000..2b6a1ae5
--- /dev/null
+++ b/docs/changelogs/v1.2.2.md
@@ -0,0 +1,63 @@
+<!--
+https://github.com/cozystack/cozystack/releases/tag/v1.2.2
+-->
+
+## Features and Improvements
+
+* **[linstor] Update piraeus-server to v1.33.2 with selected backports**: Bumps LINSTOR server from v1.33.1 to v1.33.2 and adds backported patches for improved storage reliability: a stale bitmap adjust retry mechanism for automatic recovery after bitmap attach errors, LUKS2 header sizing and optimal I/O size detection improvements for more reliable disk formatting, and the maintainer implementation backport. All patches verified against upstream v1.33.2 with `git apply --check` and `gradlew compileJava` ([**@kvaps**](https://github.com/kvaps) in #2331, #2377).
+
+## Fixes
+
+* **[postgres] Fix system PostgreSQL images to 17.7-standard-trixie**: Hardcodes PostgreSQL 17.7-standard-trixie images for system PostgreSQL instances. This ensures system databases use the correct image variant consistent with the monitoring stack requirements introduced in v1.2.1 ([**@myasnikovdaniil**](https://github.com/myasnikovdaniil) in #2364, #2369).
+
+* **[cilium] Opt-out of cri-containerd.apparmor.d for nsenter init containers**: On Ubuntu 22.04+, Debian, and other distributions that load the `cri-containerd.apparmor.d` AppArmor profile by default for containerd workloads, the kernel denied `nsenter` namespace entry in cilium-agent init containers (`mount-cgroup`, `apply-sysctl-overwrites`, `clean-cilium-state`), causing the agent to land in `Init:CrashLoopBackOff` and cascading platform failures. Per-container `container.apparmor.security.beta.kubernetes.io` annotations now opt the affected containers out of this profile, applied only on non-Talos cilium variants (`cilium-generic`, `kubeovn-cilium-generic`). The vendored daemonset template is also patched to strip the upstream `semverCompare "<1.30.0"` AppArmor block, preventing duplicate annotation keys. Talos variants are untouched as Talos does not load the AppArmor LSM ([**@lexfrei**](https://github.com/lexfrei) in #2370, #2378).
+
+* **[virtual-machine] Exclude external VM services from Cilium BPF LB**: Adds the `service.kubernetes.io/service-proxy-name: "cozy-proxy"` label to VM LoadBalancer services when `external: true`, telling Cilium to skip BPF processing entirely for these services. This fixes two issues: inter-tenant connectivity via public LB IPs (Cilium's DNAT caused cross-tenant pod-to-pod flow classification, triggering CiliumClusterwideNetworkPolicy blocks) and WholeIP broken on Cilium 1.19+ (wildcard service drop entries blocked traffic to LB IPs on undeclared ports before it reached netfilter/cozy-proxy). MetalLB L2 advertisement and kube-ovn routing remain unaffected ([**@mattia-eleuteri**](https://github.com/mattia-eleuteri) in #2357, #2361).
+
+* **[monitoring] Fix infra dashboards missing in default variant**: The default platform variant deploys the monitoring chart to the `cozy-monitoring` namespace, but the dashboard rendering condition introduced in #2197 only checked for `tenant-root`. Infrastructure dashboards were not rendered in the default variant. The `cozy-monitoring` namespace is now included in the rendering condition, consistent with the existing pattern in `vmagent.yaml` ([**@mattia-eleuteri**](https://github.com/mattia-eleuteri) in #2365, #2367).
+
+* **[build] Filter git describe to match only v* tags**: Adds `--match 'v*'` to all `git describe` calls in `hack/common-envs.mk`. The `api/apps/v1alpha1/*` subtags share the same commit as release tags, causing `git describe --exact-match` to pick `api/apps/v1alpha1/vX.Y.Z` instead of `vX.Y.Z`, producing invalid Docker image tags ([**@kvaps**](https://github.com/kvaps) in #2386, #2389).
+
+## Development, Testing, and CI/CD
+
+* **[ci] Replace cozystack-bot PAT with cozystack-ci GitHub App**: Replaces the long-lived `cozystack-bot` personal access token with short-lived, scoped tokens from the `cozystack-ci` GitHub App across all release workflows (`tags.yaml`, `auto-release.yaml`, `pull-requests-release.yaml`). Improves security and auditability of CI operations ([**@tym83**](https://github.com/tym83) in #2351).
+
+* **[ci] Use cozystack org noreply email for bot commits**: Updates CI workflows to use the cozystack organization noreply email for bot commits ([**@kvaps**](https://github.com/kvaps) in #2392, #2393).
+
+* **[ci] Replace GH_PAT with cozystack-ci GitHub App token in pull-requests workflow**: Switches the pull-requests release workflow to use the cozystack-ci GitHub App token instead of the personal access token ([**@kvaps**](https://github.com/kvaps) in #2383, #2384).
+
+## Documentation
+
+* **[website] Add ApplicationDefinition naming convention reference**: Added reference documentation on ApplicationDefinition naming conventions and how `cozystack-api` resolves kinds to their backing definitions ([**@lexfrei**](https://github.com/lexfrei) in cozystack/website#478).
+
+* **[website] Document Talos / talosctl / Cozystack version pairing**: Added documentation covering Talos, talosctl, and Cozystack version compatibility matrix for installation ([**@lexfrei**](https://github.com/lexfrei) in cozystack/website#484).
+
+* **[website] Fix KubeOVN MASTER_NODES example path and key in troubleshooting**: Corrected the MASTER_NODES example path and key in the KubeOVN troubleshooting guide ([**@lexfrei**](https://github.com/lexfrei) in cozystack/website#483).
+
+* **[website] Prefix bundle package names with cozystack. in v1 examples**: Updated documentation examples to use the correct `cozystack.` prefix for bundle package names in enabled/disabledPackages ([**@lexfrei**](https://github.com/lexfrei) in cozystack/website#482).
+
+* **[website] Finish isolated-field removal and document opt-in policy labels**: Removed the obsolete `isolated` field from tenant documentation and documented the new opt-in policy labels approach ([**@lexfrei**](https://github.com/lexfrei) in cozystack/website#481).
+
+* **[website] Add --take-ownership flag and describe networking.* fields**: Added documentation for the `--take-ownership` flag and described the `networking.*` fields in the installation guide ([**@lexfrei**](https://github.com/lexfrei) in cozystack/website#480).
+
+* **[website] Add bonding (LACP) configuration how-to guide**: Added a guide for configuring network bonding with LACP on Cozystack installations ([**@sircthulhu**](https://github.com/sircthulhu) in cozystack/website#459).
+
+* **[website] Improve registry mirrors for tenant Kubernetes in air-gapped guide**: Improved documentation for configuring registry mirrors in tenant Kubernetes clusters for air-gapped environments ([**@sircthulhu**](https://github.com/sircthulhu) in cozystack/website#461).
+
+* **[website] Update backup/restore documentation for VMI/VMDisk**: Updated backup documentation with information related to VM instance and VM disk restore improvements ([**@androndo**](https://github.com/androndo) in cozystack/website#466).
+
+* **[website] Add updated OpenAPI spec**: Updated the OpenAPI specification for managed applications reference ([**@myasnikovdaniil**](https://github.com/myasnikovdaniil) in cozystack/website#469).
+
+* **[website] Add OSS Health pages and OpenSSF badge**: Added OSS Health section with OpenSSF Scorecard and Best Practices badge to the website footer ([**@tym83**](https://github.com/tym83) in cozystack/website#470).
+
+* **[website] Add CozySummit Virtual 2026 program announcement**: Published the CozySummit Virtual 2026 program announcement blog post ([**@tym83**](https://github.com/tym83) in cozystack/website#472).
+
+* **[website] Add missing release announcements for v0.1–v0.41**: Backfilled missing release announcement blog posts for Cozystack versions v0.1 through v0.41 ([**@tym83**](https://github.com/tym83) in cozystack/website#468).
+
+* **[talm] Render templates online in apply to resolve lookups**: Fixed talm `apply` command to render templates online, resolving template lookup failures when using modeline templates ([**@myasnikovdaniil**](https://github.com/myasnikovdaniil) in cozystack/talm#119).
+
+* **[talm] Update default Talos image to v1.12.6**: Updated the default Talos image version to v1.12.6 in talm ([**@kvaps**](https://github.com/kvaps) in cozystack/talm@03e9b6e).
+
+---
+
+**Full Changelog**: https://github.com/cozystack/cozystack/compare/v1.2.1...v1.2.2
diff --git a/docs/changelogs/v1.2.3.md b/docs/changelogs/v1.2.3.md
new file mode 100644
index 00000000..6d1b3b74
--- /dev/null
+++ b/docs/changelogs/v1.2.3.md
@@ -0,0 +1,58 @@
+<!--
+https://github.com/cozystack/cozystack/releases/tag/v1.2.3
+-->
+
+# v1.2.3 (2026-04-20)
+
+A patch release with bug fixes and documentation updates.
+
+## Features and Improvements
+
+_No notable features in this patch release._
+
+## Fixes
+
+* **fix(kubernetes): set explicit ephemeral-storage on virt-launcher pods**: Prevents VM crashes caused by ephemeral-storage eviction by setting explicit `domain.resources` ephemeral-storage on the VirtualMachine spec. Uses sanitized limits and requests so virt-launcher pods do not inherit too-small namespace defaults. ([**@kvaps**](https://github.com/kvaps) in #2317, backport #2423).
+
+## Documentation
+
+* **[website] feat: add Telemetry page under OSS Health section**: Add Telemetry page and initial data seeding to OSS Health docs ([**@tym83**](https://github.com/tym83) in cozystack/website#471).
+* **[website] Refactor docs versions to major.minor variants**: Move docs to major.minor versioning for v1.x series ([**@myasnikovdaniil**](https://github.com/myasnikovdaniil) in cozystack/website#477).
+* **[website] docs(tenants): document namespace layout and parent/child derivation** ([**@lexfrei**](https://github.com/lexfrei) in cozystack/website#479).
+* **[website] docs(tenants): document the checkbox-then-edit-CR customization pattern** ([**@lexfrei**](https://github.com/lexfrei) in cozystack/website#485).
+* **[website] docs: fix 14 broken links and stale talm anchor across v1 docs** ([**@lexfrei**](https://github.com/lexfrei) in cozystack/website#486).
+* **[website] fix(og): update social badge image and title** ([**@tym83**](https://github.com/tym83) in cozystack/website#487).
+* **[website] docs(external-apps): rewrite guide for ApplicationDefinition API** ([**@kitsunoff**](https://github.com/kitsunoff) in cozystack/website#488).
+* **[website] docs: add CLAUDE.md for AI agent guidance** ([**@myasnikovdaniil**](https://github.com/myasnikovdaniil) in cozystack/website#489).
+* **[website] fix: update /docs/v1/ redirect to latest v1.2** ([**@myasnikovdaniil**](https://github.com/myasnikovdaniil) in cozystack/website#492).
+* **[website] fix(ci): add OpenAPI spec download to GitHub Pages build** ([**@myasnikovdaniil**](https://github.com/myasnikovdaniil) in cozystack/website#494).
+* **[website] feat(blog): add managed PostgreSQL with synchronous replication post** ([**@tym83**](https://github.com/tym83) in cozystack/website#497).
+* **[website] chore(blog): add images frontmatter for social preview on existing posts** ([**@tym83**](https://github.com/tym83) in cozystack/website#498).
+* **[website] feat(blog): taxonomies and client-side filter UI** ([**@tym83**](https://github.com/tym83) in cozystack/website#499).
+* **[website] style(oss-health): add breathing room between navbar and hero** ([**@tym83**](https://github.com/tym83) in cozystack/website#500).
+
+## Other repositories
+
+* **[talm] feat(config): migrate to Talos v1.12 multi-document config format**: Upgrade Talos config format and modernize configuration handling ([**@lexfrei**](https://github.com/lexfrei) in cozystack/talm#116).
+* **[talm] chore(deps): bump dependencies and modernize codebase** ([**@lexfrei**](https://github.com/lexfrei) in cozystack/talm#124).
+* **[external-apps-example] feat: replace MongoDB example with Minecraft apps from cozylex** ([**@lexfrei**](https://github.com/lexfrei) in cozystack/external-apps-example#2).
+* **[ansible-cozystack] fix(examples): add v prefix to collection version in requirements.yml** ([**@lexfrei**](https://github.com/lexfrei) in cozystack/ansible-cozystack#23).
+* **[ansible-cozystack] fix(plugins): replace ansible.utils.ipaddr with stdlib-based test plugin** ([**@lexfrei**](https://github.com/lexfrei) in cozystack/ansible-cozystack#24).
+* **[ansible-cozystack] feat(examples): comprehensive node prerequisites audit (fixes #19)** ([**@lexfrei**](https://github.com/lexfrei) in cozystack/ansible-cozystack#27).
+* **[ansible-cozystack] chore(deps): update dependency cozystack.installer to v1.2.3** ([**@app/renovate**](https://github.com/apps/renovate) in cozystack/ansible-cozystack#29).
+* **[ansible-cozystack] feat(role): expose publishing.externalIPs and tenant-root ingress via role variables** ([**@lexfrei**](https://github.com/lexfrei) in cozystack/ansible-cozystack#30).
+
+## Contributors
+
+Thanks to everyone who contributed to this patch release:
+
+* [**@app/github-actions**](https://github.com/apps/github-actions)
+* [**@app/renovate**](https://github.com/apps/renovate)
+* [**@kitsunoff**](https://github.com/kitsunoff)
+* [**@kvaps**](https://github.com/kvaps)
+* [**@lexfrei**](https://github.com/lexfrei)
+* [**@myasnikovdaniil**](https://github.com/myasnikovdaniil)
+* [**@tym83**](https://github.com/tym83)
+
+
+**Full Changelog**: https://github.com/cozystack/cozystack/compare/v1.2.2...v1.2.3
diff --git a/docs/changelogs/v1.3.0-rc.1.md b/docs/changelogs/v1.3.0-rc.1.md
new file mode 100644
index 00000000..9b63c106
--- /dev/null
+++ b/docs/changelogs/v1.3.0-rc.1.md
@@ -0,0 +1,173 @@
+<!--
+https://github.com/cozystack/cozystack/releases/tag/v1.3.0-rc.1
+-->
+
+# Cozystack v1.3.0-rc.1
+
+Cozystack v1.3.0-rc.1 is the first release candidate for v1.3.0, bringing **storage-aware scheduling** via the LINSTOR scheduler extender, a managed **LINSTOR GUI** web UI with Keycloak SSO, a **VM Default Images** catalog for out-of-the-box virtual machine provisioning, **WorkloadsReady conditions** with a real-time Events tab in the dashboard, and **cross-namespace VM backup restore** capabilities. Additional highlights include stricter tenant name validation, VM network selector improvements, Keycloak theme injection and SMTP configuration, and a comprehensive host runtime preflight check.
+
+> **Note:** Fixes marked with *(backported to v1.2.x)* were also included in v1.2.1 or v1.2.2 patch releases.
+
+## Feature Highlights
+
+### Storage-Aware Scheduling via LINSTOR Extender
+
+The `cozystack-scheduler` now calls the **LINSTOR scheduler extender** for storage-locality-aware pod placement. When a pod declares both a `SchedulingClass` and LINSTOR-backed PVCs, the scheduler consults LINSTOR to prefer nodes where volume replicas already exist — reducing cross-node replication traffic and improving I/O latency for storage-heavy workloads ([**@lllamnyp**](https://github.com/lllamnyp) in #2330).
+
+### LINSTOR GUI: Managed Web UI for Storage Administration
+
+A new opt-in `linstor-gui` system package deploys **LINBIT's linstor-gui web UI** alongside the LINSTOR controller with mTLS client authentication, non-root security context, and ClusterIP-only service. An optional **Keycloak-protected Ingress** (via oauth2-proxy) can be enabled for SSO-authenticated browser access when OIDC is configured on the platform ([**@myasnikovdaniil**](https://github.com/myasnikovdaniil) in #2382, #2390).
+
+### VM Default Images: Out-of-the-Box VM Provisioning
+
+The new `vm-default-images` package provides a curated set of **cluster-wide virtual machine images** (Ubuntu, Debian, CentOS Stream, and others) as pre-populated DataVolumes. The package is opt-in via the `iaas` bundle and defaults to replicated storage for high availability. A companion migration (migration 38) renames legacy `vm-image-*` DataVolumes to the new `vm-default-images-*` naming scheme. The `vm-disk` chart also gains a new "disk" source type for cloning from existing vm-disks in the same namespace ([**@myasnikovdaniil**](https://github.com/myasnikovdaniil) in #2258).
+
+### WorkloadsReady Condition and Events Tab
+
+Applications now expose a **WorkloadsReady** condition on their status by querying associated WorkloadMonitor resources, giving operators a single place to check whether all underlying workloads (Deployments, StatefulSets, DaemonSets) are healthy. The dashboard gains a new **Events tab** showing namespace-scoped Kubernetes events for each application, with fallback to `.firstTimestamp` when `.eventTime` is absent. A bug where WorkloadMonitor's `Operational` status was never persisted is also fixed ([**@lexfrei**](https://github.com/lexfrei) in #2356).
+
+### Cross-Namespace VM Backup Restore
+
+The backup system now supports **restoring VMInstance backups into a different namespace** (cross-namespace copy restores), with IP/MAC preservation and safe rename semantics. In-place backup/restores for VMDisk and VMInstance are improved: HelmReleases and DataVolumes are properly handled, and Velero failure messages are propagated to the Application status. The backup status structure has been refactored to store underlying resources as a generic opaque JSON object, enabling arbitrary application-specific metadata ([**@androndo**](https://github.com/androndo) in #2251, #2329, #2319).
+
+## Major Features and Improvements
+
+* **[api] Reject tenant names with dashes at Create time**: Enforces alphanumeric-only naming for Tenants at the API level, preventing names with hyphens that would silently fail during Helm reconciliation. A corresponding regex tightening and regression test suite hardens the validation ([**@lexfrei**](https://github.com/lexfrei) in #2380).
+
+* **[platform] Validate computed tenant namespace length**: Rejects Tenant creation when the computed ancestor-chain namespace would exceed the 63-character Kubernetes namespace limit, preventing opaque HelmRelease reconcile errors downstream ([**@lexfrei**](https://github.com/lexfrei) in #2376).
+
+* **[vm-instance] Rename subnets to networks and add dropdown selector**: Renames the misleading `subnets` field to `networks` in VMInstance for clarity, adds a dropdown selector for available networks in the dashboard form, and includes a migration to copy existing `subnets` values. The old field remains supported for backward compatibility ([**@sircthulhu**](https://github.com/sircthulhu) in #2263).
+
+* **[keycloak] Enable injecting themes**: Cozystack administrators can now inject custom Keycloak themes via `initContainers` for UI white-labeling and customization ([**@lllamnyp**](https://github.com/lllamnyp) in #2142).
+
+* **[keycloak-configure] Add email verification and SMTP configuration**: Adds configurable Keycloak settings for user self-registration, email verification, and SMTP server configuration, enabling automated user onboarding flows ([**@BROngineer**](https://github.com/BROngineer) in #2318).
+
+* **[postgres] Hardcode PostgreSQL 17 for monitoring databases**: Pins PostgreSQL 17.7 images for system databases (Grafana, Alerta, Harbor, Keycloak, SeaweedFS) and adds migration 37 to backfill `spec.version=v17` for existing PostgreSQL resources, preventing CNPG from defaulting to PostgreSQL 18 *(backported to v1.2.1)* ([**@IvanHunters**](https://github.com/IvanHunters) in #2304).
+
+* **[hack] Add host runtime preflight check**: New `check-host-runtime.sh` script and `make preflight` target that warns operators when a standalone containerd or docker runtime is running alongside the embedded k3s runtime, helping diagnose container runtime conflicts ([**@lexfrei**](https://github.com/lexfrei) in #2371).
+
+* **[hack] Add check-readiness.sh diagnostic script**: A new diagnostic script for tracking platform reconciliation by checking readiness of Packages, ArtifactGenerators, ExternalArtifacts, and HelmReleases, with support for watch mode and continuous monitoring ([**@myasnikovdaniil**](https://github.com/myasnikovdaniil) in #2294).
+
+* **[mariadb] Always enable replication for consistent service naming**: MariaDB now always enables replication, creating `-primary`/`-secondary` services even for single-replica instances. This fixes dashboard visibility and backup functionality for single-replica setups ([**@sircthulhu**](https://github.com/sircthulhu) in #2279).
+
+* **[platform] Prevent installed packages deletion**: Adds `helm.sh/resource-policy: keep` annotation to packages, preventing automatic deletion when packages are disabled and restoring documented behavior *(backported to v1.2.1)* ([**@kvaps**](https://github.com/kvaps) in #2273).
+
+## Bug Fixes
+
+* **[cilium] Opt-out of cri-containerd.apparmor.d for nsenter init containers**: Opts cilium-agent init containers out of the `cri-containerd.apparmor.d` AppArmor profile on non-Talos variants, fixing `Init:CrashLoopBackOff` on Ubuntu 22.04+ and Debian *(backported to v1.2.2)* ([**@lexfrei**](https://github.com/lexfrei) in #2370).
+
+* **[virtual-machine] Exclude external VM services from Cilium BPF LB**: Adds `service-proxy-name: cozy-proxy` label to VM LoadBalancer services, telling Cilium to skip BPF processing. Fixes inter-tenant connectivity via public LB IPs and WholeIP functionality on Cilium 1.19+ *(backported to v1.2.2)* ([**@mattia-eleuteri**](https://github.com/mattia-eleuteri) in #2357).
+
+* **[monitoring] Fix infra dashboards missing in default variant**: Includes `cozy-monitoring` namespace in the dashboard rendering condition, fixing infrastructure Grafana dashboards not rendering in the default platform variant *(backported to v1.2.2)* ([**@mattia-eleuteri**](https://github.com/mattia-eleuteri) in #2365).
+
+* **[postgres] Fix system PostgreSQL images to 17.7-standard-trixie**: Normalizes system PostgreSQL image tags to use `17.7-standard-trixie` variant with migration logic for existing CNPG clusters *(backported to v1.2.2)* ([**@myasnikovdaniil**](https://github.com/myasnikovdaniil) in #2364).
+
+* **[build] Filter git describe to match only v\* tags**: Adds `--match 'v*'` to `git describe` calls, preventing API subtags from being picked up instead of release tags and producing invalid Docker image tags *(backported to v1.2.2)* ([**@kvaps**](https://github.com/kvaps) in #2386).
+
+* **[platform] Fix resource allocation ratios not propagated to packages**: Restores propagation of CPU, memory, and ephemeral-storage allocation ratios to managed applications and KubeVirt, which were silently ignored since the bundle restructure *(backported to v1.2.1)* ([**@sircthulhu**](https://github.com/sircthulhu) in #2296).
+
+* **[kubernetes] Set explicit ephemeral-storage on virt-launcher pods**: Sets explicit `domain.resources` with ephemeral-storage on VirtualMachine spec to prevent virt-launcher pods from being evicted due to LimitRange defaults being too low for actual emptyDisk capacity ([**@kvaps**](https://github.com/kvaps) in #2317).
+
+* **[multus] Pin master CNI to 05-cilium.conflist**: Prevents a boot-time race condition where multus could auto-detect kube-ovn's conflist instead of Cilium's *(backported to v1.2.1)* ([**@kvaps**](https://github.com/kvaps) in #2315).
+
+* **[multus] Build custom image with DEL cache fix**: Fixes sandbox cleanup deadlock when CNI ADD never completes, preventing stale sandbox name reservations from permanently blocking pod creation *(backported to v1.2.1)* ([**@kvaps**](https://github.com/kvaps) in #2313).
+
+* **[linstor] Set verify-alg to crc32c**: Prevents DRBD connection failures on kernels where `crct10dif` is unavailable (e.g., Talos v1.12.6 with kernel 6.18.18) *(backported to v1.2.1)* ([**@kvaps**](https://github.com/kvaps) in #2303).
+
+* **[linstor] Preserve TCP ports during toggle-disk operations**: Fixes TCP port mismatches after toggle-disk operations that could cause DRBD resources to enter StandAlone state *(backported to v1.2.1)* ([**@kvaps**](https://github.com/kvaps) in #2292).
+
+## Dependencies & Version Updates
+
+* **[linstor] Update piraeus-server to v1.33.2 with selected backports**: Bumps LINSTOR server from v1.33.1 to v1.33.2 with backported patches for stale bitmap adjust retry, LUKS2 header sizing, and optimal I/O size detection *(backported to v1.2.2)* ([**@kvaps**](https://github.com/kvaps) in #2331).
+
+* **[kamaji] Update to 26.3.5-edge, drop upstreamed patches**: Updates Kamaji from edge-26.2.4 to 26.3.5-edge and removes two patches accepted upstream. Adds configurable probe tuning and DataStore readiness conditions ([**@myasnikovdaniil**](https://github.com/myasnikovdaniil) in #2260).
+
+* **[talm] Release v0.23.0, v0.23.1, v0.24.0** (github.com/cozystack/talm): Migrates to the Talos v1.12 multi-document machine config format ([**@lexfrei**](https://github.com/lexfrei) in cozystack/talm#116); fixes template rendering in `apply` command to resolve lookups ([**@myasnikovdaniil**](https://github.com/myasnikovdaniil) in cozystack/talm#119); bumps dependencies and modernizes codebase ([**@lexfrei**](https://github.com/lexfrei) in cozystack/talm#124).
+
+* **[ansible-cozystack] Release v1.2.1, v1.2.2** (github.com/cozystack/ansible-cozystack): Exposes `publishing.externalIPs` and tenant-root ingress via role variables ([**@lexfrei**](https://github.com/lexfrei) in cozystack/ansible-cozystack#30); adds comprehensive node prerequisites audit ([**@lexfrei**](https://github.com/lexfrei) in cozystack/ansible-cozystack#27); replaces `ansible.utils.ipaddr` with a stdlib-based test plugin ([**@lexfrei**](https://github.com/lexfrei) in cozystack/ansible-cozystack#24).
+
+## Security
+
+* **docs: add SECURITY.md**: Adds vulnerability reporting procedures, disclosure expectations, and supported release lines ([**@kvaps**](https://github.com/kvaps) in #2230).
+
+* **docs: add OpenSSF Best Practices badge to README**: Adds the OpenSSF Best Practices passing badge to the project README ([**@lexfrei**](https://github.com/lexfrei) in #2320).
+
+## Development, Testing, and CI/CD
+
+* **[ci] Replace cozystack-bot PAT with cozystack-ci GitHub App**: Replaces the long-lived cozystack-bot personal access token with short-lived, scoped tokens from the cozystack-ci GitHub App across all CI release workflows ([**@tym83**](https://github.com/tym83) in #2351; [**@kvaps**](https://github.com/kvaps) in #2383, #2392).
+
+* **[ci] Add Gemini Code Assist and CodeRabbit configuration**: Adds repository-level configuration for AI code reviewers with ignore patterns for vendored/generated code and incremental review settings ([**@lexfrei**](https://github.com/lexfrei) in #2385).
+
+* **[ci] Make tags workflow idempotent on re-runs**: Fixes CI to force-update API subtags and handle re-runs gracefully ([**@kvaps**](https://github.com/kvaps)).
+
+* **[tests] Fix Kafka E2E test timeout and retry race condition**: Increases Kafka E2E test timeout from 60s to 300s and fixes a retry race condition where `kubectl apply` could hit a still-deleting resource ([**@lexfrei**](https://github.com/lexfrei) in #2358).
+
+* **docs: adopt Conventional Commits for commit and PR titles**: Standardizes commit and PR title format to `type(scope): description` across all contributing docs and the PR template ([**@lexfrei**](https://github.com/lexfrei) in #2395).
+
+* **docs(ci): require screenshots for UI changes in PR template**: Adds a mandatory screenshots section to the PR template for UI-related changes ([**@kitsunoff**](https://github.com/kitsunoff) in #2407).
+
+## Documentation
+
+* **[website] Add ApplicationDefinition naming convention reference**: Documents how `cozystack-api` resolves kinds to their backing definitions ([**@lexfrei**](https://github.com/lexfrei) in cozystack/website#478).
+
+* **[website] Document Talos / talosctl / Cozystack version pairing**: Adds version compatibility matrix for installation ([**@lexfrei**](https://github.com/lexfrei) in cozystack/website#484).
+
+* **[website] Document namespace layout and parent/child derivation**: Explains tenant namespace hierarchy and parent/child namespace derivation rules ([**@lexfrei**](https://github.com/lexfrei) in cozystack/website#479).
+
+* **[website] Document the checkbox-then-edit-CR customization pattern for tenants**: Describes the workflow for customizing tenant settings via the CR after initial checkbox-based creation ([**@lexfrei**](https://github.com/lexfrei) in cozystack/website#485).
+
+* **[website] Add custom Keycloak themes documentation**: Covers the theme image contract, configuration, `imagePullSecrets`, and theme activation in the Keycloak admin console ([**@lexfrei**](https://github.com/lexfrei) in cozystack/website#463).
+
+* **[website] Add bonding (LACP) configuration how-to guide**: Covers network bonding configuration for Cozystack installations ([**@sircthulhu**](https://github.com/sircthulhu) in cozystack/website#459).
+
+* **[website] Improve registry mirrors for tenant Kubernetes in air-gapped guide**: Improved documentation for configuring registry mirrors in air-gapped environments ([**@sircthulhu**](https://github.com/sircthulhu) in cozystack/website#461).
+
+* **[website] Rewrite guide for ApplicationDefinition API (external-apps)**: Comprehensive rewrite of the external apps guide using the ApplicationDefinition API ([**@kitsunoff**](https://github.com/kitsunoff) in cozystack/website#488).
+
+* **[website] Add documentation for Go types usage**: Guide for using generated Go types for Cozystack managed applications as a Go module ([**@myasnikovdaniil**](https://github.com/myasnikovdaniil) in cozystack/website#465).
+
+* **[website] Update backup/restore documentation for VMI/VMDisk**: Updated backup documentation with VM instance and VM disk restore improvements ([**@androndo**](https://github.com/androndo) in cozystack/website#466).
+
+* **[website] Add OSS Health pages and OpenSSF badge**: Added OSS Health section with OpenSSF Scorecard and Best Practices badge to the website ([**@tym83**](https://github.com/tym83) in cozystack/website#470).
+
+* **[website] Add CozySummit Virtual 2026 program announcement**: Published the CozySummit Virtual 2026 program announcement blog post ([**@tym83**](https://github.com/tym83) in cozystack/website#472).
+
+* **[website] Add missing release announcements for v0.1–v0.41**: Backfilled missing release announcement blog posts for historical Cozystack versions ([**@tym83**](https://github.com/tym83) in cozystack/website#468).
+
+* **[website] Fix broken links and stale anchors across v1 docs**: Fixes 14 broken links and stale talm anchors ([**@lexfrei**](https://github.com/lexfrei) in cozystack/website#486).
+
+* **[website] Prefix bundle package names with cozystack. in v1 examples**: Corrects package naming in documentation examples ([**@lexfrei**](https://github.com/lexfrei) in cozystack/website#482).
+
+* **[website] Finish isolated-field removal and document opt-in policy labels**: Removes obsolete `isolated` field from tenant documentation and documents the new approach ([**@lexfrei**](https://github.com/lexfrei) in cozystack/website#481).
+
+* **[website] Add --take-ownership flag and describe networking.* fields**: Documents the `--take-ownership` flag and `networking.*` fields in the installation guide ([**@lexfrei**](https://github.com/lexfrei) in cozystack/website#480).
+
+* **[website] Fix KubeOVN MASTER_NODES example path and key in troubleshooting**: Corrects the MASTER_NODES example path ([**@lexfrei**](https://github.com/lexfrei) in cozystack/website#483).
+
+* **[external-apps-example] Replace MongoDB example with Minecraft apps**: Refactors the external apps example to use ApplicationDefinition API with Minecraft server applications ([**@lexfrei**](https://github.com/lexfrei) in cozystack/external-apps-example#2).
+
+## Governance
+
+* **Add Mattia Eleuteri ([@mattia-eleuteri](https://github.com/mattia-eleuteri)) as Maintainer**: CSI, Storage, Networking & Security ([**@tym83**](https://github.com/tym83) in #2345).
+
+* **Add Matthieu Robin ([@matthieu-robin](https://github.com/matthieu-robin)) as Maintainer**: Managed applications, platform quality, and benchmarking ([**@tym83**](https://github.com/tym83) in #2346).
+
+## Contributors
+
+We'd like to thank all contributors who made this release possible:
+
+* [**@androndo**](https://github.com/androndo)
+* [**@BROngineer**](https://github.com/BROngineer)
+* [**@IvanHunters**](https://github.com/IvanHunters)
+* [**@kitsunoff**](https://github.com/kitsunoff)
+* [**@kvaps**](https://github.com/kvaps)
+* [**@lexfrei**](https://github.com/lexfrei)
+* [**@lllamnyp**](https://github.com/lllamnyp)
+* [**@mattia-eleuteri**](https://github.com/mattia-eleuteri)
+* [**@myasnikovdaniil**](https://github.com/myasnikovdaniil)
+* [**@sircthulhu**](https://github.com/sircthulhu)
+* [**@tym83**](https://github.com/tym83)
+
+---
+
+**Full Changelog**: https://github.com/cozystack/cozystack/compare/v1.2.0...v1.3.0-rc.1
diff --git a/docs/changelogs/v1.3.0.md b/docs/changelogs/v1.3.0.md
new file mode 100644
index 00000000..60c800a8
--- /dev/null
+++ b/docs/changelogs/v1.3.0.md
@@ -0,0 +1,238 @@
+<!--
+https://github.com/cozystack/cozystack/releases/tag/v1.3.0
+-->
+
+# Cozystack v1.3.0
+
+Cozystack v1.3.0 brings **storage-aware pod scheduling** via a LINSTOR scheduler extender, a managed **LINSTOR GUI** web console with Keycloak SSO, a curated **VM Default Images** catalog for out-of-the-box virtual-machine provisioning, a new **WorkloadsReady / Events** observability surface with S3 bucket metering, and **cross-namespace VMInstance backup restore** with a full **RestoreJob dashboard** flow. The release also ships stricter tenant-name validation, VMInstance network-selector improvements, Keycloak theme injection and SMTP configuration, a host-runtime preflight check, and rolls up every fix from the v1.2.1 → v1.2.4 patch line.
+
+> **Note:** Items marked *(backported to v1.2.x)* were also shipped in v1.2.1, v1.2.2, v1.2.3, or v1.2.4 patch releases.
+
+## Feature Highlights
+
+### Storage-Aware Scheduling via the LINSTOR Extender
+
+The `cozystack-scheduler` now calls a **LINSTOR scheduler extender** for storage-locality-aware pod placement. When a pod declares both a `SchedulingClass` and LINSTOR-backed PVCs, the scheduler consults LINSTOR to prefer nodes where volume replicas already exist — reducing cross-node replication traffic and improving I/O latency for storage-heavy workloads such as databases, object stores, and VMs.
+
+The integration builds on the existing `SchedulingClass` tenant workload placement system introduced in v1.2.0 and requires no tenant-side configuration — workloads simply benefit once a SchedulingClass is assigned. Administrators can mix storage locality with the existing data-center / hardware-generation constraints defined on SchedulingClass CRs ([**@lllamnyp**](https://github.com/lllamnyp) in #2330).
+
+### LINSTOR GUI: Managed Web Console for Storage Administration
+
+A new opt-in `linstor-gui` system package deploys **LINBIT's linstor-gui web UI** alongside the LINSTOR controller with mTLS client authentication, non-root security context, and a ClusterIP-only service by default. When OIDC is configured on the platform, an optional **Keycloak-protected Ingress** (via oauth2-proxy) exposes the UI for browser access. Access is restricted to members of the `cozystack-cluster-admin` Keycloak group, consistent with host-cluster admin RBAC, and the gatekeeper blocks in-app LINSTOR authentication setup at the nginx proxy layer so the managed configuration cannot be subverted through the UI.
+
+Operators who prefer CLI access keep the existing `linstor` command; the GUI is strictly additive and stays disabled by default ([**@myasnikovdaniil**](https://github.com/myasnikovdaniil) in #2382, #2390, #2415, #2419).
+
+### VM Default Images: Out-of-the-Box VM Provisioning
+
+The new `vm-default-images` package provides a curated set of **cluster-wide virtual-machine images** (Ubuntu, Debian, CentOS Stream, and others) as pre-populated DataVolumes, so tenants can provision VMs against well-known base images without first having to upload them. The package is opt-in via the `iaas` bundle and defaults to replicated storage for high availability. Migration 38 renames legacy `vm-image-*` DataVolumes to the new `vm-default-images-*` naming scheme, and the `vm-disk` chart gains a new "disk" source type for cloning from existing vm-disks in the same namespace ([**@myasnikovdaniil**](https://github.com/myasnikovdaniil) in #2258).
+
+### Application Observability: WorkloadsReady, Events, and S3 Bucket Metering
+
+Applications now expose a **WorkloadsReady** condition on their status by querying associated WorkloadMonitor resources, giving operators a single place to check whether all underlying workloads (Deployments, StatefulSets, DaemonSets, PVCs) are healthy. The dashboard gains a new **Events tab** showing namespace-scoped Kubernetes events per application, with fallback to `.firstTimestamp` when `.eventTime` is absent. A long-standing bug where WorkloadMonitor's `Operational` status was never persisted is fixed in the same change ([**@lexfrei**](https://github.com/lexfrei) in #2356).
+
+The WorkloadMonitor reconciler is extended to track **COSI BucketClaim** objects as first-class Workloads, and the bucket controller now queries SeaweedFS logical and physical bucket-size metrics from VictoriaMetrics via a namespace-scoped monitoring endpoint, enabling S3 billing integration on par with Pods and PVCs ([**@kitsunoff**](https://github.com/kitsunoff) in #2391). Workloads are also enriched with `workloads.cozystack.io/resource-preset` and source-object labels so downstream billing pipelines can correlate monitors with the tenant preset that produced them ([**@androndo**](https://github.com/androndo) in #2416).
+
+### Cross-Namespace VM Backup Restore and RestoreJob Dashboard
+
+The backup system now supports **restoring VMInstance backups into a different namespace** (cross-namespace copy restores) with IP/MAC preservation and safe rename semantics. In-place backup and restore flows for VMDisk and VMInstance are improved: HelmReleases and DataVolumes are properly handled, and Velero failure messages are propagated to the Application status. The backup status structure has been refactored to store underlying resources as a generic opaque JSON object, enabling arbitrary application-specific metadata without status-schema churn ([**@androndo**](https://github.com/androndo) in #2251, #2319, #2329).
+
+The dashboard now ships a complete **RestoreJob experience**: list view, details page, create form, and sidebar entry, with a "Same as backup" fallback rendering when `spec.targetApplicationRef` is omitted. Non-CRD-backed sidebar factories (`kube-*`, `plan`, `backupjob`, `backup`, `restorejob`) are marked static so they pick up consistent managed-by labels across reconciles ([**@myasnikovdaniil**](https://github.com/myasnikovdaniil) in #2437).
+
+## Major Features and Improvements
+
+* **[api] Reject tenant names with dashes at Create time**: Enforces alphanumeric-only naming for Tenants at the API level, preventing names with hyphens that would silently fail during Helm reconciliation. A corresponding regex tightening and regression test suite hardens the validation ([**@lexfrei**](https://github.com/lexfrei) in #2380).
+
+* **[platform] Validate computed tenant namespace length**: Rejects Tenant creation when the computed ancestor-chain namespace would exceed the 63-character Kubernetes namespace limit, preventing opaque HelmRelease reconcile errors downstream ([**@lexfrei**](https://github.com/lexfrei) in #2376).
+
+* **[vm-instance] Rename subnets to networks and add dropdown selector**: Renames the misleading `subnets` field to `networks` in VMInstance for clarity, adds a dropdown selector for available networks in the dashboard form, and includes migration 36 to copy existing `subnets` values. The old field remains supported for backward compatibility ([**@sircthulhu**](https://github.com/sircthulhu) in #2263).
+
+* **[keycloak] Enable injecting themes**: Cozystack administrators can now inject custom Keycloak themes via `initContainers` for UI white-labeling and customization ([**@lllamnyp**](https://github.com/lllamnyp) in #2142).
+
+* **[keycloak-configure] Add email verification and SMTP configuration**: Adds configurable Keycloak settings for user self-registration, email verification, and SMTP server configuration, enabling automated user onboarding flows ([**@BROngineer**](https://github.com/BROngineer) in #2318).
+
+* **[postgres] Pin system PostgreSQL to 17.7-standard-trixie**: Pins the PostgreSQL image for system databases (Grafana, Alerta, Harbor, Keycloak, SeaweedFS) to `17.7-standard-trixie` across chart templates and `values.yaml`, and ships migration 37 to patch existing CNPG Cluster `imageName` fields to the same variant (handling unset, any PG 17 tag, and bare-version tags). This prevents CNPG from defaulting to PostgreSQL 18 and locks system databases to the trixie variant consistent with the monitoring stack requirements *(related backports shipped in v1.2.1 via #2309 and v1.2.2 via #2364)* ([**@myasnikovdaniil**](https://github.com/myasnikovdaniil) in #2369).
+
+* **[platform] Prevent installed packages deletion**: Adds the `helm.sh/resource-policy: keep` annotation to platform packages so disabling a package no longer triggers automatic Helm deletion, restoring the documented behavior where operators must explicitly delete a package *(backported to v1.2.1)* ([**@kvaps**](https://github.com/kvaps) in #2273).
+
+* **[mariadb] Always enable replication for consistent service naming**: MariaDB now always enables replication, creating `-primary`/`-secondary` services even for single-replica instances. This fixes dashboard visibility and backup functionality for single-replica setups ([**@sircthulhu**](https://github.com/sircthulhu) in #2279).
+
+* **[hack] Add host runtime preflight check**: New `check-host-runtime.sh` script and `make preflight` target that warns operators when a standalone containerd or docker runtime is running alongside the embedded k3s runtime, helping diagnose container-runtime conflicts early in an installation ([**@lexfrei**](https://github.com/lexfrei) in #2371).
+
+* **[hack] Add check-readiness.sh diagnostic script**: A new diagnostic script for tracking platform reconciliation by checking readiness of Packages, ArtifactGenerators, ExternalArtifacts, and HelmReleases, with support for watch mode and continuous monitoring ([**@myasnikovdaniil**](https://github.com/myasnikovdaniil) in #2294).
+
+* **[platform] Add resourcePreset labels to WorkloadMonitor labels**: WorkloadMonitor labels with the `workloads.cozystack.io/` prefix are now propagated onto created Workloads; created Workloads always include the reserved `workloads.cozystack.io/monitor` label, and Helm app charts add `workloads.cozystack.io/resource-preset` metadata to WorkloadMonitor manifests, enabling downstream billing pipelines to correlate monitors with the tenant preset that produced them ([**@androndo**](https://github.com/androndo) in #2416).
+
+## Bug Fixes
+
+* **[platform] Migrate ACME HTTP-01 to ingressClassName API**: Switches ACME HTTP-01 issuance from the deprecated `acme.cert-manager.io/http01-ingress-class` annotation to the modern `ingressClassName` field on `ClusterIssuer` and solver pods. Previously, ClusterIssuers referenced a non-existent `nginx` class while each Ingress individually overrode it via annotation — producing `ingressClassName and class cannot be set at the same time` errors when tenants attempted to migrate to the modern field. The migration is atomic: both the ClusterIssuer and consuming Ingresses are updated together *(backported to v1.2.4)* ([**@myasnikovdaniil**](https://github.com/myasnikovdaniil) in #2436).
+
+* **[harbor] Remove incorrect tenant module flags**: Harbor is a PaaS service, not a tenant module. Incorrect `spec.dashboard.module: true` and `internal.cozystack.io/tenantmodule` flags caused Harbor to appear in the sidebar "Modules" section and be misclassified by controllers handling tenant modules. The flags are now removed so Harbor is displayed in its proper PaaS category and is no longer treated as a tenant-scoped HelmRelease ([**@kvaps**](https://github.com/kvaps) in #2444).
+
+* **[kube-ovn] Resolve kubeovn-plunger RBAC forbidden on deployments**: Grants `kube-ovn-plunger` the RBAC needed to list Deployments so it can reconcile `ovn-central`, fixing `deployments.apps is forbidden` errors in `cozy-kubeovn` ([**@kvaps**](https://github.com/kvaps) in #2441).
+
+* **[cilium] Opt-out of cri-containerd.apparmor.d for nsenter init containers**: Opts cilium-agent init containers out of the `cri-containerd.apparmor.d` AppArmor profile on non-Talos variants (`cilium-generic`, `kubeovn-cilium-generic`), fixing `Init:CrashLoopBackOff` on Ubuntu 22.04+ and Debian where the profile denies `nsenter` namespace entry. Talos variants are untouched as Talos does not load the AppArmor LSM *(backported to v1.2.2)* ([**@lexfrei**](https://github.com/lexfrei) in #2370).
+
+* **[virtual-machine] Exclude external VM services from Cilium BPF LB**: Adds the `service.kubernetes.io/service-proxy-name: cozy-proxy` label to VM LoadBalancer services with `external: true`, telling Cilium to skip BPF processing entirely. Fixes inter-tenant connectivity via public LB IPs (Cilium's DNAT caused cross-tenant pod-to-pod flow classification, triggering CiliumClusterwideNetworkPolicy blocks) and restores WholeIP behavior on Cilium 1.19+ where wildcard service drop entries previously blocked traffic to LB IPs on undeclared ports *(backported to v1.2.2)* ([**@mattia-eleuteri**](https://github.com/mattia-eleuteri) in #2357).
+
+* **[monitoring] Fix infra dashboards missing in default variant**: Includes the `cozy-monitoring` namespace in the dashboard rendering condition, fixing infrastructure Grafana dashboards not rendering in the default platform variant (only the `tenant-root` namespace was previously checked) *(backported to v1.2.2)* ([**@mattia-eleuteri**](https://github.com/mattia-eleuteri) in #2365).
+
+* **[build] Filter git describe to match only v* tags**: Adds `--match 'v*'` to all `git describe` calls in `hack/common-envs.mk`, preventing the `api/apps/v1alpha1/vX.Y.Z` subtag from being picked up instead of the release tag and producing invalid Docker image tags *(backported to v1.2.2)* ([**@kvaps**](https://github.com/kvaps) in #2386).
+
+* **[platform] Fix resource allocation ratios not propagated to packages**: Restores propagation of `cpuAllocationRatio`, `memoryAllocationRatio`, and `ephemeralStorageAllocationRatio` from `platform/values.yaml` to the `cozystack-values` Secret that managed applications and KubeVirt read, fixing a regression introduced in the bundle restructure that silently ignored operator-configured ratios *(backported to v1.2.1)* ([**@sircthulhu**](https://github.com/sircthulhu) in #2296).
+
+* **[kubernetes] Set explicit ephemeral-storage on virt-launcher pods**: Sets explicit `domain.resources` ephemeral-storage on the VirtualMachine spec to prevent virt-launcher pods from being evicted because LimitRange defaults were too small for the actual emptyDisk capacity *(backported to v1.2.3)* ([**@kvaps**](https://github.com/kvaps) in #2317).
+
+* **[multus] Pin master CNI to 05-cilium.conflist**: Prevents a boot-time race where multus could auto-detect kube-ovn's conflist instead of Cilium's, which would cause pods to bypass the Cilium chain entirely and lose their endpoint *(backported to v1.2.1)* ([**@kvaps**](https://github.com/kvaps) in #2315).
+
+* **[multus] Build custom image with DEL cache fix**: Fixes sandbox cleanup deadlock when CNI ADD never completes, preventing stale sandbox name reservations from permanently blocking pod creation after a node disruption *(backported to v1.2.1)* ([**@kvaps**](https://github.com/kvaps) in #2313).
+
+* **[linstor] Set verify-alg to crc32c**: Prevents DRBD connection failures on kernels where `crct10dif` is unavailable (e.g., Talos v1.12.6 with kernel 6.18.18) by setting the LINSTOR verify-alg controller default to `crc32c` *(backported to v1.2.1)* ([**@kvaps**](https://github.com/kvaps) in #2303).
+
+* **[linstor] Preserve TCP ports during toggle-disk operations**: Saves existing TCP ports into the `LayerPayload` before `removeLayerData()` deletes them, preventing DRBD resources from entering StandAlone state when a satellite misses the resulting update *(backported to v1.2.1)* ([**@kvaps**](https://github.com/kvaps) in #2292).
+
+* **[linstor] Increase satellite startup probe failure threshold**: Raises the LINSTOR satellite `startupProbe` `failureThreshold` from 3 to 30 (30s → 300s) in the `LinstorSatelliteConfiguration` pod template, giving satellites with slow storage initialization enough time to come up without being killed and restarted ([**@Arsolitt**](https://github.com/Arsolitt) in #2425).
+
+## Security
+
+* **docs: add SECURITY.md**: Adds vulnerability reporting procedures, disclosure expectations, and supported release lines ([**@kvaps**](https://github.com/kvaps) in #2230).
+
+* **docs: add OpenSSF Best Practices badge to README**: Adds the OpenSSF Best Practices passing badge to the project README ([**@lexfrei**](https://github.com/lexfrei) in #2320).
+
+## Dependencies & Version Updates
+
+* **[kube-ovn] Bump kube-ovn to v1.15.10 with port-group regression fix**: Updates `packages/system/kubeovn` to upstream v1.15.10 (from v1.15.3) and carries a patch for `pkg/controller/pod.go` that preserves a VM LSP's port-group memberships when Kubernetes GCs a completed virt-launcher pod while another virt-launcher pod of the same VM is still running. Without the patch, the destination pod of a successful live migration lost its security groups, network policies, and node-scoped routing until `kube-ovn-controller` was restarted ([**@kvaps**](https://github.com/kvaps) in #2443).
+
+* **[monitoring] Upgrade victoria-metrics-operator to v0.68.4**: Bumps the vendored `victoria-metrics-operator` Helm chart from 0.59.1 to 0.61.0 (operator appVersion v0.68.1 → v0.68.4), picking up upstream fixes for `VMPodScrape` port routing on VMAgent/VLAgent and `StatefulSet` pod deletion (not eviction) when `maxUnavailable=100%` ([**@lexfrei**](https://github.com/lexfrei) in #2426).
+
+* **[linstor] Update piraeus-server to v1.33.2 with selected backports**: Bumps LINSTOR server from v1.33.1 to v1.33.2 with backported patches for stale bitmap adjust retry, LUKS2 header sizing, optimal I/O size detection, and the maintainer implementation. All patches verified against upstream v1.33.2 with `git apply --check` and `gradlew compileJava` *(backported to v1.2.2)* ([**@kvaps**](https://github.com/kvaps) in #2331).
+
+* **[kamaji] Update to 26.3.5-edge, drop upstreamed patches**: Updates Kamaji from edge-26.2.4 to 26.3.5-edge and removes two patches accepted upstream. Adds configurable probe tuning and DataStore readiness conditions ([**@myasnikovdaniil**](https://github.com/myasnikovdaniil) in #2260).
+
+* **[talm] Release v0.23.0, v0.23.1, v0.24.0** (github.com/cozystack/talm): Migrates to the Talos v1.12 multi-document machine config format ([**@lexfrei**](https://github.com/lexfrei) in cozystack/talm#116); renders templates online in `apply` to resolve lookups ([**@myasnikovdaniil**](https://github.com/myasnikovdaniil) in cozystack/talm#119); bumps dependencies and modernizes the codebase ([**@lexfrei**](https://github.com/lexfrei) in cozystack/talm#124).
+
+* **[ansible-cozystack] Release v1.2.1, v1.2.2, v1.2.4** (github.com/cozystack/ansible-cozystack): Exposes `publishing.externalIPs` and tenant-root ingress via role variables ([**@lexfrei**](https://github.com/lexfrei) in cozystack/ansible-cozystack#30); adds a comprehensive node prerequisites audit ([**@lexfrei**](https://github.com/lexfrei) in cozystack/ansible-cozystack#27); replaces `ansible.utils.ipaddr` with a stdlib-based test plugin ([**@lexfrei**](https://github.com/lexfrei) in cozystack/ansible-cozystack#24); adds `v` prefix to collection version in requirements.yml examples ([**@lexfrei**](https://github.com/lexfrei) in cozystack/ansible-cozystack#23); tracks installer releases v1.2.1 through v1.2.4 ([**@app/renovate**](https://github.com/apps/renovate) in cozystack/ansible-cozystack#20, #22, #29, #31, #32).
+
+## Development, Testing, and CI/CD
+
+* **[ci] Replace cozystack-bot PAT with cozystack-ci GitHub App**: Replaces the long-lived `cozystack-bot` personal access token with short-lived, scoped tokens from the `cozystack-ci` GitHub App across all release workflows (`tags.yaml`, `auto-release.yaml`, `pull-requests-release.yaml`), improving security and auditability of CI operations ([**@tym83**](https://github.com/tym83) in #2351; [**@kvaps**](https://github.com/kvaps) in #2383, #2392).
+
+* **[ci] Add Gemini Code Assist and CodeRabbit configuration**: Adds repository-level configuration for AI code reviewers with ignore patterns for vendored/generated code and incremental review settings ([**@lexfrei**](https://github.com/lexfrei) in #2385).
+
+* **[ci] Promote next/ trunk on new minor/major releases**: Updates `update-website-docs` in `tags.yaml` to match the new docs-versioning contract — the website repo replaces the old "pre-create `vX.Y/` draft directory" scheme with a permanent `content/en/docs/next/` trunk, and released version directories are promoted explicitly by the release workflow ([**@myasnikovdaniil**](https://github.com/myasnikovdaniil) in #2433).
+
+* **[tests] Fix Kafka E2E test timeout and retry race condition**: Increases Kafka E2E test timeout from 60s to 300s and fixes a retry race where `kubectl apply` could hit a still-deleting resource ([**@lexfrei**](https://github.com/lexfrei) in #2358).
+
+* **docs: adopt Conventional Commits for commit and PR titles**: Standardizes commit and PR title format to `type(scope): description` across all contributing docs and the PR template ([**@lexfrei**](https://github.com/lexfrei) in #2395).
+
+* **docs(ci): require screenshots for UI changes in PR template**: Adds a mandatory screenshots section to the PR template for UI-related changes ([**@kitsunoff**](https://github.com/kitsunoff) in #2407).
+
+* **chore(maintenance): add @myasnikovdaniil to CODEOWNERS**: Adds @myasnikovdaniil to the default owners in `.github/CODEOWNERS` for automatic review requests ([**@myasnikovdaniil**](https://github.com/myasnikovdaniil) in #2434).
+
+## Documentation
+
+* **[website] Add ApplicationDefinition naming convention reference**: Documents how `cozystack-api` resolves kinds to their backing definitions ([**@lexfrei**](https://github.com/lexfrei) in cozystack/website#478).
+
+* **[website] Document Talos / talosctl / Cozystack version pairing**: Adds a version compatibility matrix for installation ([**@lexfrei**](https://github.com/lexfrei) in cozystack/website#484).
+
+* **[website] Document namespace layout and parent/child derivation**: Explains tenant namespace hierarchy and parent/child namespace derivation rules ([**@lexfrei**](https://github.com/lexfrei) in cozystack/website#479).
+
+* **[website] Document the checkbox-then-edit-CR customization pattern for tenants**: Describes the workflow for customizing tenant settings via the CR after initial checkbox-based creation ([**@lexfrei**](https://github.com/lexfrei) in cozystack/website#485).
+
+* **[website] Add custom Keycloak themes documentation**: Covers the theme image contract, configuration, `imagePullSecrets`, and theme activation in the Keycloak admin console ([**@lexfrei**](https://github.com/lexfrei) in cozystack/website#463).
+
+* **[website] Add bonding (LACP) configuration how-to guide**: Covers network bonding configuration for Cozystack installations ([**@sircthulhu**](https://github.com/sircthulhu) in cozystack/website#459).
+
+* **[website] Improve registry mirrors for tenant Kubernetes in air-gapped guide**: Improves documentation for configuring registry mirrors in air-gapped environments ([**@sircthulhu**](https://github.com/sircthulhu) in cozystack/website#461).
+
+* **[website] Rewrite guide for ApplicationDefinition API (external-apps)**: Comprehensive rewrite of the external apps guide using the ApplicationDefinition API with Minecraft server examples ([**@kitsunoff**](https://github.com/kitsunoff) in cozystack/website#488).
+
+* **[website] Add documentation for Go types usage**: Guide for using generated Go types for Cozystack managed applications as a Go module ([**@myasnikovdaniil**](https://github.com/myasnikovdaniil) in cozystack/website#465).
+
+* **[website] Update backup/restore documentation for VMI/VMDisk**: Updates backup documentation with VM instance and VM disk restore improvements ([**@androndo**](https://github.com/androndo) in cozystack/website#466).
+
+* **[website] Refactor docs versions to major.minor variants**: Moves docs to major.minor versioning for the v1.x series ([**@myasnikovdaniil**](https://github.com/myasnikovdaniil) in cozystack/website#477).
+
+* **[website] Trunk-based versioning with permanent next/ directory**: Replaces the old "pre-create `vX.Y/` draft directory" scheme with a permanent `content/en/docs/next/` trunk; released version directories are promoted explicitly by `hack/release_next.sh` on new minor/major releases, and routing between `next/` and `vX.Y/` is Makefile-driven ([**@myasnikovdaniil**](https://github.com/myasnikovdaniil) in cozystack/website#495).
+
+* **[website] Add updated OpenAPI spec**: Updates the OpenAPI specification for managed applications reference ([**@myasnikovdaniil**](https://github.com/myasnikovdaniil) in cozystack/website#469).
+
+* **[website] Add OpenAPI spec download to GitHub Pages build**: Fixes the GitHub Pages build to include the OpenAPI spec download ([**@myasnikovdaniil**](https://github.com/myasnikovdaniil) in cozystack/website#494).
+
+* **[website] Add OSS Health pages and OpenSSF badge**: Adds OSS Health section with OpenSSF Scorecard and Best Practices badges to the website ([**@tym83**](https://github.com/tym83) in cozystack/website#470).
+
+* **[website] Add Telemetry page under OSS Health section**: Adds the Telemetry page with initial data seeding to the OSS Health docs ([**@tym83**](https://github.com/tym83) in cozystack/website#471, cozystack/website#504).
+
+* **[website] Blog: OSS Health section launch announcement**: Publishes the announcement blog post for the OSS Health section ([**@tym83**](https://github.com/tym83) in cozystack/website#474).
+
+* **[website] Fix OpenSSF canonical status URL**: Changes the OpenSSF canonical status URL from pt-BR to en ([**@tym83**](https://github.com/tym83) in cozystack/website#475).
+
+* **[website] Add CozySummit Virtual 2026 program announcement**: Publishes the CozySummit Virtual 2026 program announcement blog post ([**@tym83**](https://github.com/tym83) in cozystack/website#472).
+
+* **[website] Add missing release announcements for v0.1–v0.41**: Backfills missing release announcement blog posts for historical Cozystack versions ([**@tym83**](https://github.com/tym83) in cozystack/website#468).
+
+* **[website] Blog: managed PostgreSQL with synchronous replication**: Adds a post covering the managed PostgreSQL synchronous-replication feature ([**@tym83**](https://github.com/tym83) in cozystack/website#497).
+
+* **[website] Blog taxonomies and client-side filter UI**: Registers article-type and topic taxonomies and adds a client-side filter on the blog list page ([**@tym83**](https://github.com/tym83) in cozystack/website#499).
+
+* **[website] Add images frontmatter for social preview on existing posts**: Adds images frontmatter for social preview on existing blog posts ([**@tym83**](https://github.com/tym83) in cozystack/website#498).
+
+* **[website] Fix broken links and stale anchors across v1 docs**: Fixes 14 broken links and stale talm anchors ([**@lexfrei**](https://github.com/lexfrei) in cozystack/website#486).
+
+* **[website] Prefix bundle package names with cozystack. in v1 examples**: Corrects package naming in documentation examples ([**@lexfrei**](https://github.com/lexfrei) in cozystack/website#482).
+
+* **[website] Finish isolated-field removal and document opt-in policy labels**: Removes the obsolete `isolated` field from tenant documentation and documents the new opt-in policy labels approach ([**@lexfrei**](https://github.com/lexfrei) in cozystack/website#481).
+
+* **[website] Add --take-ownership flag and describe networking.* fields**: Documents the `--take-ownership` flag and `networking.*` fields in the installation guide ([**@lexfrei**](https://github.com/lexfrei) in cozystack/website#480).
+
+* **[website] Fix KubeOVN MASTER_NODES example path and key in troubleshooting**: Corrects the MASTER_NODES example path and key ([**@lexfrei**](https://github.com/lexfrei) in cozystack/website#483).
+
+* **[website] Add CLAUDE.md for AI agent guidance**: Adds a CLAUDE.md file describing the trunk-based docs architecture for AI agent guidance ([**@myasnikovdaniil**](https://github.com/myasnikovdaniil) in cozystack/website#489).
+
+* **[website] Update /docs/v1/ redirect to latest v1.2**: Updates the `/docs/v1/` redirect target to point to the latest v1.2 docs on GitHub Pages ([**@myasnikovdaniil**](https://github.com/myasnikovdaniil) in cozystack/website#492).
+
+* **[website] Remove nbykov from CODEOWNERS and CLAUDE.md**: Cleans up CODEOWNERS and CLAUDE.md entries ([**@myasnikovdaniil**](https://github.com/myasnikovdaniil) in cozystack/website#491).
+
+* **[website] Add Ahrefs Analytics tracker**: Adds the Ahrefs Analytics tracker to the website ([**@tym83**](https://github.com/tym83) in cozystack/website#503).
+
+* **[website] Add breathing room between navbar and hero on OSS Health**: Minor styling fix for the OSS Health section ([**@tym83**](https://github.com/tym83) in cozystack/website#500).
+
+* **[website] Fix og social badge image and title**: Updates the social badge image and title ([**@tym83**](https://github.com/tym83) in cozystack/website#487).
+
+* **[website] Update managed apps reference for v1.2.1**: Automated managed-apps reference update ([**@cozystack-bot**](https://github.com/cozystack-bot) in cozystack/website#464).
+
+* **[external-apps-example] Replace MongoDB example with Minecraft apps**: Refactors the external apps example to use the ApplicationDefinition API with Minecraft server applications ([**@lexfrei**](https://github.com/lexfrei) in cozystack/external-apps-example#2).
+
+* **docs: update README introductory description**: Refines the platform positioning and improves clarity on core capabilities in the main README ([**@tym83**](https://github.com/tym83) in #2409).
+
+## Governance
+
+* **Add Mattia Eleuteri ([@mattia-eleuteri](https://github.com/mattia-eleuteri)) as Maintainer**: CSI, Storage, Networking & Security ([**@tym83**](https://github.com/tym83) in #2345).
+
+* **Add Matthieu Robin ([@matthieu-robin](https://github.com/matthieu-robin)) as Maintainer**: Managed applications, platform quality, and benchmarking ([**@tym83**](https://github.com/tym83) in #2346).
+
+## Contributors
+
+We'd like to thank all contributors who made this release possible:
+
+* [**@androndo**](https://github.com/androndo)
+* [**@Arsolitt**](https://github.com/Arsolitt)
+* [**@BROngineer**](https://github.com/BROngineer)
+* [**@IvanHunters**](https://github.com/IvanHunters)
+* [**@kitsunoff**](https://github.com/kitsunoff)
+* [**@kvaps**](https://github.com/kvaps)
+* [**@lexfrei**](https://github.com/lexfrei)
+* [**@lllamnyp**](https://github.com/lllamnyp)
+* [**@mattia-eleuteri**](https://github.com/mattia-eleuteri)
+* [**@myasnikovdaniil**](https://github.com/myasnikovdaniil)
+* [**@sircthulhu**](https://github.com/sircthulhu)
+* [**@tym83**](https://github.com/tym83)
+
+### New Contributors
+
+We're excited to welcome our first-time contributors:
+
+* [**@Arsolitt**](https://github.com/Arsolitt) — First contribution!
+
+---
+
+**Full Changelog**: https://github.com/cozystack/cozystack/compare/v1.2.0...v1.3.0
diff --git a/docs/changelogs/v1.3.1.md b/docs/changelogs/v1.3.1.md
new file mode 100644
index 00000000..0f149d0d
--- /dev/null
+++ b/docs/changelogs/v1.3.1.md
@@ -0,0 +1,37 @@
+<!--
+https://github.com/cozystack/cozystack/releases/tag/v1.3.1
+-->
+
+# v1.3.1 (2026-04-28)
+
+Patch release covering a TenantNamespace IDOR fix in the API, a destructive `post-upgrade` hook removed from the etcd chart, kamaji controller stability, a `linstor-csi` bump that fixes live migration on Protocol-A/B DRBD resources, the missing `linstor-gui` build wiring, and a velero RBAC fix that unblocked installs on bundles without Velero.
+
+## Security
+
+* **fix(api): prevent IDOR in TenantNamespace Get and Watch handlers**: Two IDOR (Insecure Direct Object Reference) vulnerabilities allowed authenticated users to read TenantNamespace metadata they had no RoleBinding for. The `Get` and `Watch` handlers now go through a new `hasAccessToNamespace()` helper that lists RoleBindings scoped only to the target namespace (orders of magnitude faster than the previous all-cluster scan), returns `NotFound` instead of leaking existence on unauthorized access, and applies the same check on the `Watch` filter path. Includes regression tests for the unauthorized paths. ([**@IvanHunters**](https://github.com/IvanHunters) in #2471, backport #2524)
+
+## Features
+
+* **feat(linstor): bump linstor-csi to v1.10.6 with Protocol-C dual-attach fix**: Live migration of KubeVirt VMs on Protocol-A/B (async) DRBD volumes no longer fails with `Protocol C required`. `linstor-csi` v1.10.6 now installs a `Protocol=C` override on the resource-definition during dual-attach and reverts it on detach, so `replicated-async` StorageClasses and other Protocol-A/B resource groups support live migration without manual `drbdadm adjust` intervention. ([**@kvaps**](https://github.com/kvaps) in #2496, backport #2505)
+
+## Fixes
+
+* **fix(backups): move velero-configmap Role to velero chart**: The `backupstrategy-controller` (a default package) declared a Role/RoleBinding scoped to the `cozy-velero` namespace for managing `ResourceModifier` ConfigMaps. On bundles where Velero was not enabled, that namespace did not exist and the HelmRelease failed with `namespaces "cozy-velero" not found`, blocking installation. The Role/RoleBinding now lives in the velero chart, so it is created only when velero is actually deployed. ([**@myasnikovdaniil**](https://github.com/myasnikovdaniil) in #2459, backport #2467)
+
+* **fix(etcd): remove destructive post-upgrade cert-regeneration hook**: The etcd chart ran a `post-upgrade` Helm hook on every upgrade that deleted etcd TLS Secrets (`etcd-ca-tls`, `etcd-peer-ca-tls`, `etcd-client-tls`, `etcd-peer-tls`, `etcd-server-tls`) and then deleted etcd pods, forcing cert-manager to re-issue the entire etcd CA chain. On clusters with Kamaji-managed tenant control planes this put every tenant `kube-apiserver` into CrashLoopBackOff until each DataStore was manually re-reconciled. The hook was a one-shot `2.6.0 → 2.6.1` migration that became a permanent footgun once chart versioning moved to `0.0.0+<git-hash>` (always `< 2.6.1` per semver) and after the underlying `rotationPolicy: Always` issue was fixed in `47d81f70`. The hook is now removed entirely. ([**@myasnikovdaniil**](https://github.com/myasnikovdaniil) in #2462, backport #2511)
+
+* **fix(kamaji): increase memory limits and add startup probe**: The kamaji controller frequently entered CrashLoopBackOff due to OOMKills (exit 137) within ~20–25 seconds of startup, with the readiness probe failing while the controller was still finishing initialization. Memory limit raised from 500Mi to 512Mi, request from 100Mi to 256Mi, and a 60-second startup probe (12 attempts × 5s periods) is added so the controller has room to boot before liveness/readiness probes engage. ([**@IvanHunters**](https://github.com/IvanHunters) in #2421, backport #2491)
+
+## Build
+
+* **build(linstor): include linstor-gui in root image build target**: The `linstor-gui` package (added in #2382) was never wired into the root `Makefile`'s `build:` target, so CI never built or published the image. `ghcr.io/cozystack/cozystack/linstor-gui` returned `NAME_UNKNOWN` and `values.yaml` stayed pinned to `tag: 2.3.0` without a digest. The missing build line is added so the next CI run publishes the image and the per-package `Makefile` digest-pins `values.yaml` automatically. ([**@myasnikovdaniil**](https://github.com/myasnikovdaniil) in #2498, backport #2518)
+
+## Contributors
+
+Thanks to everyone who contributed to this patch release:
+
+* [**@IvanHunters**](https://github.com/IvanHunters)
+* [**@kvaps**](https://github.com/kvaps)
+* [**@myasnikovdaniil**](https://github.com/myasnikovdaniil)
+
+**Full Changelog**: https://github.com/cozystack/cozystack/compare/v1.3.0...v1.3.1
diff --git a/go.mod b/go.mod
index 00e8e129..7b7d5bc4 100644
--- a/go.mod
+++ b/go.mod
@@ -5,6 +5,7 @@ module github.com/cozystack/cozystack
 go 1.25.0
 
 require (
+	github.com/cozystack/cozystack-scheduler/pkg/apis v0.1.1
 	github.com/emicklei/dot v1.10.0
 	github.com/fluxcd/helm-controller/api v1.4.3
 	github.com/fluxcd/source-controller/api v1.7.4
@@ -28,8 +29,10 @@ require (
 	k8s.io/klog/v2 v2.130.1
 	k8s.io/kube-openapi v0.0.0-20250710124328-f3f2b991d03b
 	k8s.io/utils v0.0.0-20250820121507-0af2bda4dd1d
+	sigs.k8s.io/container-object-storage-interface-api v0.1.0
 	sigs.k8s.io/controller-runtime v0.22.4
 	sigs.k8s.io/structured-merge-diff/v6 v6.3.0
+	sigs.k8s.io/yaml v1.6.0
 )
 
 require (
@@ -42,10 +45,8 @@ require (
 	github.com/cespare/xxhash/v2 v2.3.0 // indirect
 	github.com/coreos/go-semver v0.3.1 // indirect
 	github.com/coreos/go-systemd/v22 v22.5.0 // indirect
-	github.com/cozystack/cozystack-scheduler/pkg/apis v0.1.1 // indirect
 	github.com/davecgh/go-spew v1.1.2-0.20180830191138-d8f796af33cc // indirect
 	github.com/emicklei/go-restful/v3 v3.12.2 // indirect
-	github.com/evanphx/json-patch v4.12.0+incompatible // indirect
 	github.com/evanphx/json-patch/v5 v5.9.11 // indirect
 	github.com/felixge/httpsnoop v1.0.4 // indirect
 	github.com/fluxcd/pkg/apis/acl v0.9.0 // indirect
@@ -126,7 +127,6 @@ require (
 	sigs.k8s.io/apiserver-network-proxy/konnectivity-client v0.31.2 // indirect
 	sigs.k8s.io/json v0.0.0-20250730193827-2d320260d730 // indirect
 	sigs.k8s.io/randfill v1.0.0 // indirect
-	sigs.k8s.io/yaml v1.6.0 // indirect
 )
 
 // See: issues.k8s.io/135537
diff --git a/go.sum b/go.sum
index 52942562..81e1779f 100644
--- a/go.sum
+++ b/go.sum
@@ -321,6 +321,8 @@ k8s.io/utils v0.0.0-20250820121507-0af2bda4dd1d h1:wAhiDyZ4Tdtt7e46e9M5ZSAJ/MnPG
 k8s.io/utils v0.0.0-20250820121507-0af2bda4dd1d/go.mod h1:OLgZIPagt7ERELqWJFomSt595RzquPNLL48iOWgYOg0=
 sigs.k8s.io/apiserver-network-proxy/konnectivity-client v0.31.2 h1:jpcvIRr3GLoUoEKRkHKSmGjxb6lWwrBlJsXc+eUYQHM=
 sigs.k8s.io/apiserver-network-proxy/konnectivity-client v0.31.2/go.mod h1:Ve9uj1L+deCXFrPOk1LpFXqTg7LCFzFso6PA48q/XZw=
+sigs.k8s.io/container-object-storage-interface-api v0.1.0 h1:8tB6JFQhbQIC1hwGQ+q4+tmSSNfjKemb7bFI6C0CK/4=
+sigs.k8s.io/container-object-storage-interface-api v0.1.0/go.mod h1:YiB+i/UGkzqgODDhRG3u7jkbWkQcoUeLEJ7hwOT/2Qk=
 sigs.k8s.io/controller-runtime v0.22.4 h1:GEjV7KV3TY8e+tJ2LCTxUTanW4z/FmNB7l327UfMq9A=
 sigs.k8s.io/controller-runtime v0.22.4/go.mod h1:+QX1XUpTXN4mLoblf4tqr5CQcyHPAki2HLXqQMY6vh8=
 sigs.k8s.io/json v0.0.0-20250730193827-2d320260d730 h1:IpInykpT6ceI+QxKBbEflcR5EXP7sU1kvOlxwZh5txg=
diff --git a/hack/admin-kubeconfig-invariant.bats b/hack/admin-kubeconfig-invariant.bats
new file mode 100644
index 00000000..a4b98b00
--- /dev/null
+++ b/hack/admin-kubeconfig-invariant.bats
@@ -0,0 +1,145 @@
+#!/usr/bin/env bats
+# -----------------------------------------------------------------------------
+# Chart-wide invariant for packages/apps/kubernetes:
+#
+# Every Deployment in this chart that mounts <release>-admin-kubeconfig as a
+# Secret volume MUST:
+#   - declare that volume optional: true (so kubelet does not FailedMount
+#     while Kamaji is still provisioning the Secret), AND
+#   - include the wait-for-kubeconfig init container (so the pod becomes
+#     Ready only after Kamaji publishes the Secret).
+#
+# The per-template unittests in packages/apps/kubernetes/tests/ lock in
+# today's three Deployments (cluster-autoscaler, kccm, csi controller) by
+# name. This invariant is stricter: any future Deployment added to this
+# chart that mounts the same Secret but forgets the guard will fail here.
+#
+# Requires: helm, yq (mikefarah v4+), jq. All three are available on the
+# project's CI runners and on the maintainer workstation.
+# -----------------------------------------------------------------------------
+
+@test "every Deployment mounting admin-kubeconfig has optional:true and wait-for-kubeconfig init" {
+  values_file="packages/apps/kubernetes/tests/values-ci.yaml"
+  [ -f "$values_file" ]
+
+  tmp=$(mktemp -d)
+  trap 'rm -rf "$tmp"' EXIT
+
+  helm template invariant packages/apps/kubernetes \
+    --namespace tenant-root \
+    --values "$values_file" \
+    2>/dev/null > "$tmp/rendered.yaml"
+
+  # yq streams one JSON object per input document. jq -s slurps the stream
+  # into an array so we can treat all Deployments as a single collection.
+  yq --output-format=json eval-all '.' "$tmp/rendered.yaml" \
+    | jq -s --raw-output '
+        map(select(.kind == "Deployment")) |
+        map({
+          name: .metadata.name,
+          volumes: (.spec.template.spec.volumes // []),
+          initNames: ((.spec.template.spec.initContainers // []) | map(.name)),
+        }) |
+        map(
+          .name as $n |
+          .initNames as $ins |
+          (.volumes[] | select(.secret.secretName | test("-admin-kubeconfig$")?))
+          | {
+              name: $n,
+              optional: (.secret.optional == true),
+              hasInit: ($ins | index("wait-for-kubeconfig") != null),
+            }
+        )
+      ' > "$tmp/summary.json"
+
+  # At least one Deployment must match; if a refactor removes every
+  # admin-kubeconfig volume from this chart, the test must be updated
+  # deliberately rather than silently passing.
+  matched=$(jq 'length' "$tmp/summary.json")
+  [ "$matched" -ge 1 ]
+
+  offenders=$(jq --raw-output '.[] | select(.optional != true or .hasInit != true) | .name' "$tmp/summary.json")
+
+  if [ -n "$offenders" ]; then
+    echo "Deployments mounting *-admin-kubeconfig without optional:true + wait-for-kubeconfig init:" >&2
+    echo "$offenders" >&2
+    echo "Full summary:" >&2
+    cat "$tmp/summary.json" >&2
+    exit 1
+  fi
+
+  echo "Invariant holds for $matched Deployment(s)"
+}
+
+@test "chart emits zero admin-kubeconfig Deployments when tenant has no etcd DataStore" {
+  # Without a DataStore (parent Tenant has not populated _namespace.etcd yet)
+  # the control-plane-side Deployments must NOT render at all. If they did,
+  # the wait-for-kubeconfig init would CrashLoopBackOff indefinitely - there
+  # would be no KamajiControlPlane to provision the Secret - consuming the
+  # HelmRelease wait budget and triggering exactly the remediation cycle the
+  # rest of this chart tries to avoid. This test renders the whole chart
+  # with etcd empty and asserts no Deployment references the admin-kubeconfig
+  # Secret.
+  tmp=$(mktemp -d)
+  trap 'rm -rf "$tmp"' EXIT
+
+  helm template invariant packages/apps/kubernetes \
+    --namespace tenant-root \
+    --values packages/apps/kubernetes/tests/values-ci-no-etcd.yaml \
+    2>/dev/null > "$tmp/rendered.yaml"
+
+  matched=$(
+    yq --output-format=json eval-all '.' "$tmp/rendered.yaml" \
+      | jq -s '
+          map(select(.kind == "Deployment")) |
+          map(select(
+            (.spec.template.spec.volumes // [])
+              | any(.secret.secretName | test("-admin-kubeconfig$")?)
+          )) |
+          length
+        '
+  )
+
+  if [ "$matched" -ne 0 ]; then
+    echo "Expected zero Deployments mounting *-admin-kubeconfig when etcd is empty, got $matched:" >&2
+    yq --output-format=json eval-all '.' "$tmp/rendered.yaml" \
+      | jq -s 'map(select(.kind == "Deployment") | .metadata.name)' >&2
+    exit 1
+  fi
+
+  echo "No admin-kubeconfig Deployments rendered for empty etcd (as expected)"
+}
+
+@test "chart emits zero admin-kubeconfig HelmReleases when tenant has no etcd DataStore" {
+  # Same principle as the Deployment variant above, extended to every child
+  # HelmRelease (cilium, coredns, csi, cert-manager, ...). They reference
+  # *-admin-kubeconfig via kubeConfig.secretRef and would otherwise sit in
+  # NotReady forever on an etcd-less tenant, polluting the HelmRelease list
+  # the operator sees and contradicting the "awaiting-etcd beacon only"
+  # contract of the soft-skip path.
+  tmp=$(mktemp -d)
+  trap 'rm -rf "$tmp"' EXIT
+
+  helm template invariant packages/apps/kubernetes \
+    --namespace tenant-root \
+    --values packages/apps/kubernetes/tests/values-ci-no-etcd.yaml \
+    2>/dev/null > "$tmp/rendered.yaml"
+
+  matched=$(
+    yq --output-format=json eval-all '.' "$tmp/rendered.yaml" \
+      | jq -s '
+          map(select(.kind == "HelmRelease")) |
+          map(select(.spec.kubeConfig.secretRef.name | test("-admin-kubeconfig$")?)) |
+          length
+        '
+  )
+
+  if [ "$matched" -ne 0 ]; then
+    echo "Expected zero HelmReleases referencing *-admin-kubeconfig when etcd is empty, got $matched:" >&2
+    yq --output-format=json eval-all '.' "$tmp/rendered.yaml" \
+      | jq -s 'map(select(.kind == "HelmRelease") | .metadata.name)' >&2
+    exit 1
+  fi
+
+  echo "No admin-kubeconfig HelmReleases rendered for empty etcd (as expected)"
+}
diff --git a/hack/cdi_golden_image_create.sh b/hack/cdi_golden_image_create.sh
index be1558d0..f19f9970 100644
--- a/hack/cdi_golden_image_create.sh
+++ b/hack/cdi_golden_image_create.sh
@@ -16,7 +16,7 @@ kubectl create -f - <<EOF
 apiVersion: cdi.kubevirt.io/v1beta1
 kind: DataVolume
 metadata:
-  name: "vm-image-$name"
+  name: "vm-default-images-$name"
   namespace: cozy-public
   annotations:
     cdi.kubevirt.io/storage.bind.immediate.requested: "true"
diff --git a/hack/check-gpu-recording-rules.bats b/hack/check-gpu-recording-rules.bats
new file mode 100644
index 00000000..bae6785e
--- /dev/null
+++ b/hack/check-gpu-recording-rules.bats
@@ -0,0 +1,206 @@
+#!/usr/bin/env bats
+# -----------------------------------------------------------------------------
+# Cross-validation between GPU recording rules, the dashboards that consume
+# them, and the DCGM Exporter metric set the cluster actually scrapes. Catches:
+#
+#   1. dangling references — a dashboard query mentions a recording rule name
+#      that doesn't exist in gpu-recording.rules.yaml. This is the bug the
+#      pre-merge review caught: gpu-efficiency.json shipped panels keyed on
+#      pod:tensor_saturation:avg5m without the rule being defined, so the
+#      panel showed "No data" everywhere.
+#
+#   2. typos in rule names — same bug class, manifested as a single-character
+#      difference between rule and reference.
+#
+#   3. undeclared DCGM metrics — a dashboard query or recording rule mentions
+#      a DCGM_FI_* metric that is neither in the upstream default CSV nor in
+#      the project's custom CSV (dcgm-custom-metrics.yaml), meaning DCGM
+#      Exporter would never emit it and the panel silently shows "No data".
+#      Example regression: gpu-fleet.json shipped a TDP panel referencing
+#      DCGM_FI_DEV_POWER_MGMT_LIMIT before the custom CSV declared it.
+#
+# Scope: only dashboards listed in packages/system/monitoring/dashboards-infra.list
+# under the "gpu/" prefix (i.e. shipped to production), not every JSON file in
+# dashboards/gpu/. Untracked drafts stay out of scope on purpose — adding one
+# to dashboards-infra.list is what brings it under the test.
+#
+# Reverse direction (rule defined but never consumed) is intentionally NOT
+# enforced: some rules exist for ad-hoc PromQL or upcoming dashboards. Treat
+# unused rules as an editorial concern, not a regression.
+#
+# Title syntax constraints from cozytest.sh's awk parser:
+#   - Titles delimited by ASCII double quotes; embedded quotes truncate.
+#   - Only [A-Za-z0-9] from the title survives into the function name; titles
+#     differing only in punctuation collapse to the same function.
+#
+# Run with: hack/cozytest.sh hack/check-gpu-recording-rules.bats
+# -----------------------------------------------------------------------------
+
+REPO_ROOT="$(cd "$(dirname "${BATS_TEST_FILENAME:-$0}")/.." && pwd)"
+RULES_FILE="$REPO_ROOT/packages/system/monitoring-agents/alerts/gpu-recording.rules.yaml"
+DASHBOARDS_LIST="$REPO_ROOT/packages/system/monitoring/dashboards-infra.list"
+DASHBOARDS_DIR="$REPO_ROOT/dashboards"
+DCGM_DEFAULT_CSV="$REPO_ROOT/hack/dcgm-default-counters.csv"
+DCGM_CUSTOM_CSV="$REPO_ROOT/packages/system/gpu-operator/examples/dcgm-custom-metrics.yaml"
+
+# Extract the set of "- record: NAME" entries from the rules YAML.
+# Outputs one rule name per line, sorted and deduplicated.
+extract_rules() {
+  awk '/^[[:space:]]*-[[:space:]]*record:[[:space:]]/ {
+    sub(/^[[:space:]]*-[[:space:]]*record:[[:space:]]*/, "")
+    sub(/[[:space:]]*$/, "")
+    print
+  }' "$RULES_FILE" | sort -u
+}
+
+# Extract the set of recording-rule references from a dashboard JSON.
+# A recording-rule reference is matched by the pattern
+#   <segment>:<segment>(:<segment>)+
+# where each <segment> is [a-z0-9_]. Raw DCGM metrics (DCGM_FI_*),
+# kube-state-metrics (kube_*) and similar uppercase / single-word metric
+# names do not match because the leading segment must be lowercase and the
+# whole expression must contain at least two ':' characters.
+extract_refs() {
+  json_file=$1
+  # Prometheus convention allows 2-segment rule names (level:metric); this
+  # regex is tuned to the 3+ segment convention used in this repo
+  # (level:metric:op — e.g. cluster:gpu_count:total). Update if future
+  # rules use 2 segments, otherwise they will be silently skipped.
+  grep -hoE '[a-z][a-z0-9_]*:[a-z0-9_]+:[a-z0-9_]+' "$json_file" | sort -u
+}
+
+# Resolve "gpu/foo" -> "$DASHBOARDS_DIR/gpu/foo.json"
+list_tracked_gpu_dashboards() {
+  awk '/^gpu\// { print $0 ".json" }' "$DASHBOARDS_LIST"
+}
+
+# Extract the set of DCGM_FI_* metric names declared in a CSV file. Handles
+# both the upstream-style default CSV (unindented) and the ConfigMap-style
+# custom CSV (YAML-indented). A declaration line starts — after any leading
+# whitespace — with "DCGM_FI_<NAME>," ; comment lines begin with "#" and are
+# skipped. Uses POSIX awk's match()+RSTART/RLENGTH so no GNU extensions
+# are required.
+extract_csv_metrics() {
+  file=$1
+  awk '
+    {
+      line = $0
+      sub(/^[[:space:]]+/, "", line)
+      if (line ~ /^#/) next
+      if (match(line, /^DCGM_FI_[A-Z0-9_]+/)) {
+        print substr(line, RSTART, RLENGTH)
+      }
+    }
+  ' "$file" | sort -u
+}
+
+# Extract the set of DCGM_FI_* metric references from a text file (dashboard
+# JSON or rules YAML). A DCGM metric name has at least two underscore-delimited
+# segments after the "DCGM_FI_" prefix (e.g. DCGM_FI_DEV_GPU_UTIL, DCGM_FI_PROF_
+# PIPE_TENSOR_ACTIVE, DCGM_FI_DRIVER_VERSION). Requiring two segments keeps
+# the matcher from latching onto glob stubs like "DCGM_FI_DEV_*_VIOLATION" that
+# appear in comments.
+extract_dcgm_refs() {
+  file=$1
+  grep -hoE 'DCGM_FI_[A-Z0-9]+(_[A-Z0-9]+)+' "$file" | sort -u
+}
+
+@test "every recording rule reference in tracked GPU dashboards has a matching record" {
+  TMP=$(mktemp -d)
+  trap 'rm -rf "$TMP"' EXIT
+
+  extract_rules > "$TMP/rules.txt"
+  [ -s "$TMP/rules.txt" ] || { echo "no recording rules extracted from $RULES_FILE" >&2; exit 1; }
+
+  list_tracked_gpu_dashboards > "$TMP/dashboards.txt"
+  [ -s "$TMP/dashboards.txt" ] || { echo "no gpu/* dashboards listed in $DASHBOARDS_LIST" >&2; exit 1; }
+
+  failed=0
+  while IFS= read -r dashboard_rel; do
+    dashboard="$DASHBOARDS_DIR/$dashboard_rel"
+    if [ ! -f "$dashboard" ]; then
+      echo "ERROR: dashboard listed but file missing: $dashboard" >&2
+      failed=1
+      continue
+    fi
+
+    extract_refs "$dashboard" > "$TMP/refs.txt"
+    # comm -23: lines unique to refs.txt (referenced but not defined)
+    # Both inputs must be sorted; extract_* helpers already sort.
+    comm -23 "$TMP/refs.txt" "$TMP/rules.txt" > "$TMP/missing.txt"
+    if [ -s "$TMP/missing.txt" ]; then
+      echo "ERROR: $dashboard_rel references undefined recording rules:" >&2
+      sed 's/^/  - /' "$TMP/missing.txt" >&2
+      failed=1
+    fi
+  done < "$TMP/dashboards.txt"
+
+  [ "$failed" -eq 0 ]
+}
+
+@test "every DCGM metric referenced in tracked dashboards and rules is declared" {
+  TMP=$(mktemp -d)
+  trap 'rm -rf "$TMP"' EXIT
+
+  [ -f "$DCGM_DEFAULT_CSV" ] || { echo "missing $DCGM_DEFAULT_CSV" >&2; exit 1; }
+  [ -f "$DCGM_CUSTOM_CSV" ]  || { echo "missing $DCGM_CUSTOM_CSV"  >&2; exit 1; }
+
+  {
+    extract_csv_metrics "$DCGM_DEFAULT_CSV"
+    extract_csv_metrics "$DCGM_CUSTOM_CSV"
+  } | sort -u > "$TMP/declared.txt"
+  [ -s "$TMP/declared.txt" ] || { echo "no DCGM metrics extracted from CSVs" >&2; exit 1; }
+
+  list_tracked_gpu_dashboards > "$TMP/dashboards.txt"
+  [ -s "$TMP/dashboards.txt" ] || { echo "no gpu/* dashboards listed in $DASHBOARDS_LIST" >&2; exit 1; }
+
+  failed=0
+
+  # Dashboard coverage — every dashboard's DCGM references must resolve.
+  while IFS= read -r dashboard_rel; do
+    dashboard="$DASHBOARDS_DIR/$dashboard_rel"
+    [ -f "$dashboard" ] || continue  # handled by the existence test
+    extract_dcgm_refs "$dashboard" > "$TMP/refs.txt"
+    [ -s "$TMP/refs.txt" ] || continue  # dashboard relies entirely on recording rules
+    comm -23 "$TMP/refs.txt" "$TMP/declared.txt" > "$TMP/missing.txt"
+    if [ -s "$TMP/missing.txt" ]; then
+      echo "ERROR: $dashboard_rel references DCGM metrics not declared in any CSV:" >&2
+      sed 's/^/  - /' "$TMP/missing.txt" >&2
+      failed=1
+    fi
+  done < "$TMP/dashboards.txt"
+
+  # Rules coverage — recording rules consume DCGM directly, so their set
+  # must be declared too, otherwise derived series on every dashboard
+  # collapse to empty.
+  extract_dcgm_refs "$RULES_FILE" > "$TMP/rule-refs.txt"
+  if [ -s "$TMP/rule-refs.txt" ]; then
+    comm -23 "$TMP/rule-refs.txt" "$TMP/declared.txt" > "$TMP/rule-missing.txt"
+    if [ -s "$TMP/rule-missing.txt" ]; then
+      echo "ERROR: gpu-recording.rules.yaml references DCGM metrics not declared in any CSV:" >&2
+      sed 's/^/  - /' "$TMP/rule-missing.txt" >&2
+      failed=1
+    fi
+  fi
+
+  [ "$failed" -eq 0 ]
+}
+
+@test "every tracked GPU dashboard listed in dashboards-infra.list exists on disk" {
+  TMP=$(mktemp -d)
+  trap 'rm -rf "$TMP"' EXIT
+
+  list_tracked_gpu_dashboards > "$TMP/dashboards.txt"
+  [ -s "$TMP/dashboards.txt" ] || { echo "no gpu/* dashboards listed in $DASHBOARDS_LIST" >&2; exit 1; }
+
+  failed=0
+  while IFS= read -r dashboard_rel; do
+    dashboard="$DASHBOARDS_DIR/$dashboard_rel"
+    if [ ! -f "$dashboard" ]; then
+      echo "ERROR: $dashboard_rel listed in dashboards-infra.list but $dashboard does not exist" >&2
+      failed=1
+    fi
+  done < "$TMP/dashboards.txt"
+
+  [ "$failed" -eq 0 ]
+}
diff --git a/hack/check-host-runtime.bats b/hack/check-host-runtime.bats
new file mode 100644
index 00000000..b121b31b
--- /dev/null
+++ b/hack/check-host-runtime.bats
@@ -0,0 +1,415 @@
+#!/usr/bin/env bats
+# -----------------------------------------------------------------------------
+# Unit tests for hack/check-host-runtime.sh
+#
+# The script warns when a standalone containerd.service or docker.service is
+# active alongside the embedded k3s runtime on Ubuntu hosts running the
+# cozystack "generic" variant. Warnings go to stderr; exit code is always 0.
+#
+# Test strategy: each test builds its own temporary stub directory and prepends
+# it to PATH to inject a fake `systemctl` (and optionally `du`) binary. The
+# script itself honors a small set of COZYSTACK_PREFLIGHT_* environment
+# variables to redirect socket/dir probes into the stub tree, so tests do not
+# need root privileges or a real systemd host.
+#
+# Each test installs a `trap 'rm -rf "$STUB_DIR"' EXIT` immediately after
+# creating the stub dir so cleanup runs even when an assertion fails mid-test
+# under `set -e`. cozytest.sh runs each @test in its own subshell, so traps
+# scope per test and do not leak across tests.
+#
+# Tests are otherwise self-contained — no shared setup/teardown helpers,
+# because cozytest.sh's awk parser only recognizes @test blocks and treats a
+# bare `}` on its own line as the end of a test function.
+#
+# Title syntax constraints (inherited from cozytest.sh's awk parser):
+#   - Titles must be delimited by ASCII double quotes; embedded literal
+#     double quotes are NOT escaped and will silently truncate the title.
+#   - Only alphanumeric characters from the title survive into the shell
+#     function name (everything else becomes '_'), so titles that differ
+#     only in punctuation collapse to the same function name. Keep titles
+#     distinctive in their alphanumeric run.
+#
+# Run with: hack/cozytest.sh hack/check-host-runtime.bats
+#           (or `bats hack/check-host-runtime.bats` if the bats binary is
+#           installed; cozytest.sh is the CI path.)
+# -----------------------------------------------------------------------------
+
+@test "clean host with no runtime services exits silently" {
+  STUB_DIR=$(mktemp -d)
+  trap 'rm -rf "$STUB_DIR"' EXIT
+
+  cat >"$STUB_DIR/systemctl" <<'STUBEOF'
+#!/bin/sh
+if [ "$1" = "--version" ]; then
+  echo "systemd stub"
+  exit 0
+fi
+exit 1
+STUBEOF
+  chmod +x "$STUB_DIR/systemctl"
+
+  STDERR_FILE="$STUB_DIR/stderr"
+  COZYSTACK_CONTAINERD_SOCKET="$STUB_DIR/missing-containerd.sock" \
+  COZYSTACK_DOCKER_SOCKET_PATHS="$STUB_DIR/missing-docker1.sock $STUB_DIR/missing-docker2.sock" \
+  COZYSTACK_CONTAINERD_DIR="$STUB_DIR/missing-containerd-dir" \
+  COZYSTACK_DOCKER_DIR="$STUB_DIR/missing-docker-dir" \
+  PATH="$STUB_DIR:$PATH" \
+    bash hack/check-host-runtime.sh >"$STUB_DIR/stdout" 2>"$STDERR_FILE"
+
+  [ ! -s "$STDERR_FILE" ]
+  [ ! -s "$STUB_DIR/stdout" ]
+}
+
+@test "standalone containerd service active prints warning" {
+  STUB_DIR=$(mktemp -d)
+  trap 'rm -rf "$STUB_DIR"' EXIT
+
+  cat >"$STUB_DIR/systemctl" <<'STUBEOF'
+#!/bin/sh
+if [ "$1" = "--version" ]; then
+  echo "systemd stub"
+  exit 0
+fi
+if [ "$1" = "is-active" ] && [ "$2" = "containerd.service" ]; then
+  echo active
+  exit 0
+fi
+exit 1
+STUBEOF
+  chmod +x "$STUB_DIR/systemctl"
+
+  mkdir -p "$STUB_DIR/var-lib-containerd"
+  echo dummy >"$STUB_DIR/var-lib-containerd/dummy"
+
+  STDERR_FILE="$STUB_DIR/stderr"
+  COZYSTACK_CONTAINERD_SOCKET="$STUB_DIR/missing-containerd.sock" \
+  COZYSTACK_DOCKER_SOCKET_PATHS="$STUB_DIR/missing-docker.sock" \
+  COZYSTACK_CONTAINERD_DIR="$STUB_DIR/var-lib-containerd" \
+  COZYSTACK_DOCKER_DIR="$STUB_DIR/missing-docker-dir" \
+  PATH="$STUB_DIR:$PATH" \
+    bash hack/check-host-runtime.sh 2>"$STDERR_FILE"
+
+  grep -q 'standalone containerd.service' "$STDERR_FILE"
+  if grep -q 'standalone docker.service' "$STDERR_FILE"; then
+    echo "unexpected docker warning found:" >&2
+    cat "$STDERR_FILE" >&2
+    exit 1
+  fi
+  # HINT line must name only the detected service, not advise disabling
+  # docker.service when only containerd.service is running. The sudo
+  # prefix is also required — without it the command silently no-ops
+  # for a non-root operator, so the prefix is part of the contract.
+  grep -q 'sudo systemctl disable --now containerd.service' "$STDERR_FILE"
+  if grep -q 'systemctl disable --now.*docker' "$STDERR_FILE"; then
+    echo "HINT unexpectedly mentions docker:" >&2
+    cat "$STDERR_FILE" >&2
+    exit 1
+  fi
+}
+
+@test "standalone docker service active prints warning" {
+  STUB_DIR=$(mktemp -d)
+  trap 'rm -rf "$STUB_DIR"' EXIT
+
+  cat >"$STUB_DIR/systemctl" <<'STUBEOF'
+#!/bin/sh
+if [ "$1" = "--version" ]; then
+  echo "systemd stub"
+  exit 0
+fi
+if [ "$1" = "is-active" ] && [ "$2" = "docker.service" ]; then
+  echo active
+  exit 0
+fi
+exit 1
+STUBEOF
+  chmod +x "$STUB_DIR/systemctl"
+
+  mkdir -p "$STUB_DIR/var-lib-docker"
+  echo dummy >"$STUB_DIR/var-lib-docker/dummy"
+
+  STDERR_FILE="$STUB_DIR/stderr"
+  COZYSTACK_CONTAINERD_SOCKET="$STUB_DIR/missing-containerd.sock" \
+  COZYSTACK_DOCKER_SOCKET_PATHS="$STUB_DIR/missing-docker.sock" \
+  COZYSTACK_CONTAINERD_DIR="$STUB_DIR/missing-containerd-dir" \
+  COZYSTACK_DOCKER_DIR="$STUB_DIR/var-lib-docker" \
+  PATH="$STUB_DIR:$PATH" \
+    bash hack/check-host-runtime.sh 2>"$STDERR_FILE"
+
+  grep -q 'standalone docker.service' "$STDERR_FILE"
+  if grep -q 'standalone containerd.service' "$STDERR_FILE"; then
+    echo "unexpected containerd warning found:" >&2
+    cat "$STDERR_FILE" >&2
+    exit 1
+  fi
+  # HINT line must name only the detected service. As in the
+  # containerd test, the sudo prefix is part of the contract.
+  grep -q 'sudo systemctl disable --now docker.service' "$STDERR_FILE"
+  if grep -q 'systemctl disable --now.*containerd' "$STDERR_FILE"; then
+    echo "HINT unexpectedly mentions containerd:" >&2
+    cat "$STDERR_FILE" >&2
+    exit 1
+  fi
+}
+
+@test "both services active prints two warnings and the HINT block" {
+  STUB_DIR=$(mktemp -d)
+  trap 'rm -rf "$STUB_DIR"' EXIT
+
+  cat >"$STUB_DIR/systemctl" <<'STUBEOF'
+#!/bin/sh
+if [ "$1" = "--version" ]; then
+  echo "systemd stub"
+  exit 0
+fi
+if [ "$1" = "is-active" ]; then
+  case "$2" in
+    containerd.service|docker.service) echo active; exit 0 ;;
+  esac
+fi
+exit 1
+STUBEOF
+  chmod +x "$STUB_DIR/systemctl"
+
+  mkdir -p "$STUB_DIR/var-lib-containerd" "$STUB_DIR/var-lib-docker"
+
+  STDERR_FILE="$STUB_DIR/stderr"
+  # Capture exit code explicitly: the script contract says exit 0
+  # unconditionally (warning, not blocker). `set -e` in the test
+  # function body would already fail on a nonzero exit, but an
+  # explicit status check locks in the contract and makes a
+  # regression show up as "expected 0, got N" rather than as a
+  # generic test failure.
+  status=0
+  COZYSTACK_CONTAINERD_SOCKET="$STUB_DIR/missing-containerd.sock" \
+  COZYSTACK_DOCKER_SOCKET_PATHS="$STUB_DIR/missing-docker.sock" \
+  COZYSTACK_CONTAINERD_DIR="$STUB_DIR/var-lib-containerd" \
+  COZYSTACK_DOCKER_DIR="$STUB_DIR/var-lib-docker" \
+  PATH="$STUB_DIR:$PATH" \
+    bash hack/check-host-runtime.sh 2>"$STDERR_FILE" || status=$?
+
+  [ "$status" -eq 0 ]
+  grep -q 'standalone containerd.service' "$STDERR_FILE"
+  grep -q 'standalone docker.service' "$STDERR_FILE"
+  # HINT block must fire whenever warnings exist; otherwise a future silent
+  # removal of the HINT would go unnoticed. When both services fire the HINT
+  # must list both in a single sudo systemctl disable invocation — the sudo
+  # prefix is as important as the systemctl verb, otherwise the operator
+  # would be told to run it as a non-root user and quietly fail.
+  grep -q 'HINT:' "$STDERR_FILE"
+  grep -q 'sudo systemctl disable --now containerd.service docker.service' "$STDERR_FILE"
+}
+
+@test "failing du does not suppress the containerd warning" {
+  STUB_DIR=$(mktemp -d)
+  trap 'rm -rf "$STUB_DIR"' EXIT
+
+  cat >"$STUB_DIR/systemctl" <<'STUBEOF'
+#!/bin/sh
+if [ "$1" = "--version" ]; then
+  echo "systemd stub"
+  exit 0
+fi
+if [ "$1" = "is-active" ] && [ "$2" = "containerd.service" ]; then
+  echo active
+  exit 0
+fi
+exit 1
+STUBEOF
+  chmod +x "$STUB_DIR/systemctl"
+  cat >"$STUB_DIR/du" <<'DUEOF'
+#!/bin/sh
+exit 1
+DUEOF
+  chmod +x "$STUB_DIR/du"
+
+  mkdir -p "$STUB_DIR/var-lib-containerd"
+
+  STDERR_FILE="$STUB_DIR/stderr"
+  COZYSTACK_CONTAINERD_SOCKET="$STUB_DIR/missing-containerd.sock" \
+  COZYSTACK_DOCKER_SOCKET_PATHS="$STUB_DIR/missing-docker.sock" \
+  COZYSTACK_CONTAINERD_DIR="$STUB_DIR/var-lib-containerd" \
+  COZYSTACK_DOCKER_DIR="$STUB_DIR/missing-docker-dir" \
+  PATH="$STUB_DIR:$PATH" \
+    bash hack/check-host-runtime.sh 2>"$STDERR_FILE"
+
+  grep -q 'standalone containerd.service' "$STDERR_FILE"
+}
+
+@test "containerd socket fallback fires when systemctl is unavailable" {
+  STUB_DIR=$(mktemp -d)
+  trap 'rm -rf "$STUB_DIR"' EXIT
+
+  # The script uses `[ -e "$sock" ]`, not `[ -S ... ]`, so a regular
+  # file is a valid stand-in for a unix socket in tests. This also
+  # removes any optional runtime dependency on python3 and makes the
+  # test unconditional on every CI runner.
+  SOCK="$STUB_DIR/containerd.sock"
+  touch "$SOCK"
+
+  STDERR_FILE="$STUB_DIR/stderr"
+  COZYSTACK_PREFLIGHT_FORCE_NO_SYSTEMCTL=1 \
+  COZYSTACK_CONTAINERD_SOCKET="$SOCK" \
+  COZYSTACK_DOCKER_SOCKET_PATHS="$STUB_DIR/missing-docker.sock" \
+  COZYSTACK_CONTAINERD_DIR="$STUB_DIR/missing-containerd-dir" \
+  COZYSTACK_DOCKER_DIR="$STUB_DIR/missing-docker-dir" \
+    bash hack/check-host-runtime.sh 2>"$STDERR_FILE"
+
+  grep -q 'standalone containerd.service' "$STDERR_FILE"
+}
+
+@test "docker socket fallback fires when systemctl is unavailable" {
+  STUB_DIR=$(mktemp -d)
+  trap 'rm -rf "$STUB_DIR"' EXIT
+
+  SOCK="$STUB_DIR/docker.sock"
+  touch "$SOCK"
+
+  STDERR_FILE="$STUB_DIR/stderr"
+  COZYSTACK_PREFLIGHT_FORCE_NO_SYSTEMCTL=1 \
+  COZYSTACK_CONTAINERD_SOCKET="$STUB_DIR/missing-containerd.sock" \
+  COZYSTACK_DOCKER_SOCKET_PATHS="$SOCK" \
+  COZYSTACK_CONTAINERD_DIR="$STUB_DIR/missing-containerd-dir" \
+  COZYSTACK_DOCKER_DIR="$STUB_DIR/missing-docker-dir" \
+    bash hack/check-host-runtime.sh 2>"$STDERR_FILE"
+
+  grep -q 'standalone docker.service' "$STDERR_FILE"
+  if grep -q 'standalone containerd.service' "$STDERR_FILE"; then
+    echo "unexpected containerd warning found:" >&2
+    cat "$STDERR_FILE" >&2
+    exit 1
+  fi
+}
+
+@test "clean host without systemctl exits silently" {
+  STUB_DIR=$(mktemp -d)
+  trap 'rm -rf "$STUB_DIR"' EXIT
+
+  STDERR_FILE="$STUB_DIR/stderr"
+  COZYSTACK_PREFLIGHT_FORCE_NO_SYSTEMCTL=1 \
+  COZYSTACK_CONTAINERD_SOCKET="$STUB_DIR/missing-containerd.sock" \
+  COZYSTACK_DOCKER_SOCKET_PATHS="$STUB_DIR/missing-docker1.sock $STUB_DIR/missing-docker2.sock" \
+  COZYSTACK_CONTAINERD_DIR="$STUB_DIR/missing-containerd-dir" \
+  COZYSTACK_DOCKER_DIR="$STUB_DIR/missing-docker-dir" \
+    bash hack/check-host-runtime.sh >"$STUB_DIR/stdout" 2>"$STDERR_FILE"
+
+  [ ! -s "$STDERR_FILE" ]
+  [ ! -s "$STUB_DIR/stdout" ]
+}
+
+@test "docker service plus socket still emits exactly one warning" {
+  STUB_DIR=$(mktemp -d)
+  trap 'rm -rf "$STUB_DIR"' EXIT
+
+  cat >"$STUB_DIR/systemctl" <<'STUBEOF'
+#!/bin/sh
+if [ "$1" = "--version" ]; then
+  echo "systemd stub"
+  exit 0
+fi
+if [ "$1" = "is-active" ] && [ "$2" = "docker.service" ]; then
+  echo active
+  exit 0
+fi
+exit 1
+STUBEOF
+  chmod +x "$STUB_DIR/systemctl"
+
+  SOCK="$STUB_DIR/docker.sock"
+  touch "$SOCK"
+
+  mkdir -p "$STUB_DIR/var-lib-docker"
+
+  STDERR_FILE="$STUB_DIR/stderr"
+  COZYSTACK_CONTAINERD_SOCKET="$STUB_DIR/missing-containerd.sock" \
+  COZYSTACK_DOCKER_SOCKET_PATHS="$SOCK" \
+  COZYSTACK_CONTAINERD_DIR="$STUB_DIR/missing-containerd-dir" \
+  COZYSTACK_DOCKER_DIR="$STUB_DIR/var-lib-docker" \
+  PATH="$STUB_DIR:$PATH" \
+    bash hack/check-host-runtime.sh 2>"$STDERR_FILE"
+
+  count=$(grep -c 'standalone docker.service' "$STDERR_FILE")
+  if [ "$count" != "1" ]; then
+    echo "expected exactly one docker warning, got $count" >&2
+    cat "$STDERR_FILE" >&2
+    exit 1
+  fi
+}
+
+@test "docker socket paths with glob chars do not expand" {
+  STUB_DIR=$(mktemp -d)
+  trap 'rm -rf "$STUB_DIR"' EXIT
+
+  # Create two directories that a naive `for sock in $PATHS` loop
+  # would glob-expand and treat as existing "sockets". With the
+  # array-based parsing the literal path "$STUB_DIR/var-lib-*" does
+  # not exist and no warning must fire.
+  mkdir -p "$STUB_DIR/var-lib-docker" "$STUB_DIR/var-lib-containerd"
+
+  cat >"$STUB_DIR/systemctl" <<'STUBEOF'
+#!/bin/sh
+if [ "$1" = "--version" ]; then
+  echo "systemd stub"
+  exit 0
+fi
+exit 1
+STUBEOF
+  chmod +x "$STUB_DIR/systemctl"
+
+  STDERR_FILE="$STUB_DIR/stderr"
+  COZYSTACK_CONTAINERD_SOCKET="$STUB_DIR/missing-containerd.sock" \
+  COZYSTACK_DOCKER_SOCKET_PATHS="$STUB_DIR/var-lib-*" \
+  COZYSTACK_CONTAINERD_DIR="$STUB_DIR/missing-containerd-dir" \
+  COZYSTACK_DOCKER_DIR="$STUB_DIR/missing-docker-dir" \
+  PATH="$STUB_DIR:$PATH" \
+    bash hack/check-host-runtime.sh 2>"$STDERR_FILE"
+
+  if grep -q 'standalone docker.service' "$STDERR_FILE"; then
+    echo "glob pattern expanded — docker warning should not fire:" >&2
+    cat "$STDERR_FILE" >&2
+    exit 1
+  fi
+}
+
+@test "containerd service plus socket still emits exactly one warning" {
+  STUB_DIR=$(mktemp -d)
+  trap 'rm -rf "$STUB_DIR"' EXIT
+
+  cat >"$STUB_DIR/systemctl" <<'STUBEOF'
+#!/bin/sh
+if [ "$1" = "--version" ]; then
+  echo "systemd stub"
+  exit 0
+fi
+if [ "$1" = "is-active" ] && [ "$2" = "containerd.service" ]; then
+  echo active
+  exit 0
+fi
+exit 1
+STUBEOF
+  chmod +x "$STUB_DIR/systemctl"
+
+  # The script uses `[ -e "$sock" ]`, not `[ -S ... ]`, so a regular
+  # file is a valid stand-in for a unix socket in tests. This also
+  # removes any optional runtime dependency on python3 and makes the
+  # test unconditional on every CI runner.
+  SOCK="$STUB_DIR/containerd.sock"
+  touch "$SOCK"
+
+  mkdir -p "$STUB_DIR/var-lib-containerd"
+
+  STDERR_FILE="$STUB_DIR/stderr"
+  COZYSTACK_CONTAINERD_SOCKET="$SOCK" \
+  COZYSTACK_DOCKER_SOCKET_PATHS="$STUB_DIR/missing-docker.sock" \
+  COZYSTACK_CONTAINERD_DIR="$STUB_DIR/var-lib-containerd" \
+  COZYSTACK_DOCKER_DIR="$STUB_DIR/missing-docker-dir" \
+  PATH="$STUB_DIR:$PATH" \
+    bash hack/check-host-runtime.sh 2>"$STDERR_FILE"
+
+  count=$(grep -c 'standalone containerd.service' "$STDERR_FILE")
+  if [ "$count" != "1" ]; then
+    echo "expected exactly one containerd warning, got $count" >&2
+    cat "$STDERR_FILE" >&2
+    exit 1
+  fi
+}
diff --git a/hack/check-host-runtime.sh b/hack/check-host-runtime.sh
new file mode 100755
index 00000000..917453d2
--- /dev/null
+++ b/hack/check-host-runtime.sh
@@ -0,0 +1,161 @@
+#!/usr/bin/env bash
+# -----------------------------------------------------------------------------
+# check-host-runtime.sh — operator preflight warning
+#
+# Purpose:
+#   Warn when a standalone containerd.service or docker.service is running on
+#   the host alongside the embedded k3s runtime. This mismatch is silent on
+#   day 0 (k3s uses its own containerd at /run/k3s/containerd/containerd.sock
+#   and /var/lib/rancher/k3s/agent/containerd) but over time the standalone
+#   runtime accumulates unpruned images and build cache in /var/lib/containerd
+#   — enough to trigger DiskPressure and crash cozystack-api with eviction
+#   loops. The script does NOT block install; it only prints a warning.
+#
+# When to run:
+#   Before `helm install cozy-installer` on an Ubuntu host prepared with k3s
+#   or kubeadm (cozystack "generic" variant). Irrelevant on Talos where the
+#   container runtime lifecycle is fully managed. Discoverable via
+#   `make preflight` from the repository root.
+#
+# Exit code:
+#   Always 0 (warning, not a blocker). Warnings go to stderr.
+#
+# Environment variables (test hooks — override default probe paths):
+#   COZYSTACK_CONTAINERD_SOCKET        standalone containerd socket path
+#   COZYSTACK_DOCKER_SOCKET_PATHS      space-separated list of docker socket paths
+#   COZYSTACK_CONTAINERD_DIR           standalone containerd data directory
+#   COZYSTACK_DOCKER_DIR               standalone docker data directory
+#   COZYSTACK_PREFLIGHT_FORCE_NO_SYSTEMCTL=1    pretend systemctl is absent
+# -----------------------------------------------------------------------------
+set -euo pipefail
+
+if [ -t 2 ]; then
+    YELLOW=$'\033[1;33m'
+    RESET=$'\033[0m'
+else
+    YELLOW=''
+    RESET=''
+fi
+
+CONTAINERD_SOCKET=${COZYSTACK_CONTAINERD_SOCKET:-/run/containerd/containerd.sock}
+DOCKER_SOCKET_PATHS=${COZYSTACK_DOCKER_SOCKET_PATHS:-/run/docker.sock /var/run/docker.sock}
+CONTAINERD_DIR=${COZYSTACK_CONTAINERD_DIR:-/var/lib/containerd}
+DOCKER_DIR=${COZYSTACK_DOCKER_DIR:-/var/lib/docker}
+
+CONTAINERD_WARN=0
+DOCKER_WARN=0
+
+warn() {
+    printf '%sWARNING:%s %s\n' "$YELLOW" "$RESET" "$1" >&2
+}
+
+detect_systemctl() {
+    if [ "${COZYSTACK_PREFLIGHT_FORCE_NO_SYSTEMCTL:-0}" = "1" ]; then
+        return 1
+    fi
+    if command -v systemctl >/dev/null 2>&1 && systemctl --version >/dev/null 2>&1; then
+        return 0
+    fi
+    return 1
+}
+
+disk_usage() {
+    local path=$1
+    local usage
+    if [ -d "$path" ]; then
+        # Wrap `du` in `timeout 5s` so a container data directory with
+        # millions of files (the exact scenario this script exists to
+        # warn about) cannot stall the preflight indefinitely. If the
+        # `timeout` binary is absent the pipeline still exits 0 via
+        # `|| true` and `usage` stays empty; the warning itself is
+        # still printed, just without the size detail.
+        usage=$(timeout 5s du -sh "$path" 2>/dev/null | awk '{print $1}' || true)
+        if [ -n "${usage:-}" ]; then
+            printf ' (%s uses %s)' "$path" "$usage"
+        fi
+    fi
+}
+
+service_active() {
+    local service=$1
+    if [ "$HAS_SYSTEMCTL" = "1" ]; then
+        if systemctl is-active "$service" >/dev/null 2>&1; then
+            return 0
+        fi
+    fi
+    return 1
+}
+
+check_containerd() {
+    local detail=""
+    local found=0
+    if service_active containerd.service; then
+        found=1
+    fi
+    if [ "$found" -eq 0 ] && [ -e "$CONTAINERD_SOCKET" ]; then
+        found=1
+    fi
+    if [ "$found" -eq 1 ]; then
+        detail=$(disk_usage "$CONTAINERD_DIR")
+        warn "standalone containerd.service detected alongside k3s embedded runtime${detail}"
+        CONTAINERD_WARN=1
+    fi
+}
+
+check_docker() {
+    local detail=""
+    local found=0
+    if service_active docker.service; then
+        found=1
+    fi
+    if [ "$found" -eq 0 ]; then
+        # DOCKER_SOCKET_PATHS is a space separated list of paths. Parse
+        # it into an array via `read -ra` so that word splitting is
+        # explicit AND glob expansion is suppressed — `for sock in
+        # $DOCKER_SOCKET_PATHS` would both word split and glob, so a
+        # path containing a literal `*` or `?` could expand into
+        # directory entries and produce false positives.
+        local -a _socks
+        read -ra _socks <<<"$DOCKER_SOCKET_PATHS"
+        for sock in "${_socks[@]}"; do
+            if [ -e "$sock" ]; then
+                found=1
+                break
+            fi
+        done
+    fi
+    if [ "$found" -eq 1 ]; then
+        detail=$(disk_usage "$DOCKER_DIR")
+        warn "standalone docker.service detected alongside k3s embedded runtime${detail}"
+        DOCKER_WARN=1
+    fi
+}
+
+if detect_systemctl; then
+    HAS_SYSTEMCTL=1
+else
+    HAS_SYSTEMCTL=0
+fi
+
+check_containerd
+check_docker
+
+if [ "$CONTAINERD_WARN" -eq 1 ] || [ "$DOCKER_WARN" -eq 1 ]; then
+    services=""
+    if [ "$CONTAINERD_WARN" -eq 1 ]; then
+        services="containerd.service"
+    fi
+    if [ "$DOCKER_WARN" -eq 1 ]; then
+        if [ -n "$services" ]; then
+            services="$services docker.service"
+        else
+            services="docker.service"
+        fi
+    fi
+    printf '%sHINT:%s cozystack runs its own containerd under k3s. To stop the shadow runtime:\n' "$YELLOW" "$RESET" >&2
+    printf '  sudo systemctl disable --now %s\n' "$services" >&2
+    printf 'Inspect and reclaim standalone runtime storage separately — it may contain container data\n' >&2
+    printf 'that the operator still needs; do not delete it blindly.\n' >&2
+fi
+
+exit 0
diff --git a/hack/common-envs.mk b/hack/common-envs.mk
index 56eb5478..9b8db5db 100644
--- a/hack/common-envs.mk
+++ b/hack/common-envs.mk
@@ -6,13 +6,13 @@ else
 endif
 
 REGISTRY ?= ghcr.io/cozystack/cozystack
-TAG = $(shell git describe --tags --exact-match 2>/dev/null || echo latest)
+TAG = $(shell git describe --tags --exact-match --match 'v*' 2>/dev/null || echo latest)
 PUSH := 1
 LOAD := 0
 BUILDER ?=
 PLATFORM ?= 
 BUILDX_EXTRA_ARGS ?=
-COZYSTACK_VERSION = $(patsubst v%,%,$(shell git describe --tags))
+COZYSTACK_VERSION = $(patsubst v%,%,$(shell git describe --tags --match 'v*'))
 
 BUILDX_ARGS := --provenance=false --push=$(PUSH) --load=$(LOAD) \
   --label org.opencontainers.image.source=https://github.com/cozystack/cozystack \
@@ -28,6 +28,6 @@ endef
 ifeq ($(COZYSTACK_VERSION),)
     $(shell git remote add upstream https://github.com/cozystack/cozystack.git || true)
     $(shell git fetch upstream --tags)
-    COZYSTACK_VERSION = $(patsubst v%,%,$(shell git describe --tags))
+    COZYSTACK_VERSION = $(patsubst v%,%,$(shell git describe --tags --match 'v*'))
 endif
 
diff --git a/hack/dcgm-default-counters.csv b/hack/dcgm-default-counters.csv
new file mode 100644
index 00000000..b5e94540
--- /dev/null
+++ b/hack/dcgm-default-counters.csv
@@ -0,0 +1,104 @@
+# Snapshot of the upstream DCGM Exporter default-counters.csv used as a
+# fixture by hack/check-gpu-recording-rules.bats. The test verifies that
+# every DCGM_FI_* metric referenced by a tracked GPU dashboard is either
+# declared here (upstream defaults) or in
+# packages/system/gpu-operator/examples/dcgm-custom-metrics.yaml
+# (the project's custom CSV).
+#
+# Source: https://github.com/NVIDIA/dcgm-exporter/blob/4.1.1-4.0.4/etc/default-counters.csv
+# Pinned to the DCGM Exporter image tag shipped by the gpu-operator
+# chart under packages/system/gpu-operator/charts/gpu-operator/values.yaml
+# (dcgmExporter.version = 4.1.1-4.0.4-ubuntu22.04). When that image is
+# bumped, refresh this file from the matching tag in the NVIDIA/dcgm-exporter
+# repo and re-run `./hack/cozytest.sh hack/check-gpu-recording-rules.bats`.
+
+# Format
+# If line starts with a '#' it is considered a comment
+# DCGM FIELD, Prometheus metric type, help message
+
+# Clocks
+DCGM_FI_DEV_SM_CLOCK,  gauge, SM clock frequency (in MHz).
+DCGM_FI_DEV_MEM_CLOCK, gauge, Memory clock frequency (in MHz).
+
+# Temperature
+DCGM_FI_DEV_MEMORY_TEMP, gauge, Memory temperature (in C).
+DCGM_FI_DEV_GPU_TEMP,    gauge, GPU temperature (in C).
+
+# Power
+DCGM_FI_DEV_POWER_USAGE,              gauge, Power draw (in W).
+DCGM_FI_DEV_TOTAL_ENERGY_CONSUMPTION, counter, Total energy consumption since boot (in mJ).
+
+# PCIE
+# DCGM_FI_PROF_PCIE_TX_BYTES,  counter, Total number of bytes transmitted through PCIe TX via NVML.
+# DCGM_FI_PROF_PCIE_RX_BYTES,  counter, Total number of bytes received through PCIe RX via NVML.
+DCGM_FI_DEV_PCIE_REPLAY_COUNTER, counter, Total number of PCIe retries.
+
+# Utilization (the sample period varies depending on the product)
+DCGM_FI_DEV_GPU_UTIL,      gauge, GPU utilization (in %).
+DCGM_FI_DEV_MEM_COPY_UTIL, gauge, Memory utilization (in %).
+DCGM_FI_DEV_ENC_UTIL,      gauge, Encoder utilization (in %).
+DCGM_FI_DEV_DEC_UTIL ,     gauge, Decoder utilization (in %).
+
+# Errors and violations
+DCGM_FI_DEV_XID_ERRORS,              gauge,   Value of the last XID error encountered.
+# DCGM_FI_DEV_POWER_VIOLATION,       counter, Throttling duration due to power constraints (in us).
+# DCGM_FI_DEV_THERMAL_VIOLATION,     counter, Throttling duration due to thermal constraints (in us).
+# DCGM_FI_DEV_SYNC_BOOST_VIOLATION,  counter, Throttling duration due to sync-boost constraints (in us).
+# DCGM_FI_DEV_BOARD_LIMIT_VIOLATION, counter, Throttling duration due to board limit constraints (in us).
+# DCGM_FI_DEV_LOW_UTIL_VIOLATION,    counter, Throttling duration due to low utilization (in us).
+# DCGM_FI_DEV_RELIABILITY_VIOLATION, counter, Throttling duration due to reliability constraints (in us).
+
+# Memory usage
+DCGM_FI_DEV_FB_FREE, gauge, Framebuffer memory free (in MiB).
+DCGM_FI_DEV_FB_USED, gauge, Framebuffer memory used (in MiB).
+
+# ECC
+# DCGM_FI_DEV_ECC_SBE_VOL_TOTAL, counter, Total number of single-bit volatile ECC errors.
+# DCGM_FI_DEV_ECC_DBE_VOL_TOTAL, counter, Total number of double-bit volatile ECC errors.
+# DCGM_FI_DEV_ECC_SBE_AGG_TOTAL, counter, Total number of single-bit persistent ECC errors.
+# DCGM_FI_DEV_ECC_DBE_AGG_TOTAL, counter, Total number of double-bit persistent ECC errors.
+
+# Retired pages
+# DCGM_FI_DEV_RETIRED_SBE,     counter, Total number of retired pages due to single-bit errors.
+# DCGM_FI_DEV_RETIRED_DBE,     counter, Total number of retired pages due to double-bit errors.
+# DCGM_FI_DEV_RETIRED_PENDING, counter, Total number of pages pending retirement.
+
+# NVLink
+# DCGM_FI_DEV_NVLINK_CRC_FLIT_ERROR_COUNT_TOTAL, counter, Total number of NVLink flow-control CRC errors.
+# DCGM_FI_DEV_NVLINK_CRC_DATA_ERROR_COUNT_TOTAL, counter, Total number of NVLink data CRC errors.
+# DCGM_FI_DEV_NVLINK_REPLAY_ERROR_COUNT_TOTAL,   counter, Total number of NVLink retries.
+# DCGM_FI_DEV_NVLINK_RECOVERY_ERROR_COUNT_TOTAL, counter, Total number of NVLink recovery errors.
+DCGM_FI_DEV_NVLINK_BANDWIDTH_TOTAL,            counter, Total number of NVLink bandwidth counters for all lanes.
+# DCGM_FI_DEV_NVLINK_BANDWIDTH_L0,               counter, The number of bytes of active NVLink rx or tx data including both header and payload.
+
+# VGPU License status
+DCGM_FI_DEV_VGPU_LICENSE_STATUS, gauge, vGPU License status
+
+# Remapped rows
+DCGM_FI_DEV_UNCORRECTABLE_REMAPPED_ROWS, counter, Number of remapped rows for uncorrectable errors
+DCGM_FI_DEV_CORRECTABLE_REMAPPED_ROWS,   counter, Number of remapped rows for correctable errors
+DCGM_FI_DEV_ROW_REMAP_FAILURE,           gauge,   Whether remapping of rows has failed
+
+# Static configuration information. These appear as labels on the other metrics
+DCGM_FI_DRIVER_VERSION,        label, Driver Version
+# DCGM_FI_NVML_VERSION,          label, NVML Version
+# DCGM_FI_DEV_BRAND,             label, Device Brand
+# DCGM_FI_DEV_SERIAL,            label, Device Serial Number
+# DCGM_FI_DEV_OEM_INFOROM_VER,   label, OEM inforom version
+# DCGM_FI_DEV_ECC_INFOROM_VER,   label, ECC inforom version
+# DCGM_FI_DEV_POWER_INFOROM_VER, label, Power management object inforom version
+# DCGM_FI_DEV_INFOROM_IMAGE_VER, label, Inforom image version
+# DCGM_FI_DEV_VBIOS_VERSION,     label, VBIOS version of the device
+
+# Datacenter Profiling (DCP) metrics
+# NOTE: supported on Nvidia datacenter Volta GPUs and newer
+DCGM_FI_PROF_GR_ENGINE_ACTIVE,   gauge, Ratio of time the graphics engine is active.
+# DCGM_FI_PROF_SM_ACTIVE,          gauge, The ratio of cycles an SM has at least 1 warp assigned.
+# DCGM_FI_PROF_SM_OCCUPANCY,       gauge, The ratio of number of warps resident on an SM.
+DCGM_FI_PROF_PIPE_TENSOR_ACTIVE, gauge, Ratio of cycles the tensor (HMMA) pipe is active.
+DCGM_FI_PROF_DRAM_ACTIVE,        gauge, Ratio of cycles the device memory interface is active sending or receiving data.
+# DCGM_FI_PROF_PIPE_FP64_ACTIVE,   gauge, Ratio of cycles the fp64 pipes are active.
+# DCGM_FI_PROF_PIPE_FP32_ACTIVE,   gauge, Ratio of cycles the fp32 pipes are active.
+# DCGM_FI_PROF_PIPE_FP16_ACTIVE,   gauge, Ratio of cycles the fp16 pipes are active.
+DCGM_FI_PROF_PCIE_TX_BYTES,      gauge, The rate of data transmitted over the PCIe bus - including both protocol headers and data payloads - in bytes per second.
+DCGM_FI_PROF_PCIE_RX_BYTES,      gauge, The rate of data received over the PCIe bus - including both protocol headers and data payloads - in bytes per second.
diff --git a/hack/e2e-apps/remediation-guard.sh b/hack/e2e-apps/remediation-guard.sh
new file mode 100644
index 00000000..b45561b9
--- /dev/null
+++ b/hack/e2e-apps/remediation-guard.sh
@@ -0,0 +1,39 @@
+# Helpers for asserting that a Flux HelmRelease did not fall into an
+# install/upgrade remediation cycle during an e2e run.
+#
+# Background: Flux helm-controller's ClearFailures() zeroes
+# .status.installFailures / .status.upgradeFailures on every successful
+# reconciliation (see the upstream ClearFailures method on
+# HelmReleaseStatus). That makes those counters useless for a guard that
+# runs after the HelmRelease has reached Ready - the values are always 0.
+#
+# What survives a successful reconciliation is .status.history, a bounded
+# list of release Snapshots. Each Snapshot carries a status field that
+# tracks the Helm release state: deployed, superseded, failed, uninstalled,
+# and so on. A remediation cycle leaves the footprint behind: a snapshot
+# with status "uninstalled" (from install/upgrade remediation) or "failed"
+# (Helm release failure that remediation then uninstalled). Those stay in
+# history even after a subsequent successful reinstall.
+#
+# helmrelease_has_remediation_cycle takes a newline-delimited list of
+# snapshot statuses (whatever the caller extracted via kubectl -o jsonpath
+# or equivalent) and returns 0 (detected) when any entry is "failed" or
+# "uninstalled", 1 otherwise. Empty input is treated as "no history yet,
+# no cycle observed".
+
+helmrelease_has_remediation_cycle() {
+    statuses="$1"
+    if [ -z "${statuses}" ]; then
+        return 1
+    fi
+    # printf + grep over the pipe, rather than a heredoc plus while read.
+    # printf %s treats the status string as a literal payload, so any stray
+    # $ in a future caller's input does not trigger shell expansion. grep
+    # returns 0 iff at least one line matches the allowlist, which is
+    # exactly the contract the caller wants, so we can return its exit
+    # status directly.
+    if printf '%s\n' "${statuses}" | grep --extended-regexp --quiet '^(failed|uninstalled)$'; then
+        return 0
+    fi
+    return 1
+}
diff --git a/hack/e2e-apps/run-kubernetes.sh b/hack/e2e-apps/run-kubernetes.sh
index be6dcd6f..761ee57d 100644
--- a/hack/e2e-apps/run-kubernetes.sh
+++ b/hack/e2e-apps/run-kubernetes.sh
@@ -1,3 +1,5 @@
+. hack/e2e-apps/remediation-guard.sh
+
 run_kubernetes_test() {
     local version_expr="$1"
     local test_name="$2"
@@ -320,6 +322,35 @@ EOF
     done
     kubectl wait hr kubernetes-${test_name}-ingress-nginx -n tenant-test --timeout=5m --for=condition=ready
 
+  # Guard: parent HelmRelease must not have entered an install/upgrade remediation cycle.
+  # A non-zero installFailures/upgradeFailures indicates the helm-wait budget expired while
+  # admin-kubeconfig was still being provisioned, which would trigger uninstall remediation
+  # and churn the Cluster CR.
+  # Flux helm-controller v2 retains per-revision release Snapshots in
+  # .status.history; each Snapshot's .status reflects the Helm release
+  # state (deployed/superseded/failed/uninstalled). A remediation cycle
+  # leaves a "failed" or "uninstalled" entry behind that survives a later
+  # successful reinstall, unlike the installFailures/upgradeFailures
+  # counters (which ClearFailures zeroes on every successful reconcile).
+  # The shape is pinned by hack/remediation-guard.bats; the upstream
+  # types are github.com/fluxcd/helm-controller/api v2 Snapshot.
+  history_statuses=$(kubectl get hr -n tenant-test "kubernetes-${test_name}" \
+    -ojsonpath='{range .status.history[*]}{.status}{"\n"}{end}')
+  # Always emit the raw value so a silent future-Flux field rename shows
+  # up as "empty history on a Ready HR" in CI logs rather than vanishing.
+  echo "Parent HelmRelease history statuses:"
+  printf '%s\n' "${history_statuses:-<empty>}"
+  if [ -z "${history_statuses}" ]; then
+    echo "Unexpected empty .status.history on a Ready HelmRelease - Flux API shape may have changed." >&2
+    kubectl -n tenant-test describe hr "kubernetes-${test_name}" >&2
+    exit 1
+  fi
+  if helmrelease_has_remediation_cycle "${history_statuses}"; then
+    echo "Parent HelmRelease entered remediation cycle." >&2
+    kubectl -n tenant-test describe hr "kubernetes-${test_name}" >&2
+    exit 1
+  fi
+
   # Clean up
   pkill -f "port-forward.*${port}:" 2>/dev/null || true
   rm -f "tenantkubeconfig-${test_name}"
diff --git a/hack/e2e-install-cozystack.bats b/hack/e2e-install-cozystack.bats
index 743782e9..cf66b73f 100644
--- a/hack/e2e-install-cozystack.bats
+++ b/hack/e2e-install-cozystack.bats
@@ -200,8 +200,60 @@ EOF
   kubectl wait hr/keycloak hr/keycloak-configure hr/keycloak-operator -n cozy-keycloak --timeout=10m --for=condition=ready
 }
 
+@test "Aggregated API rejects Tenant name with dashes" {
+  # Regression guard: the tenant Helm chart's tenant.name helper splits the
+  # Release.Name on "-" and fails unless the result is exactly
+  # ["tenant", "<name>"]. The aggregated API must catch tenant names
+  # containing dashes up-front with a tenant-specific error, instead of
+  # silently accepting the Application and letting Flux fail later.
+
+  # Defensive cleanup: if a prior regression left foo-bar in the cluster,
+  # remove it before exercising the validation so we are not observing
+  # stale state. Safe even in the happy path because of --ignore-not-found.
+  kubectl delete tenants.apps.cozystack.io foo-bar -n tenant-root --ignore-not-found
+
+  # Preflight: tenant-root is created by earlier tests in this suite. Fail
+  # loudly if it is missing so this test does not silently trigger an
+  # unrelated "namespace not found" error and misreport as a pass.
+  kubectl get namespace tenant-root
+
+  # --validate=ignore forces kubectl to skip client-side OpenAPI validation
+  # and send the payload straight to the aggregated API. This guarantees the
+  # server-side name check runs and the error we grep for is the tenant
+  # contract error, not a kubectl schema rejection. (--validate=false is the
+  # deprecated alias.)
+  local output rc
+  # Run the apply in its own subshell so we can capture BOTH stdout+stderr
+  # AND the exit code explicitly, without `|| true` swallowing a real failure
+  # mode (e.g. network error, auth failure) that should also fail the test.
+  output=$(
+    kubectl apply --validate=ignore -f - 2>&1 <<EOF
+apiVersion: apps.cozystack.io/v1alpha1
+kind: Tenant
+metadata:
+  name: foo-bar
+  namespace: tenant-root
+spec: {}
+EOF
+  ) && rc=0 || rc=$?
+  echo "kubectl apply exit=$rc, output=$output"
+  # kubectl MUST have failed: success would mean validation regressed.
+  [ "$rc" -ne 0 ]
+  # Assert the tenant-specific message is present (distinguishes from
+  # generic DNS-1035 errors and from network/auth failures).
+  echo "$output" | grep -q "tenant names must"
+  # And assert kubectl did NOT report creation — if validation regressed
+  # into a "warn" variant, the server could still accept the object.
+  ! echo "$output" | grep -qi "created"
+
+  # Post-condition cleanup: even though we expect validation to reject the
+  # create, removing foo-bar unconditionally keeps the cluster clean for
+  # subsequent tests in case validation regresses and the object is created.
+  kubectl delete tenants.apps.cozystack.io foo-bar -n tenant-root --ignore-not-found
+}
+
 @test "Create tenant with isolated mode enabled" {
-  kubectl -n tenant-root get tenants.apps.cozystack.io test || 
+  kubectl -n tenant-root get tenants.apps.cozystack.io test ||
   kubectl apply -f - <<EOF
 apiVersion: apps.cozystack.io/v1alpha1
 kind: Tenant
diff --git a/hack/remediation-guard.bats b/hack/remediation-guard.bats
new file mode 100644
index 00000000..092fe06d
--- /dev/null
+++ b/hack/remediation-guard.bats
@@ -0,0 +1,127 @@
+#!/usr/bin/env bats
+# -----------------------------------------------------------------------------
+# Unit tests for hack/e2e-apps/remediation-guard.sh
+#
+# helmrelease_has_remediation_cycle takes a newline-delimited list of
+# HelmRelease history snapshot status values (deployed/superseded/failed/
+# uninstalled/...) and returns 0 when any entry is "failed" or "uninstalled"
+# (meaning flux helm-controller performed install/upgrade remediation).
+#
+# This is used by the e2e script after the HelmRelease reaches Ready. The
+# failure/upgrade counters (.status.installFailures / .status.upgradeFailures)
+# are useless there because flux's ClearFailures zeroes them on successful
+# reconciliation; .status.history retains the snapshot trail.
+#
+# cozytest.sh's awk parser recognizes only @test blocks and a bare `}` on
+# its own line; there is no bats `run` or `$status`. Assertions are
+# expressed as direct shell tests that exit non-zero on failure.
+#
+# Run with: hack/cozytest.sh hack/remediation-guard.bats
+# -----------------------------------------------------------------------------
+
+@test "empty history returns not-detected" {
+    . hack/e2e-apps/remediation-guard.sh
+    if helmrelease_has_remediation_cycle ""; then
+        echo "expected not-detected for empty history" >&2
+        exit 1
+    fi
+}
+
+@test "single deployed snapshot returns not-detected" {
+    . hack/e2e-apps/remediation-guard.sh
+    if helmrelease_has_remediation_cycle "deployed"; then
+        echo "expected not-detected for deployed-only history" >&2
+        exit 1
+    fi
+}
+
+@test "deployed then superseded returns not-detected" {
+    . hack/e2e-apps/remediation-guard.sh
+    statuses=$(printf 'deployed\nsuperseded\n')
+    if helmrelease_has_remediation_cycle "${statuses}"; then
+        echo "expected not-detected for deployed+superseded history" >&2
+        exit 1
+    fi
+}
+
+@test "single failed snapshot returns detected" {
+    . hack/e2e-apps/remediation-guard.sh
+    if ! helmrelease_has_remediation_cycle "failed"; then
+        echo "expected detected when history contains failed snapshot" >&2
+        exit 1
+    fi
+}
+
+@test "single uninstalled snapshot returns detected" {
+    # The exact signature of the install-remediation race: the first install
+    # exceeded flux's wait budget, remediation uninstalled, the next retry
+    # eventually succeeded. History still carries the uninstalled snapshot.
+    . hack/e2e-apps/remediation-guard.sh
+    if ! helmrelease_has_remediation_cycle "uninstalled"; then
+        echo "expected detected when history contains uninstalled snapshot" >&2
+        exit 1
+    fi
+}
+
+@test "uninstalled then deployed still returns detected" {
+    . hack/e2e-apps/remediation-guard.sh
+    statuses=$(printf 'uninstalled\ndeployed\n')
+    if ! helmrelease_has_remediation_cycle "${statuses}"; then
+        echo "expected detected despite later successful deploy" >&2
+        exit 1
+    fi
+}
+
+@test "deployed then failed still returns detected" {
+    . hack/e2e-apps/remediation-guard.sh
+    statuses=$(printf 'deployed\nfailed\n')
+    if ! helmrelease_has_remediation_cycle "${statuses}"; then
+        echo "expected detected when any entry is failed" >&2
+        exit 1
+    fi
+}
+
+@test "status.history extraction pins HR v2 status.history shape" {
+    # Pins the Flux HelmRelease v2 .status.history[].status shape that
+    # run-kubernetes.sh relies on. If a future flux release renames the
+    # field, the jsonpath returns nothing, the guard reports no cycle,
+    # and real remediation loops slip past the e2e assertion. This test
+    # uses yq to read the exact path used in the e2e script; the upstream
+    # Snapshot type lives at
+    # github.com/fluxcd/helm-controller/api/v2.Snapshot (via go.mod).
+    tmp=$(mktemp -d)
+    trap 'rm -rf "$tmp"' EXIT
+
+    cat > "$tmp/hr.yaml" <<'YAML'
+apiVersion: helm.toolkit.fluxcd.io/v2
+kind: HelmRelease
+metadata:
+  name: kubernetes-test
+  namespace: tenant-test
+status:
+  history:
+    - name: kubernetes-test
+      namespace: tenant-test
+      version: 1
+      status: uninstalled
+    - name: kubernetes-test
+      namespace: tenant-test
+      version: 2
+      status: deployed
+YAML
+
+    # Default yq output is yaml scalar format, which for string values emits
+    # bare unquoted tokens - matching what kubectl -o jsonpath produces in
+    # e2e. Do not switch to JSON output here; that would quote the values
+    # and break the loop in helmrelease_has_remediation_cycle.
+    statuses=$(yq '.status.history[].status' "$tmp/hr.yaml")
+
+    [ -n "$statuses" ]
+    echo "$statuses" | grep --quiet '^uninstalled$'
+
+    . hack/e2e-apps/remediation-guard.sh
+    if ! helmrelease_has_remediation_cycle "$statuses"; then
+        echo "expected detected for pinned HR snippet with uninstalled + deployed history" >&2
+        exit 1
+    fi
+}
diff --git a/internal/controller/dashboard/customformsoverride.go b/internal/controller/dashboard/customformsoverride.go
index 907ebedd..0a20ae71 100644
--- a/internal/controller/dashboard/customformsoverride.go
+++ b/internal/controller/dashboard/customformsoverride.go
@@ -141,7 +141,10 @@ func (m *Manager) ensureCFOMapping(ctx context.Context, crd *cozyv1alpha1.Applic
 }
 
 // buildMultilineStringSchema parses OpenAPI schema and creates schema with multilineString
-// for all string fields inside spec that don't have enum
+// for all string fields inside spec that don't have enum.
+// It handles two structures:
+// - properties.spec.properties (most resources)
+// - properties.properties (VMDisk and similar resources without spec wrapper)
 func buildMultilineStringSchema(openAPISchema string) (map[string]any, error) {
 	if openAPISchema == "" {
 		return map[string]any{}, nil
@@ -161,15 +164,25 @@ func buildMultilineStringSchema(openAPISchema string) (map[string]any, error) {
 		"properties": map[string]any{},
 	}
 
-	// Check if there's a spec property
-	specProp, ok := props["spec"].(map[string]any)
-	if !ok {
-		return map[string]any{}, nil
+	var specProps map[string]any
+	var hasSpec bool
+
+	// First try to find properties under spec
+	if specProp, ok := props["spec"].(map[string]any); ok {
+		specProps, hasSpec = specProp["properties"].(map[string]any)
 	}
 
-	specProps, ok := specProp["properties"].(map[string]any)
-	if !ok {
-		return map[string]any{}, nil
+	// If no spec wrapper, use top-level properties directly (VMDisk pattern)
+	if !hasSpec {
+		specProps = props
+		// Still wrap in spec for consistency with applyListInputOverrides
+		schemaProps := schema["properties"].(map[string]any)
+		specSchema := map[string]any{
+			"properties": map[string]any{},
+		}
+		schemaProps["spec"] = specSchema
+		processSpecProperties(specProps, specSchema["properties"].(map[string]any))
+		return schema, nil
 	}
 
 	// Create spec.properties structure in schema
@@ -231,10 +244,45 @@ func applyListInputOverrides(schema map[string]any, kind string, openAPIProps ma
 		}
 
 	case "ClickHouse", "Harbor", "HTTPCache", "Kubernetes", "MariaDB", "MongoDB",
-		"NATS", "OpenBAO", "Postgres", "Qdrant", "RabbitMQ", "Redis", "VMDisk":
+		"NATS", "OpenBAO", "Postgres", "Qdrant", "RabbitMQ", "Redis":
 		specProps := ensureSchemaPath(schema, "spec")
 		specProps["storageClass"] = storageClassListInput()
 
+	case "VMDisk":
+		specProps := ensureSchemaPath(schema, "spec")
+		specProps["storageClass"] = storageClassListInput()
+
+		// Override source.image.name to be an API-backed dropdown listing default images
+		if sourceObj, ok := specProps["source"].(map[string]any); ok {
+			if imgProps, ok := sourceObj["properties"].(map[string]any); ok {
+				if imgName, ok := imgProps["image"].(map[string]any); ok {
+					if imgNameProps, ok := imgName["properties"].(map[string]any); ok {
+						imgNameProps["name"] = map[string]any{
+							"type": "listInput",
+							"customProps": map[string]any{
+								"valueUri":    "/api/clusters/{cluster}/k8s/apis/cdi.kubevirt.io/v1beta1/namespaces/cozy-public/datavolumes",
+								"keysToValue": []any{"metadata", "annotations", "vm-default-images.cozystack.io/name"},
+								"keysToLabel": []any{"metadata", "annotations", "vm-default-images.cozystack.io/description"},
+							},
+						}
+					}
+				}
+				// Override source.disk.name to be an API-backed dropdown listing VMDisk resources
+				if diskName, ok := imgProps["disk"].(map[string]any); ok {
+					if diskNameProps, ok := diskName["properties"].(map[string]any); ok {
+						diskNameProps["name"] = map[string]any{
+							"type": "listInput",
+							"customProps": map[string]any{
+								"valueUri":    "/api/clusters/{cluster}/k8s/apis/apps.cozystack.io/v1alpha1/namespaces/{namespace}/vmdisks",
+								"keysToValue": []any{"metadata", "name"},
+								"keysToLabel": []any{"metadata", "name"},
+							},
+						}
+					}
+				}
+			}
+		}
+
 	case "FoundationDB":
 		storageProps := ensureSchemaPath(schema, "spec", "storage")
 		storageProps["storageClass"] = storageClassListInput()
diff --git a/internal/controller/dashboard/customformsoverride_test.go b/internal/controller/dashboard/customformsoverride_test.go
index 7833d2e9..6a83d62e 100644
--- a/internal/controller/dashboard/customformsoverride_test.go
+++ b/internal/controller/dashboard/customformsoverride_test.go
@@ -317,6 +317,108 @@ func TestApplyListInputOverrides_StorageClassFoundationDB(t *testing.T) {
 	assertStorageClassListInput(t, sc)
 }
 
+func TestApplyListInputOverrides_VMDisk_SourceFields(t *testing.T) {
+	openAPISchema := `{
+		"properties":{
+			"optical":{"type":"boolean"},
+			"source":{
+				"type":"object",
+				"properties":{
+					"image":{
+						"type":"object",
+						"properties":{
+							"name":{"type":"string"}
+						}
+					},
+					"disk":{
+						"type":"object",
+						"properties":{
+							"name":{"type":"string"}
+						}
+					}
+				}
+			},
+			"storage":{"type":"string"},
+			"storageClass":{"type":"string"}
+		}
+	}`
+
+	schema, err := buildMultilineStringSchema(openAPISchema)
+	if err != nil {
+		t.Fatalf("buildMultilineStringSchema failed: %v", err)
+	}
+
+	applyListInputOverrides(schema, "VMDisk", map[string]any{})
+
+	specProps := schema["properties"].(map[string]any)["spec"].(map[string]any)["properties"].(map[string]any)
+
+	// Check storageClass
+	sc, ok := specProps["storageClass"].(map[string]any)
+	if !ok {
+		t.Fatal("storageClass not found in spec.properties")
+	}
+	assertStorageClassListInput(t, sc)
+
+	// Check source.image.name listInput
+	// Structure: specProps["source"]["properties"]["image"]["properties"]["name"]
+	sourceObj, ok := specProps["source"].(map[string]any)
+	if !ok {
+		t.Fatal("source not found in spec.properties")
+	}
+	sourceObjProps, ok := sourceObj["properties"].(map[string]any)
+	if !ok {
+		t.Fatal("source.properties not found")
+	}
+	imageObj, ok := sourceObjProps["image"].(map[string]any)
+	if !ok {
+		t.Fatal("image not found in source.properties")
+	}
+	imageObjProps, ok := imageObj["properties"].(map[string]any)
+	if !ok {
+		t.Fatal("image.properties not found")
+	}
+	imgName, ok := imageObjProps["name"].(map[string]any)
+	if !ok {
+		t.Fatal("name not found in image.properties")
+	}
+	if imgName["type"] != "listInput" {
+		t.Errorf("expected type listInput, got %v", imgName["type"])
+	}
+	imgNameCustomProps, ok := imgName["customProps"].(map[string]any)
+	if !ok {
+		t.Fatal("name.customProps not found")
+	}
+	expectedImageURI := "/api/clusters/{cluster}/k8s/apis/cdi.kubevirt.io/v1beta1/namespaces/cozy-public/datavolumes"
+	if imgNameCustomProps["valueUri"] != expectedImageURI {
+		t.Errorf("expected valueUri %s, got %v", expectedImageURI, imgNameCustomProps["valueUri"])
+	}
+
+	// Check source.disk.name listInput
+	diskObj, ok := sourceObjProps["disk"].(map[string]any)
+	if !ok {
+		t.Fatal("disk not found in source.properties")
+	}
+	diskObjProps, ok := diskObj["properties"].(map[string]any)
+	if !ok {
+		t.Fatal("disk.properties not found")
+	}
+	diskName, ok := diskObjProps["name"].(map[string]any)
+	if !ok {
+		t.Fatal("name not found in disk.properties")
+	}
+	if diskName["type"] != "listInput" {
+		t.Errorf("expected type listInput, got %v", diskName["type"])
+	}
+	diskNameCustomProps, ok := diskName["customProps"].(map[string]any)
+	if !ok {
+		t.Fatal("disk name.customProps not found")
+	}
+	expectedDiskURI := "/api/clusters/{cluster}/k8s/apis/apps.cozystack.io/v1alpha1/namespaces/{namespace}/vmdisks"
+	if diskNameCustomProps["valueUri"] != expectedDiskURI {
+		t.Errorf("expected valueUri %s, got %v", expectedDiskURI, diskNameCustomProps["valueUri"])
+	}
+}
+
 func TestApplyListInputOverrides_StorageClassKafka(t *testing.T) {
 	schema := map[string]any{}
 	applyListInputOverrides(schema, "Kafka", map[string]any{})
diff --git a/internal/controller/dashboard/factory.go b/internal/controller/dashboard/factory.go
index 539e503b..fe0cadfd 100644
--- a/internal/controller/dashboard/factory.go
+++ b/internal/controller/dashboard/factory.go
@@ -47,6 +47,7 @@ func (m *Manager) ensureFactory(ctx context.Context, crd *cozyv1alpha1.Applicati
 	if prefix, ok := vncTabPrefix(kind); ok {
 		tabs = append(tabs, vncTab(prefix))
 	}
+	tabs = append(tabs, eventsTab(kind))
 	tabs = append(tabs, yamlTab(g, v, plural))
 
 	// Use unified factory creation
@@ -358,6 +359,37 @@ func secretsTab(kind string) map[string]any {
 	}
 }
 
+// eventsTab shows Kubernetes Events scoped to the application's namespace.
+// Events are namespace-scoped because Kubernetes Events don't carry application
+// labels and cannot be filtered by label selector. In Cozystack's multi-tenancy
+// model, each tenant namespace maps to a single application scope, so namespace
+// filtering provides the correct event scope.
+// For Tenant applications, events are fetched from status.namespace (the tenant's
+// own namespace) instead of the parent namespace where the Tenant object lives.
+func eventsTab(kind string) map[string]any {
+	nsPlaceholder := "{3}"
+	if kind == "Tenant" {
+		nsPlaceholder = "{reqsJsonPath[0]['.status.namespace']}"
+	}
+	return map[string]any{
+		"key":   "events",
+		"label": "Events",
+		"children": []any{
+			map[string]any{
+				"type": "EnrichedTable",
+				"data": map[string]any{
+					"id":              "events-table",
+					"fetchUrl":        "/api/clusters/{2}/k8s/api/v1/namespaces/" + nsPlaceholder + "/events",
+					"cluster":         "{2}",
+					"baseprefix":      "/openapi-ui",
+					"customizationId": "factory-details-events",
+					"pathToItems":     []any{"items"},
+				},
+			},
+		},
+	}
+}
+
 func yamlTab(group, version, plural string) map[string]any {
 	return map[string]any{
 		"key":   "yaml",
diff --git a/internal/controller/dashboard/factory_test.go b/internal/controller/dashboard/factory_test.go
new file mode 100644
index 00000000..b3742564
--- /dev/null
+++ b/internal/controller/dashboard/factory_test.go
@@ -0,0 +1,60 @@
+package dashboard
+
+import (
+	"testing"
+)
+
+func TestEventsTab_Structure(t *testing.T) {
+	tab := eventsTab("PostgreSQL")
+
+	if tab["key"] != "events" {
+		t.Errorf("expected key=events, got %v", tab["key"])
+	}
+	if tab["label"] != "Events" {
+		t.Errorf("expected label=Events, got %v", tab["label"])
+	}
+
+	children, ok := tab["children"].([]any)
+	if !ok || len(children) != 1 {
+		t.Fatal("expected exactly 1 child in events tab")
+	}
+
+	table, ok := children[0].(map[string]any)
+	if !ok {
+		t.Fatal("child is not a map")
+	}
+	if table["type"] != "EnrichedTable" {
+		t.Errorf("expected type=EnrichedTable, got %v", table["type"])
+	}
+
+	data, ok := table["data"].(map[string]any)
+	if !ok {
+		t.Fatal("table data is not a map")
+	}
+	if data["id"] != "events-table" {
+		t.Errorf("expected id=events-table, got %v", data["id"])
+	}
+	if data["fetchUrl"] != "/api/clusters/{2}/k8s/api/v1/namespaces/{3}/events" {
+		t.Errorf("unexpected fetchUrl for non-Tenant: %v", data["fetchUrl"])
+	}
+	if data["customizationId"] != "factory-details-events" {
+		t.Errorf("expected customizationId=factory-details-events, got %v", data["customizationId"])
+	}
+
+	pathToItems, ok := data["pathToItems"].([]any)
+	if !ok || len(pathToItems) != 1 || pathToItems[0] != "items" {
+		t.Errorf("expected pathToItems=[items], got %v", data["pathToItems"])
+	}
+}
+
+func TestEventsTab_TenantUsesStatusNamespace(t *testing.T) {
+	tab := eventsTab("Tenant")
+	children := tab["children"].([]any)
+	table := children[0].(map[string]any)
+	data := table["data"].(map[string]any)
+
+	expectedURL := "/api/clusters/{2}/k8s/api/v1/namespaces/{reqsJsonPath[0]['.status.namespace']}/events"
+	if data["fetchUrl"] != expectedURL {
+		t.Errorf("expected Tenant fetchUrl to use status.namespace, got %v", data["fetchUrl"])
+	}
+}
diff --git a/internal/controller/dashboard/sidebar.go b/internal/controller/dashboard/sidebar.go
index 90721b3d..69c933fb 100644
--- a/internal/controller/dashboard/sidebar.go
+++ b/internal/controller/dashboard/sidebar.go
@@ -144,6 +144,9 @@ func (m *Manager) ensureSidebar(ctx context.Context, crd *cozyv1alpha1.Applicati
 	// Add sidebar for backups.cozystack.io Backup resource
 	keysAndTags["backups"] = []any{"backup-sidebar"}
 
+	// Add sidebar for backups.cozystack.io RestoreJob resource
+	keysAndTags["restorejobs"] = []any{"restorejob-sidebar"}
+
 	// 3) Sort items within each category by Weight (desc), then Label (A→Z)
 	for cat := range categories {
 		sort.Slice(categories[cat], func(i, j int) bool {
@@ -215,6 +218,11 @@ func (m *Manager) ensureSidebar(ctx context.Context, crd *cozyv1alpha1.Applicati
 				"label": "Backups",
 				"link":  "/openapi-ui/{cluster}/{namespace}/api-table/backups.cozystack.io/v1alpha1/backups",
 			},
+			map[string]any{
+				"key":   "restorejobs",
+				"label": "RestoreJobs",
+				"link":  "/openapi-ui/{cluster}/{namespace}/api-table/backups.cozystack.io/v1alpha1/restorejobs",
+			},
 		},
 	})
 
@@ -258,6 +266,7 @@ func (m *Manager) ensureSidebar(ctx context.Context, crd *cozyv1alpha1.Applicati
 		"stock-project-factory-plan-details",
 		"stock-project-factory-backupjob-details",
 		"stock-project-factory-backup-details",
+		"stock-project-factory-restorejob-details",
 		"stock-project-factory-external-ips",
 		"stock-project-api-form",
 		"stock-project-api-table",
@@ -272,7 +281,12 @@ func (m *Manager) ensureSidebar(ctx context.Context, crd *cozyv1alpha1.Applicati
 		"stock-instance-builtin-table",
 	}
 
-	// Add details sidebars for all CRDs with dashboard config
+	// Add details sidebars for all CRDs with dashboard config, and collect
+	// the set of IDs that are genuinely CRD-backed (dynamic). The hardcoded
+	// `-details` IDs above (e.g. kube-* and backup/backupjob/plan/restorejob)
+	// are not tied to an ApplicationDefinition and must be treated as static
+	// so they receive consistent labels via upsertMultipleSidebars().
+	dynamicDetailsIDs := map[string]bool{}
 	for i := range all {
 		def := &all[i]
 		if def.Spec.Dashboard == nil {
@@ -282,17 +296,22 @@ func (m *Manager) ensureSidebar(ctx context.Context, crd *cozyv1alpha1.Applicati
 		lowerKind := strings.ToLower(kind)
 		detailsID := fmt.Sprintf("stock-project-factory-%s-details", lowerKind)
 		targetIDs = append(targetIDs, detailsID)
+		dynamicDetailsIDs[detailsID] = true
 	}
 
 	// 7) Upsert all target sidebars with identical menuItems and keysAndTags
-	return m.upsertMultipleSidebars(ctx, crd, targetIDs, keysAndTags, menuItems)
+	return m.upsertMultipleSidebars(ctx, crd, targetIDs, dynamicDetailsIDs, keysAndTags, menuItems)
 }
 
 // upsertMultipleSidebars creates/updates several Sidebar resources with the same menu spec.
+// dynamicDetailsIDs identifies `stock-project-factory-<kind>-details` sidebars that are
+// backed by an ApplicationDefinition and should therefore be owned by that CRD.
+// Any other ID is treated as a static sidebar (managed-by labels, no owner ref).
 func (m *Manager) upsertMultipleSidebars(
 	ctx context.Context,
 	crd *cozyv1alpha1.ApplicationDefinition,
 	ids []string,
+	dynamicDetailsIDs map[string]bool,
 	keysAndTags map[string]any,
 	menuItems []any,
 ) error {
@@ -308,8 +327,10 @@ func (m *Manager) upsertMultipleSidebars(
 
 		if _, err := controllerutil.CreateOrUpdate(ctx, m.Client, obj, func() error {
 			// Only set owner reference for dynamic sidebars (stock-project-factory-{kind}-details)
-			// Static sidebars (stock-instance-*, stock-project-*) should not have owner references
-			if strings.HasPrefix(id, "stock-project-factory-") && strings.HasSuffix(id, "-details") {
+			// that are actually backed by an ApplicationDefinition. Static sidebars — including
+			// hardcoded details sidebars for built-in/backup resources — must fall through to the
+			// static-label branch so they're managed consistently.
+			if strings.HasPrefix(id, "stock-project-factory-") && strings.HasSuffix(id, "-details") && dynamicDetailsIDs[id] {
 				// This is a dynamic sidebar, set owner reference only if it matches the current CRD
 				_, _, kind := pickGVK(crd)
 				lowerKind := strings.ToLower(kind)
diff --git a/internal/controller/dashboard/static_helpers.go b/internal/controller/dashboard/static_helpers.go
index 1a1018cb..db37e05b 100644
--- a/internal/controller/dashboard/static_helpers.go
+++ b/internal/controller/dashboard/static_helpers.go
@@ -650,12 +650,31 @@ func createStringColumn(name, jsonPath string) map[string]any {
 	}
 }
 
-// createTimestampColumn creates a timestamp column with custom formatting
-func createTimestampColumn(name, jsonPath string) map[string]any {
+// createTimestampColumn creates a timestamp column with custom formatting.
+// Extra jsonPaths act as ordered fallbacks: the first non-null path wins,
+// and `-` is used only when every path is null. Depth is capped at two
+// because the template parser's non-greedy matcher cannot track more than
+// one level of alternating quotes in a nested reqsJsonPath fallback.
+func createTimestampColumn(name string, jsonPaths ...string) map[string]any {
+	if len(jsonPaths) == 0 {
+		panic("createTimestampColumn requires at least one jsonPath")
+	}
+	if len(jsonPaths) > 2 {
+		panic("createTimestampColumn supports at most two jsonPaths (primary + one fallback)")
+	}
+
+	var text string
+	switch len(jsonPaths) {
+	case 1:
+		text = "{reqsJsonPath[0]['" + jsonPaths[0] + "']['-']}"
+	case 2:
+		text = "{reqsJsonPath[0]['" + jsonPaths[0] + "'][\"{reqsJsonPath[0]['" + jsonPaths[1] + "']['-']}\"]}"
+	}
+
 	return map[string]any{
 		"name":     name,
 		"type":     "factory",
-		"jsonPath": jsonPath,
+		"jsonPath": jsonPaths[0],
 		"customProps": map[string]any{
 			"disableEventBubbling": true,
 			"items": []any{
@@ -673,7 +692,7 @@ func createTimestampColumn(name, jsonPath string) map[string]any {
 							"data": map[string]any{
 								"formatter": "timestamp",
 								"id":        "time-value",
-								"text":      "{reqsJsonPath[0]['" + jsonPath + "']['-']}",
+								"text":      text,
 							},
 						},
 					},
diff --git a/internal/controller/dashboard/static_refactored.go b/internal/controller/dashboard/static_refactored.go
index 06c4239f..9f28c0fb 100644
--- a/internal/controller/dashboard/static_refactored.go
+++ b/internal/controller/dashboard/static_refactored.go
@@ -224,6 +224,20 @@ func CreateAllCustomColumnsOverrides() []*dashboardv1alpha1.CustomColumnsOverrid
 			createStringColumn("Name", ".name"),
 		}),
 
+		// Factory details events. Event Time falls back to `.firstTimestamp`
+		// because core/v1 Events (Helm, controller-runtime) populate only the
+		// legacy timestamps while events.k8s.io/v1 Events populate only
+		// `.eventTime`; taking the first non-null of the two covers both.
+		createCustomColumnsOverride("factory-details-events", []any{
+			createTimestampColumn("Last Seen", ".lastTimestamp", ".eventTime"),
+			createTimestampColumn("Event Time", ".eventTime", ".firstTimestamp"),
+			createStringColumn("Type", ".type"),
+			createStringColumn("Reason", ".reason"),
+			createStringColumn("Object", ".involvedObject.kind"),
+			createStringColumn("Name", ".involvedObject.name"),
+			createStringColumn("Message", ".message"),
+		}),
+
 		// Factory status conditions
 		createCustomColumnsOverride("factory-status-conditions", []any{
 			createStringColumn("Type", ".type"),
@@ -411,6 +425,14 @@ func CreateAllCustomColumnsOverrides() []*dashboardv1alpha1.CustomColumnsOverrid
 			createTimestampColumn("Taken At", ".spec.takenAt"),
 			createTimestampColumn("Created", ".metadata.creationTimestamp"),
 		}),
+
+		// Stock namespace backups cozystack io v1alpha1 restorejobs
+		createCustomColumnsOverride("stock-namespace-/backups.cozystack.io/v1alpha1/restorejobs", []any{
+			createCustomColumnWithJsonPath("Name", ".metadata.name", "RestoreJob", "", "/openapi-ui/{2}/{reqsJsonPath[0]['.metadata.namespace']['-']}/factory/restorejob-details/{reqsJsonPath[0]['.metadata.name']['-']}"),
+			createStringColumn("Phase", ".status.phase"),
+			createStringColumn("Backup", ".spec.backupRef.name"),
+			createTimestampColumn("Created", ".metadata.creationTimestamp"),
+		}),
 	}
 }
 
@@ -533,6 +555,31 @@ func CreateAllCustomFormsOverrides() []*dashboardv1alpha1.CustomFormsOverride {
 				"backupClassName": listInputScemaItemBackupClass(),
 			}),
 		}),
+
+		// RestoreJobs form override - backups.cozystack.io/v1alpha1
+		createCustomFormsOverride("default-/backups.cozystack.io/v1alpha1/restorejobs", map[string]any{
+			"formItems": []any{
+				createFormItem("metadata.name", "Name", "text"),
+				createFormItem("metadata.namespace", "Namespace", "text"),
+				createFormItem("spec.backupRef.name", "Backup", "text"),
+				// Target application: leave empty to restore into the same application
+				// as referenced by the selected Backup. Fill all three to restore into
+				// a different application (e.g. rename, or restore into a new target).
+				createFormItem("spec.targetApplicationRef.apiGroup", "Target Application API Group (optional, used to restore into a different application)", "text"),
+				createFormItem("spec.targetApplicationRef.kind", "Target Application Kind (optional, used to restore into a different application)", "text"),
+				createFormItem("spec.targetApplicationRef.name", "Target Application Name (optional, used to restore into a different application)", "text"),
+				// Driver-specific options (key-value editor). Refer to the backup driver
+				// documentation for supported keys.
+				createFormItem("spec.options", "Options (driver-specific key/value pairs)", "object"),
+			},
+			"schema": createSchema(map[string]any{
+				"backupRef": map[string]any{
+					"properties": map[string]any{
+						"name": listInputSchemaItemBackup(),
+					},
+				},
+			}),
+		}),
 	}
 }
 
@@ -1923,6 +1970,177 @@ func CreateAllFactories() []*dashboardv1alpha1.Factory {
 	}
 	backupSpec := createUnifiedFactory(backupConfig, backupTabs, []any{"/api/clusters/{2}/k8s/apis/backups.cozystack.io/v1alpha1/namespaces/{3}/backups/{6}"})
 
+	// RestoreJob details factory using unified approach
+	restoreJobConfig := UnifiedResourceConfig{
+		Name:         "restorejob-details",
+		ResourceType: "factory",
+		Kind:         "RestoreJob",
+		Plural:       "restorejobs",
+		Title:        "restorejob",
+	}
+	restoreJobTabs := []any{
+		map[string]any{
+			"key":   "details",
+			"label": "Details",
+			"children": []any{
+				contentCard("details-card", map[string]any{
+					"marginBottom": "24px",
+				}, []any{
+					antdText("details-title", true, "RestoreJob details", map[string]any{
+						"fontSize":     20,
+						"marginBottom": "12px",
+					}),
+					spacer("details-spacer", 16),
+					antdRow("details-grid", []any{48, 12}, []any{
+						antdCol("col-left", 12, []any{
+							antdFlexVertical("col-left-stack", 24, []any{
+								antdFlexVertical("meta-name-block", 4, []any{
+									antdText("meta-name-label", true, "Name", nil),
+									parsedText("meta-name-value", "{reqsJsonPath[0]['.metadata.name']['-']}", nil),
+								}),
+								antdFlexVertical("meta-namespace-block", 8, []any{
+									antdText("meta-namespace-label", true, "Namespace", nil),
+									antdFlex("header-row", 6, []any{
+										map[string]any{
+											"type": "antdText",
+											"data": map[string]any{
+												"id":    "header-badge",
+												"text":  "NS",
+												"title": "namespace",
+												"style": map[string]any{
+													"backgroundColor": "#a25792ff",
+													"borderRadius":    "20px",
+													"color":           "#fff",
+													"display":         "inline-block",
+													"fontFamily":      "RedHatDisplay, Overpass, overpass, helvetica, arial, sans-serif",
+													"fontSize":        "15px",
+													"fontWeight":      400,
+													"lineHeight":      "24px",
+													"minWidth":        24,
+													"padding":         "0 9px",
+													"textAlign":       "center",
+													"whiteSpace":      "nowrap",
+												},
+											},
+										},
+										map[string]any{
+											"type": "antdLink",
+											"data": map[string]any{
+												"id":   "namespace-link",
+												"text": "{reqsJsonPath[0]['.metadata.namespace']['-']}",
+												"href": "/openapi-ui/{2}/{reqsJsonPath[0]['.metadata.namespace']['-']}/factory/marketplace",
+											},
+										},
+									}),
+								}),
+								antdFlexVertical("meta-created-block", 4, []any{
+									antdText("created-time-label", true, "Created", nil),
+									antdFlex("created-time-block", 6, []any{
+										map[string]any{
+											"type": "antdText",
+											"data": map[string]any{
+												"id":   "created-time-icon",
+												"text": "🌐",
+											},
+										},
+										map[string]any{
+											"type": "parsedText",
+											"data": map[string]any{
+												"formatter": "timestamp",
+												"id":        "created-time-value",
+												"text":      "{reqsJsonPath[0]['.metadata.creationTimestamp']['-']}",
+											},
+										},
+									}),
+								}),
+							}),
+						}),
+						antdCol("col-right", 12, []any{
+							antdFlexVertical("col-right-stack", 24, []any{
+								antdFlexVertical("status-phase-block", 4, []any{
+									antdText("phase-label", true, "Phase", nil),
+									parsedText("phase-value", "{reqsJsonPath[0]['.status.phase']['-']}", nil),
+								}),
+								antdFlexVertical("spec-backup-ref-block", 4, []any{
+									antdText("backup-ref-label", true, "Backup Ref", nil),
+									parsedText("backup-ref-value", "{reqsJsonPath[0]['.spec.backupRef.name']['-']}", nil),
+								}),
+								antdFlexVertical("spec-target-application-ref-block", 4, []any{
+									antdText("target-application-ref-label", true, "Target Application", nil),
+									// targetApplicationRef is optional — when absent, the restore targets
+									// the same application as the selected Backup. Show a single friendly
+									// fallback in that case instead of rendering "-.-/-".
+									parsedText("target-application-ref-value", "{reqsJsonPath[0]['.spec.targetApplicationRef.name']['Same as backup']}", nil),
+								}),
+								antdFlexVertical("spec-options-target-namespace-block", 4, []any{
+									antdText("options-target-namespace-label", true, "Target Namespace", nil),
+									parsedText("options-target-namespace-value", "{reqsJsonPath[0]['.spec.options.targetNamespace']['-']}", nil),
+								}),
+								antdFlexVertical("spec-options-fail-if-target-exists-block", 4, []any{
+									antdText("options-fail-if-target-exists-label", true, "Fail If Target Exists", nil),
+									parsedText("options-fail-if-target-exists-value", "{reqsJsonPath[0]['.spec.options.failIfTargetExists']['-']}", nil),
+								}),
+								antdFlexVertical("spec-options-keep-original-pvc-block", 4, []any{
+									antdText("options-keep-original-pvc-label", true, "Keep Original PVC", nil),
+									parsedText("options-keep-original-pvc-value", "{reqsJsonPath[0]['.spec.options.keepOriginalPVC']['-']}", nil),
+								}),
+								antdFlexVertical("spec-options-keep-original-ip-mac-block", 4, []any{
+									antdText("options-keep-original-ip-mac-label", true, "Keep Original IP/MAC", nil),
+									parsedText("options-keep-original-ip-mac-value", "{reqsJsonPath[0]['.spec.options.keepOriginalIpAndMac']['-']}", nil),
+								}),
+								antdFlexVertical("status-started-at-block", 4, []any{
+									antdText("started-at-label", true, "Started At", nil),
+									antdFlex("started-at-time-block", 6, []any{
+										map[string]any{
+											"type": "antdText",
+											"data": map[string]any{
+												"id":   "started-at-time-icon",
+												"text": "🌐",
+											},
+										},
+										map[string]any{
+											"type": "parsedText",
+											"data": map[string]any{
+												"formatter": "timestamp",
+												"id":        "started-at-time-value",
+												"text":      "{reqsJsonPath[0]['.status.startedAt']['-']}",
+											},
+										},
+									}),
+								}),
+								antdFlexVertical("status-completed-at-block", 4, []any{
+									antdText("completed-at-label", true, "Completed At", nil),
+									antdFlex("completed-at-time-block", 6, []any{
+										map[string]any{
+											"type": "antdText",
+											"data": map[string]any{
+												"id":   "completed-at-time-icon",
+												"text": "🌐",
+											},
+										},
+										map[string]any{
+											"type": "parsedText",
+											"data": map[string]any{
+												"formatter": "timestamp",
+												"id":        "completed-at-time-value",
+												"text":      "{reqsJsonPath[0]['.status.completedAt']['-']}",
+											},
+										},
+									}),
+								}),
+								antdFlexVertical("status-message-block", 4, []any{
+									antdText("message-label", true, "Message", nil),
+									parsedText("message-value", "{reqsJsonPath[0]['.status.message']['-']}", nil),
+								}),
+							}),
+						}),
+					}),
+				}),
+			},
+		},
+	}
+	restoreJobSpec := createUnifiedFactory(restoreJobConfig, restoreJobTabs, []any{"/api/clusters/{2}/k8s/apis/backups.cozystack.io/v1alpha1/namespaces/{3}/restorejobs/{6}"})
+
 	// External IPs factory (filtered services)
 	externalIPsTabs := []any{
 		map[string]any{
@@ -1975,6 +2193,7 @@ func CreateAllFactories() []*dashboardv1alpha1.Factory {
 		createFactory("plan-details", planSpec),
 		createFactory("backupjob-details", backupJobSpec),
 		createFactory("backup-details", backupSpec),
+		createFactory("restorejob-details", restoreJobSpec),
 		createFactory("external-ips", externalIPsSpec),
 	}
 }
@@ -1994,9 +2213,10 @@ func CreateAllNavigations() []*dashboardv1alpha1.Navigation {
 		"base-factory-namespaced-api-networking.k8s.io-v1-ingresses":         "kube-ingress-details",
 		"base-factory-namespaced-api-cozystack.io-v1alpha1-workloadmonitors": "workloadmonitor-details",
 		// Backup resources (not ApplicationDefinitions, so ensureNavigation doesn't cover them)
-		"base-factory-namespaced-api-backups.cozystack.io-v1alpha1-plans":      "plan-details",
-		"base-factory-namespaced-api-backups.cozystack.io-v1alpha1-backupjobs": "backupjob-details",
-		"base-factory-namespaced-api-backups.cozystack.io-v1alpha1-backups":    "backup-details",
+		"base-factory-namespaced-api-backups.cozystack.io-v1alpha1-plans":       "plan-details",
+		"base-factory-namespaced-api-backups.cozystack.io-v1alpha1-backupjobs":  "backupjob-details",
+		"base-factory-namespaced-api-backups.cozystack.io-v1alpha1-backups":     "backup-details",
+		"base-factory-namespaced-api-backups.cozystack.io-v1alpha1-restorejobs": "restorejob-details",
 	}
 
 	return []*dashboardv1alpha1.Navigation{
@@ -2121,6 +2341,19 @@ func listInputScemaItemBackupClass() map[string]any {
 	}
 }
 
+// listInputSchemaItemBackup returns a listInput schema overlay for selecting a Backup
+// from the current namespace (used by RestoreJob form for spec.backupRef.name).
+func listInputSchemaItemBackup() map[string]any {
+	return map[string]any{
+		"type": "listInput",
+		"customProps": map[string]any{
+			"valueUri":    "/api/clusters/{cluster}/k8s/apis/backups.cozystack.io/v1alpha1/namespaces/{namespace}/backups",
+			"keysToValue": []any{"metadata", "name"},
+			"keysToLabel": []any{"metadata", "name"},
+		},
+	}
+}
+
 // backupClassSchema returns the schema for spec.backupClassName as listInput (BackupJob/Plan).
 func createSchema(customProps map[string]any) map[string]any {
 	return map[string]any{
diff --git a/internal/controller/workloadmonitor_controller.go b/internal/controller/workloadmonitor_controller.go
index a684df9b..35906a87 100644
--- a/internal/controller/workloadmonitor_controller.go
+++ b/internal/controller/workloadmonitor_controller.go
@@ -4,7 +4,13 @@ import (
 	"context"
 	"encoding/json"
 	"fmt"
+	"io"
+	"net/http"
+	"net/url"
 	"sort"
+	"strconv"
+	"strings"
+	"time"
 
 	apierrors "k8s.io/apimachinery/pkg/api/errors"
 	"k8s.io/apimachinery/pkg/runtime"
@@ -22,8 +28,29 @@ import (
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 
 	cozyv1alpha1 "github.com/cozystack/cozystack/api/v1alpha1"
+	cosiv1alpha1 "sigs.k8s.io/container-object-storage-interface-api/apis/objectstorage/v1alpha1"
 )
 
+const (
+	// namespaceMonitoringLabel is the namespace label that indicates which tenant
+	// namespace hosts the monitoring stack (VictoriaMetrics/Prometheus).
+	namespaceMonitoringLabel = "namespace.cozystack.io/monitoring"
+	workloadLabelPrefix      = "workloads.cozystack.io/"
+	// workloadMonitorLabel is reserved: it names the WorkloadMonitor that owns
+	// the Workload and is always set by the reconciler, so it is never copied
+	// from monitor labels.
+	workloadMonitorLabel = workloadLabelPrefix + "monitor"
+	// vmSelectService is the well-known service name for VictoriaMetrics vmselect
+	// within a monitoring namespace. Port 8481, path /select/0/prometheus.
+	vmSelectService = "vmselect-shortterm"
+	vmSelectPort    = "8481"
+	vmSelectPath    = "/select/0/prometheus"
+)
+
+// prometheusHTTPClient is a dedicated HTTP client for Prometheus queries,
+// avoiding the shared http.DefaultClient global.
+var prometheusHTTPClient = &http.Client{Timeout: 10 * time.Second}
+
 // WorkloadMonitorReconciler reconciles a WorkloadMonitor object
 type WorkloadMonitorReconciler struct {
 	client.Client
@@ -36,6 +63,13 @@ type WorkloadMonitorReconciler struct {
 // +kubebuilder:rbac:groups=cozystack.io,resources=workloads/status,verbs=get;update;patch
 // +kubebuilder:rbac:groups=core,resources=pods,verbs=get;list;watch
 // +kubebuilder:rbac:groups=core,resources=persistentvolumeclaims,verbs=get;list;watch
+// +kubebuilder:rbac:groups=core,resources=namespaces,verbs=get
+// +kubebuilder:rbac:groups=objectstorage.k8s.io,resources=bucketclaims,verbs=get;list;watch
+
+// isBucketClaimReady checks if the BucketClaim has been provisioned.
+func (r *WorkloadMonitorReconciler) isBucketClaimReady(bc *cosiv1alpha1.BucketClaim) bool {
+	return bc.Status.BucketReady
+}
 
 // isServiceReady checks if the service has an external IP bound
 func (r *WorkloadMonitorReconciler) isServiceReady(svc *corev1.Service) bool {
@@ -101,6 +135,190 @@ func updateOwnerReferences(obj metav1.Object, monitor client.Object) {
 	obj.SetOwnerReferences(owners)
 }
 
+// resolvePrometheusURL returns the Prometheus-compatible API base URL for the given namespace.
+// It reads the namespace.cozystack.io/monitoring label to find the monitoring namespace,
+// then constructs the vmselect URL. Returns empty string if monitoring is not configured.
+func (r *WorkloadMonitorReconciler) resolvePrometheusURL(ctx context.Context, namespace string) string {
+	logger := log.FromContext(ctx)
+	ns := &corev1.Namespace{}
+	if err := r.Get(ctx, types.NamespacedName{Name: namespace}, ns); err != nil {
+		logger.V(1).Info("Failed to read namespace for monitoring resolution", "namespace", namespace, "error", err)
+		return ""
+	}
+	monitoringNS := ns.Labels[namespaceMonitoringLabel]
+	if monitoringNS == "" {
+		return ""
+	}
+	return fmt.Sprintf("http://%s.%s.svc:%s%s", vmSelectService, monitoringNS, vmSelectPort, vmSelectPath)
+}
+
+// bucketMetrics holds size metrics for a single bucket, keyed by metric name.
+type bucketMetrics struct {
+	LogicalSize  int64
+	PhysicalSize int64
+	HasLogical   bool
+	HasPhysical  bool
+}
+
+// queryAllBucketMetrics fetches SeaweedFS bucket size metrics for the given
+// bucket names in a single Prometheus query and returns them keyed by bucket
+// name. The query is scoped to only the requested buckets to avoid fetching
+// metrics for buckets belonging to other WorkloadMonitors.
+func (r *WorkloadMonitorReconciler) queryAllBucketMetrics(ctx context.Context, prometheusBaseURL string, bucketNames []string) map[string]*bucketMetrics {
+	result := make(map[string]*bucketMetrics)
+	if prometheusBaseURL == "" || len(bucketNames) == 0 {
+		return result
+	}
+	logger := log.FromContext(ctx)
+
+	query := fmt.Sprintf(`{__name__=~"SeaweedFS_s3_bucket_(size|physical_size)_bytes",bucket=~"%s"}`, strings.Join(bucketNames, "|"))
+	u, err := url.Parse(strings.TrimRight(prometheusBaseURL, "/") + "/api/v1/query")
+	if err != nil {
+		logger.Error(err, "Failed to parse Prometheus URL")
+		return result
+	}
+	u.RawQuery = url.Values{"query": {query}}.Encode()
+
+	httpCtx, cancel := context.WithTimeout(ctx, 10*time.Second)
+	defer cancel()
+
+	req, err := http.NewRequestWithContext(httpCtx, http.MethodGet, u.String(), nil)
+	if err != nil {
+		logger.Error(err, "Failed to create Prometheus request")
+		return result
+	}
+
+	resp, err := prometheusHTTPClient.Do(req)
+	if err != nil {
+		logger.V(1).Info("Failed to query Prometheus for bucket metrics", "error", err)
+		return result
+	}
+	defer resp.Body.Close()
+
+	if resp.StatusCode != http.StatusOK {
+		logger.V(1).Info("Prometheus returned non-OK status for bucket metrics", "status", resp.StatusCode)
+		return result
+	}
+
+	body, err := io.ReadAll(io.LimitReader(resp.Body, 4<<20))
+	if err != nil {
+		logger.Error(err, "Failed to read Prometheus response")
+		return result
+	}
+
+	var promResp struct {
+		Status string `json:"status"`
+		Data   struct {
+			Result []struct {
+				Metric map[string]string  `json:"metric"`
+				Value  [2]json.RawMessage `json:"value"`
+			} `json:"result"`
+		} `json:"data"`
+	}
+	if err := json.Unmarshal(body, &promResp); err != nil {
+		logger.Error(err, "Failed to parse Prometheus response")
+		return result
+	}
+	if promResp.Status != "success" {
+		return result
+	}
+
+	for _, r := range promResp.Data.Result {
+		bucket := r.Metric["bucket"]
+		metricName := r.Metric["__name__"]
+		if bucket == "" || metricName == "" {
+			continue
+		}
+
+		var valueStr string
+		if err := json.Unmarshal(r.Value[1], &valueStr); err != nil {
+			continue
+		}
+		val, err := strconv.ParseFloat(valueStr, 64)
+		if err != nil {
+			continue
+		}
+
+		bm, ok := result[bucket]
+		if !ok {
+			bm = &bucketMetrics{}
+			result[bucket] = bm
+		}
+
+		switch metricName {
+		case "SeaweedFS_s3_bucket_size_bytes":
+			bm.LogicalSize = int64(val)
+			bm.HasLogical = true
+		case "SeaweedFS_s3_bucket_physical_size_bytes":
+			bm.PhysicalSize = int64(val)
+			bm.HasPhysical = true
+		}
+	}
+
+	return result
+}
+
+// reconcileBucketClaimForMonitor creates or updates a Workload object for the given BucketClaim and WorkloadMonitor.
+func (r *WorkloadMonitorReconciler) reconcileBucketClaimForMonitor(
+	ctx context.Context,
+	monitor *cozyv1alpha1.WorkloadMonitor,
+	bc cosiv1alpha1.BucketClaim,
+	allMetrics map[string]*bucketMetrics,
+) error {
+	logger := log.FromContext(ctx)
+	workload := &cozyv1alpha1.Workload{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      fmt.Sprintf("bucket-%s", bc.Name),
+			Namespace: bc.Namespace,
+			Labels:    make(map[string]string, len(bc.Labels)),
+		},
+	}
+
+	resources := make(map[string]resource.Quantity)
+
+	// Look up pre-fetched bucket metrics by the SeaweedFS bucket name.
+	// bc.Status.BucketName is the COSI Bucket name, which the COSI driver
+	// uses directly as the SeaweedFS bucket name.
+	if bm, ok := allMetrics[bc.Status.BucketName]; ok {
+		if bm.HasLogical {
+			resources["s3-storage-bytes"] = *resource.NewQuantity(bm.LogicalSize, resource.BinarySI)
+		}
+		if bm.HasPhysical {
+			resources["s3-physical-storage-bytes"] = *resource.NewQuantity(bm.PhysicalSize, resource.BinarySI)
+		}
+	}
+
+	monitorLabels := r.getMonitorLabels(monitor)
+	_, err := ctrl.CreateOrUpdate(ctx, r.Client, workload, func() error {
+		updateOwnerReferences(workload.GetObjectMeta(), &bc)
+
+		if workload.Labels == nil {
+			workload.Labels = make(map[string]string)
+		}
+		// Apply monitor-level labels first so source-object labels can override on conflict
+		for k, v := range monitorLabels {
+			workload.Labels[k] = v
+		}
+		for k, v := range bc.Labels {
+			workload.Labels[k] = v
+		}
+		workload.Labels[workloadMonitorLabel] = monitor.Name
+
+		workload.Status.Kind = monitor.Spec.Kind
+		workload.Status.Type = monitor.Spec.Type
+		workload.Status.Resources = resources
+		workload.Status.Operational = r.isBucketClaimReady(&bc)
+
+		return nil
+	})
+	if err != nil {
+		logger.Error(err, "Failed to CreateOrUpdate Workload", "workload", workload.Name)
+		return err
+	}
+
+	return nil
+}
+
 // reconcileServiceForMonitor creates or updates a Workload object for the given Service and WorkloadMonitor.
 func (r *WorkloadMonitorReconciler) reconcileServiceForMonitor(
 	ctx context.Context,
@@ -137,14 +355,22 @@ func (r *WorkloadMonitorReconciler) reconcileServiceForMonitor(
 	resourceLabel = fmt.Sprintf("%s.ipaddresspool.metallb.io/requests.ipaddresses", resourceLabel)
 	resources[resourceLabel] = quantity
 
+	monitorLabels := r.getMonitorLabels(monitor)
 	_, err := ctrl.CreateOrUpdate(ctx, r.Client, workload, func() error {
 		// Update owner references with the new monitor
 		updateOwnerReferences(workload.GetObjectMeta(), &svc)
 
+		// Apply monitor-level labels first so source-object labels can override on conflict
+		if workload.Labels == nil {
+			workload.Labels = make(map[string]string)
+		}
+		for k, v := range monitorLabels {
+			workload.Labels[k] = v
+		}
 		for k, v := range svc.Labels {
 			workload.Labels[k] = v
 		}
-		workload.Labels["workloads.cozystack.io/monitor"] = monitor.Name
+		workload.Labels[workloadMonitorLabel] = monitor.Name
 
 		// Fill Workload status fields:
 		workload.Status.Kind = monitor.Spec.Kind
@@ -188,14 +414,22 @@ func (r *WorkloadMonitorReconciler) reconcilePVCForMonitor(
 		resources[resourceLabel] = resourceQuantity
 	}
 
+	monitorLabels := r.getMonitorLabels(monitor)
 	_, err := ctrl.CreateOrUpdate(ctx, r.Client, workload, func() error {
 		// Update owner references with the new monitor
 		updateOwnerReferences(workload.GetObjectMeta(), &pvc)
 
+		// Apply monitor-level labels first so source-object labels can override on conflict
+		if workload.Labels == nil {
+			workload.Labels = make(map[string]string)
+		}
+		for k, v := range monitorLabels {
+			workload.Labels[k] = v
+		}
 		for k, v := range pvc.Labels {
 			workload.Labels[k] = v
 		}
-		workload.Labels["workloads.cozystack.io/monitor"] = monitor.Name
+		workload.Labels[workloadMonitorLabel] = monitor.Name
 
 		// Fill Workload status fields:
 		workload.Status.Kind = monitor.Spec.Kind
@@ -262,14 +496,22 @@ func (r *WorkloadMonitorReconciler) reconcilePodForMonitor(
 	}
 
 	metaLabels := r.getWorkloadMetadata(&pod)
+	monitorLabels := r.getMonitorLabels(monitor)
 	_, err := ctrl.CreateOrUpdate(ctx, r.Client, workload, func() error {
 		// Update owner references with the new monitor
 		updateOwnerReferences(workload.GetObjectMeta(), &pod)
 
+		// Apply monitor-level labels first so source-object labels can override on conflict
+		if workload.Labels == nil {
+			workload.Labels = make(map[string]string)
+		}
+		for k, v := range monitorLabels {
+			workload.Labels[k] = v
+		}
 		for k, v := range pod.Labels {
 			workload.Labels[k] = v
 		}
-		workload.Labels["workloads.cozystack.io/monitor"] = monitor.Name
+		workload.Labels[workloadMonitorLabel] = monitor.Name
 
 		// Add workload meta to labels
 		for k, v := range metaLabels {
@@ -375,6 +617,34 @@ func (r *WorkloadMonitorReconciler) Reconcile(ctx context.Context, req ctrl.Requ
 		}
 	}
 
+	bucketClaimList := &cosiv1alpha1.BucketClaimList{}
+	if err := r.List(
+		ctx,
+		bucketClaimList,
+		client.InNamespace(monitor.Namespace),
+		client.MatchingLabels(monitor.Spec.Selector),
+	); err != nil {
+		logger.Error(err, "Unable to list BucketClaims for WorkloadMonitor", "monitor", monitor.Name)
+		return ctrl.Result{}, err
+	}
+
+	if len(bucketClaimList.Items) > 0 {
+		bucketPromURL := r.resolvePrometheusURL(ctx, monitor.Namespace)
+		var bucketNames []string
+		for _, bc := range bucketClaimList.Items {
+			if bc.Status.BucketName != "" {
+				bucketNames = append(bucketNames, bc.Status.BucketName)
+			}
+		}
+		allBucketMetrics := r.queryAllBucketMetrics(ctx, bucketPromURL, bucketNames)
+		for _, bc := range bucketClaimList.Items {
+			if err := r.reconcileBucketClaimForMonitor(ctx, monitor, bc, allBucketMetrics); err != nil {
+				logger.Error(err, "Failed to reconcile Workload for BucketClaim", "BucketClaim", bc.Name)
+				continue
+			}
+		}
+	}
+
 	// Update WorkloadMonitor status based on observed pods
 	monitor.Status.ObservedReplicas = observedReplicas
 	monitor.Status.AvailableReplicas = availableReplicas
@@ -388,10 +658,12 @@ func (r *WorkloadMonitorReconciler) Reconcile(ctx context.Context, req ctrl.Requ
 		fresh.Status.ObservedReplicas = observedReplicas
 		fresh.Status.AvailableReplicas = availableReplicas
 
-		// Default to operational = true, but check MinReplicas if set
-		monitor.Status.Operational = pointer.Bool(true)
-		if monitor.Spec.MinReplicas != nil && availableReplicas < *monitor.Spec.MinReplicas {
-			monitor.Status.Operational = pointer.Bool(false)
+		// Default to operational = true, but check MinReplicas if set.
+		// Use fresh.Spec to avoid making decisions based on a stale cached copy
+		// when the spec was updated between the initial read and this retry.
+		fresh.Status.Operational = pointer.Bool(true)
+		if fresh.Spec.MinReplicas != nil && availableReplicas < *fresh.Spec.MinReplicas {
+			fresh.Status.Operational = pointer.Bool(false)
 		}
 		return r.Status().Update(ctx, fresh)
 	})
@@ -400,7 +672,12 @@ func (r *WorkloadMonitorReconciler) Reconcile(ctx context.Context, req ctrl.Requ
 		return ctrl.Result{}, err
 	}
 
-	// Return without requeue if we want purely event-driven reconciliations
+	// Requeue periodically if there are BucketClaims to keep sizes up to date.
+	// Bucket sizes come from Prometheus metrics that update every 60s.
+	if len(bucketClaimList.Items) > 0 {
+		return ctrl.Result{RequeueAfter: 60 * time.Second}, nil
+	}
+
 	return ctrl.Result{}, nil
 }
 
@@ -419,6 +696,11 @@ func (r *WorkloadMonitorReconciler) SetupWithManager(mgr ctrl.Manager) error {
 			&corev1.PersistentVolumeClaim{},
 			handler.EnqueueRequestsFromMapFunc(mapObjectToMonitor(&corev1.PersistentVolumeClaim{}, r.Client)),
 		).
+		// Watch BucketClaims for S3 bucket billing
+		Watches(
+			&cosiv1alpha1.BucketClaim{},
+			handler.EnqueueRequestsFromMapFunc(mapObjectToMonitor(&cosiv1alpha1.BucketClaim{}, r.Client)),
+		).
 		// Watch for changes to Workload objects we create (owned by WorkloadMonitor)
 		Owns(&cozyv1alpha1.Workload{}).
 		Complete(r)
@@ -472,3 +754,21 @@ func (r *WorkloadMonitorReconciler) getWorkloadMetadata(obj client.Object) map[s
 	}
 	return labels
 }
+
+// getMonitorLabels extracts workloads.cozystack.io/* labels from a WorkloadMonitor
+// so they can be propagated onto Workload objects created for pods, PVCs, services,
+// or bucket claims. The monitor label "workloads.cozystack.io/monitor" is reserved
+// and set separately per Workload, so it is excluded here.
+func (r *WorkloadMonitorReconciler) getMonitorLabels(monitor *cozyv1alpha1.WorkloadMonitor) map[string]string {
+	labels := make(map[string]string)
+	for k, v := range monitor.GetLabels() {
+		if !strings.HasPrefix(k, workloadLabelPrefix) {
+			continue
+		}
+		if k == workloadMonitorLabel {
+			continue
+		}
+		labels[k] = v
+	}
+	return labels
+}
diff --git a/internal/controller/workloadmonitor_controller_test.go b/internal/controller/workloadmonitor_controller_test.go
new file mode 100644
index 00000000..355b0379
--- /dev/null
+++ b/internal/controller/workloadmonitor_controller_test.go
@@ -0,0 +1,846 @@
+package controller
+
+import (
+	"context"
+	"fmt"
+	"net/http"
+	"net/http/httptest"
+	"strings"
+	"testing"
+
+	cozyv1alpha1 "github.com/cozystack/cozystack/api/v1alpha1"
+	corev1 "k8s.io/api/core/v1"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/apimachinery/pkg/runtime"
+	"k8s.io/apimachinery/pkg/types"
+	cosiv1alpha1 "sigs.k8s.io/container-object-storage-interface-api/apis/objectstorage/v1alpha1"
+	"sigs.k8s.io/controller-runtime/pkg/client/fake"
+	"sigs.k8s.io/controller-runtime/pkg/reconcile"
+)
+
+func TestReconcile_OperationalStatusPersisted(t *testing.T) {
+	scheme := runtime.NewScheme()
+	_ = cozyv1alpha1.AddToScheme(scheme)
+	_ = corev1.AddToScheme(scheme)
+	_ = cosiv1alpha1.AddToScheme(scheme)
+
+	minReplicas := int32(2)
+	monitor := &cozyv1alpha1.WorkloadMonitor{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "test-monitor",
+			Namespace: "default",
+		},
+		Spec: cozyv1alpha1.WorkloadMonitorSpec{
+			Selector:    map[string]string{"app": "test"},
+			MinReplicas: &minReplicas,
+		},
+	}
+
+	// Create one pod that is ready — availableReplicas=1 < minReplicas=2, so Operational should be false
+	pod := &corev1.Pod{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "test-pod-1",
+			Namespace: "default",
+			Labels:    map[string]string{"app": "test"},
+		},
+		Status: corev1.PodStatus{
+			Conditions: []corev1.PodCondition{
+				{Type: corev1.PodReady, Status: corev1.ConditionTrue},
+			},
+		},
+	}
+
+	fakeClient := fake.NewClientBuilder().
+		WithScheme(scheme).
+		WithObjects(monitor, pod).
+		WithStatusSubresource(monitor).
+		Build()
+
+	reconciler := &WorkloadMonitorReconciler{Client: fakeClient, Scheme: scheme}
+	req := reconcile.Request{NamespacedName: types.NamespacedName{Name: "test-monitor", Namespace: "default"}}
+
+	_, err := reconciler.Reconcile(context.TODO(), req)
+	if err != nil {
+		t.Fatalf("Reconcile returned error: %v", err)
+	}
+
+	// Fetch the monitor back from fake client and check Operational is persisted
+	updated := &cozyv1alpha1.WorkloadMonitor{}
+	if err := fakeClient.Get(context.TODO(), req.NamespacedName, updated); err != nil {
+		t.Fatalf("Failed to get updated WorkloadMonitor: %v", err)
+	}
+
+	if updated.Status.Operational == nil {
+		t.Fatal("Expected Operational to be set, got nil")
+	}
+	if *updated.Status.Operational {
+		t.Error("Expected Operational=false (1 available < 2 minReplicas), got true")
+	}
+	if updated.Status.ObservedReplicas != 1 {
+		t.Errorf("Expected ObservedReplicas=1, got %d", updated.Status.ObservedReplicas)
+	}
+	if updated.Status.AvailableReplicas != 1 {
+		t.Errorf("Expected AvailableReplicas=1, got %d", updated.Status.AvailableReplicas)
+	}
+}
+
+func TestReconcile_OperationalTrue_WhenEnoughReplicas(t *testing.T) {
+	scheme := runtime.NewScheme()
+	_ = cozyv1alpha1.AddToScheme(scheme)
+	_ = corev1.AddToScheme(scheme)
+	_ = cosiv1alpha1.AddToScheme(scheme)
+
+	minReplicas := int32(1)
+	monitor := &cozyv1alpha1.WorkloadMonitor{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "test-monitor",
+			Namespace: "default",
+		},
+		Spec: cozyv1alpha1.WorkloadMonitorSpec{
+			Selector:    map[string]string{"app": "test"},
+			MinReplicas: &minReplicas,
+		},
+	}
+
+	pod := &corev1.Pod{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "test-pod-1",
+			Namespace: "default",
+			Labels:    map[string]string{"app": "test"},
+		},
+		Status: corev1.PodStatus{
+			Conditions: []corev1.PodCondition{
+				{Type: corev1.PodReady, Status: corev1.ConditionTrue},
+			},
+		},
+	}
+
+	fakeClient := fake.NewClientBuilder().
+		WithScheme(scheme).
+		WithObjects(monitor, pod).
+		WithStatusSubresource(monitor).
+		Build()
+
+	reconciler := &WorkloadMonitorReconciler{Client: fakeClient, Scheme: scheme}
+	req := reconcile.Request{NamespacedName: types.NamespacedName{Name: "test-monitor", Namespace: "default"}}
+
+	_, err := reconciler.Reconcile(context.TODO(), req)
+	if err != nil {
+		t.Fatalf("Reconcile returned error: %v", err)
+	}
+
+	updated := &cozyv1alpha1.WorkloadMonitor{}
+	if err := fakeClient.Get(context.TODO(), req.NamespacedName, updated); err != nil {
+		t.Fatalf("Failed to get updated WorkloadMonitor: %v", err)
+	}
+
+	if updated.Status.Operational == nil {
+		t.Fatal("Expected Operational to be set, got nil")
+	}
+	if !*updated.Status.Operational {
+		t.Error("Expected Operational=true (1 available >= 1 minReplicas), got false")
+	}
+}
+
+func TestGetMonitorLabels(t *testing.T) {
+	tests := []struct {
+		name     string
+		labels   map[string]string
+		expected map[string]string
+	}{
+		{
+			name:     "nil labels",
+			labels:   nil,
+			expected: map[string]string{},
+		},
+		{
+			name: "only workloads.cozystack.io/* labels are propagated",
+			labels: map[string]string{
+				"workloads.cozystack.io/resource-preset": "medium",
+				"app.kubernetes.io/name":                 "postgres",
+				"custom.example.com/team":                "platform",
+			},
+			expected: map[string]string{
+				"workloads.cozystack.io/resource-preset": "medium",
+			},
+		},
+		{
+			name: "monitor label is reserved and excluded",
+			labels: map[string]string{
+				"workloads.cozystack.io/resource-preset": "small",
+				"workloads.cozystack.io/monitor":         "should-be-dropped",
+			},
+			expected: map[string]string{
+				"workloads.cozystack.io/resource-preset": "small",
+			},
+		},
+		{
+			name: "multiple workloads.cozystack.io labels propagate",
+			labels: map[string]string{
+				"workloads.cozystack.io/resource-preset": "large",
+				"workloads.cozystack.io/tier":            "db",
+			},
+			expected: map[string]string{
+				"workloads.cozystack.io/resource-preset": "large",
+				"workloads.cozystack.io/tier":            "db",
+			},
+		},
+		{
+			name: "no matching labels returns empty map",
+			labels: map[string]string{
+				"app.kubernetes.io/name": "postgres",
+			},
+			expected: map[string]string{},
+		},
+	}
+
+	r := &WorkloadMonitorReconciler{}
+	for _, tc := range tests {
+		t.Run(tc.name, func(t *testing.T) {
+			monitor := &cozyv1alpha1.WorkloadMonitor{
+				ObjectMeta: metav1.ObjectMeta{Labels: tc.labels},
+			}
+			got := r.getMonitorLabels(monitor)
+			if len(got) != len(tc.expected) {
+				t.Fatalf("expected %d labels, got %d (%v)", len(tc.expected), len(got), got)
+			}
+			for k, v := range tc.expected {
+				if gv, ok := got[k]; !ok || gv != v {
+					t.Errorf("expected label %q=%q, got %q", k, v, gv)
+				}
+			}
+		})
+	}
+}
+
+func TestReconcile_MonitorLabelsPropagatedToPodWorkload(t *testing.T) {
+	scheme := runtime.NewScheme()
+	_ = cozyv1alpha1.AddToScheme(scheme)
+	_ = corev1.AddToScheme(scheme)
+
+	monitor := &cozyv1alpha1.WorkloadMonitor{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "test-monitor",
+			Namespace: "default",
+			Labels: map[string]string{
+				"workloads.cozystack.io/resource-preset": "medium",
+				"app.kubernetes.io/name":                 "ignored-not-propagated",
+			},
+		},
+		Spec: cozyv1alpha1.WorkloadMonitorSpec{
+			Selector: map[string]string{"app": "test"},
+			Kind:     "postgres",
+			Type:     "postgres",
+		},
+	}
+
+	pod := &corev1.Pod{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "test-pod-1",
+			Namespace: "default",
+			Labels: map[string]string{
+				"app":                    "test",
+				"app.kubernetes.io/name": "pod-wins-on-conflict",
+			},
+		},
+		Status: corev1.PodStatus{
+			Conditions: []corev1.PodCondition{
+				{Type: corev1.PodReady, Status: corev1.ConditionTrue},
+			},
+		},
+	}
+
+	fakeClient := fake.NewClientBuilder().
+		WithScheme(scheme).
+		WithObjects(monitor, pod).
+		WithStatusSubresource(monitor).
+		Build()
+
+	reconciler := &WorkloadMonitorReconciler{Client: fakeClient, Scheme: scheme}
+	req := reconcile.Request{NamespacedName: types.NamespacedName{Name: "test-monitor", Namespace: "default"}}
+	if _, err := reconciler.Reconcile(context.TODO(), req); err != nil {
+		t.Fatalf("Reconcile returned error: %v", err)
+	}
+
+	workload := &cozyv1alpha1.Workload{}
+	if err := fakeClient.Get(context.TODO(), types.NamespacedName{Name: "pod-test-pod-1", Namespace: "default"}, workload); err != nil {
+		t.Fatalf("Failed to get Workload: %v", err)
+	}
+
+	if got := workload.Labels["workloads.cozystack.io/resource-preset"]; got != "medium" {
+		t.Errorf("expected monitor label propagated, got %q", got)
+	}
+	// Non-workloads.cozystack.io monitor labels must not be copied
+	if _, ok := workload.Labels["app.kubernetes.io/name"]; !ok {
+		t.Error("expected pod label to be present on Workload")
+	}
+	// Source-object label takes precedence on conflict
+	if got := workload.Labels["app.kubernetes.io/name"]; got != "pod-wins-on-conflict" {
+		t.Errorf("expected pod label to win on conflict, got %q", got)
+	}
+	// Reserved monitor label is always set from the monitor name
+	if got := workload.Labels["workloads.cozystack.io/monitor"]; got != "test-monitor" {
+		t.Errorf("expected monitor-name label, got %q", got)
+	}
+}
+
+func TestReconcile_BackwardCompat_NoMonitorLabels(t *testing.T) {
+	scheme := runtime.NewScheme()
+	_ = cozyv1alpha1.AddToScheme(scheme)
+	_ = corev1.AddToScheme(scheme)
+
+	monitor := &cozyv1alpha1.WorkloadMonitor{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "test-monitor",
+			Namespace: "default",
+		},
+		Spec: cozyv1alpha1.WorkloadMonitorSpec{
+			Selector: map[string]string{"app": "test"},
+		},
+	}
+
+	pod := &corev1.Pod{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "test-pod-1",
+			Namespace: "default",
+			Labels:    map[string]string{"app": "test"},
+		},
+		Status: corev1.PodStatus{
+			Conditions: []corev1.PodCondition{
+				{Type: corev1.PodReady, Status: corev1.ConditionTrue},
+			},
+		},
+	}
+
+	fakeClient := fake.NewClientBuilder().
+		WithScheme(scheme).
+		WithObjects(monitor, pod).
+		WithStatusSubresource(monitor).
+		Build()
+
+	reconciler := &WorkloadMonitorReconciler{Client: fakeClient, Scheme: scheme}
+	req := reconcile.Request{NamespacedName: types.NamespacedName{Name: "test-monitor", Namespace: "default"}}
+	if _, err := reconciler.Reconcile(context.TODO(), req); err != nil {
+		t.Fatalf("Reconcile returned error: %v", err)
+	}
+
+	workload := &cozyv1alpha1.Workload{}
+	if err := fakeClient.Get(context.TODO(), types.NamespacedName{Name: "pod-test-pod-1", Namespace: "default"}, workload); err != nil {
+		t.Fatalf("Failed to get Workload: %v", err)
+	}
+	for k := range workload.Labels {
+		if strings.HasPrefix(k, "workloads.cozystack.io/") && k != "workloads.cozystack.io/monitor" {
+			t.Errorf("unexpected workload label present: %q", k)
+		}
+	}
+}
+
+func TestReconcile_OperationalTrue_WhenNoMinReplicas(t *testing.T) {
+	scheme := runtime.NewScheme()
+	_ = cozyv1alpha1.AddToScheme(scheme)
+	_ = corev1.AddToScheme(scheme)
+	_ = cosiv1alpha1.AddToScheme(scheme)
+
+	monitor := &cozyv1alpha1.WorkloadMonitor{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "test-monitor",
+			Namespace: "default",
+		},
+		Spec: cozyv1alpha1.WorkloadMonitorSpec{
+			Selector: map[string]string{"app": "test"},
+			// No MinReplicas — should default to operational=true
+		},
+	}
+
+	fakeClient := fake.NewClientBuilder().
+		WithScheme(scheme).
+		WithObjects(monitor).
+		WithStatusSubresource(monitor).
+		Build()
+
+	reconciler := &WorkloadMonitorReconciler{Client: fakeClient, Scheme: scheme}
+	req := reconcile.Request{NamespacedName: types.NamespacedName{Name: "test-monitor", Namespace: "default"}}
+
+	_, err := reconciler.Reconcile(context.TODO(), req)
+	if err != nil {
+		t.Fatalf("Reconcile returned error: %v", err)
+	}
+
+	updated := &cozyv1alpha1.WorkloadMonitor{}
+	if err := fakeClient.Get(context.TODO(), req.NamespacedName, updated); err != nil {
+		t.Fatalf("Failed to get updated WorkloadMonitor: %v", err)
+	}
+
+	if updated.Status.Operational == nil {
+		t.Fatal("Expected Operational to be set, got nil")
+	}
+	if !*updated.Status.Operational {
+		t.Error("Expected Operational=true (no MinReplicas constraint), got false")
+	}
+}
+
+func newTestScheme() *runtime.Scheme {
+	s := runtime.NewScheme()
+	_ = cozyv1alpha1.AddToScheme(s)
+	_ = corev1.AddToScheme(s)
+	_ = cosiv1alpha1.AddToScheme(s)
+	return s
+}
+
+func TestReconcileBucketClaimCreatesWorkload(t *testing.T) {
+	s := newTestScheme()
+
+	monitor := &cozyv1alpha1.WorkloadMonitor{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "my-bucket",
+			Namespace: "tenant-demo",
+		},
+		Spec: cozyv1alpha1.WorkloadMonitorSpec{
+			Kind: "bucket",
+			Type: "s3",
+			Selector: map[string]string{
+				"app.kubernetes.io/instance": "my-bucket",
+			},
+		},
+	}
+
+	bc := &cosiv1alpha1.BucketClaim{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "my-bucket",
+			Namespace: "tenant-demo",
+			Labels: map[string]string{
+				"app.kubernetes.io/instance": "my-bucket",
+			},
+		},
+		Spec: cosiv1alpha1.BucketClaimSpec{
+			BucketClassName: "seaweedfs",
+			Protocols:       []cosiv1alpha1.Protocol{cosiv1alpha1.ProtocolS3},
+		},
+		Status: cosiv1alpha1.BucketClaimStatus{
+			BucketReady: true,
+			BucketName:  "cosi-abc123",
+		},
+	}
+
+	fakeClient := fake.NewClientBuilder().
+		WithScheme(s).
+		WithObjects(monitor, bc).
+		WithStatusSubresource(monitor).
+		Build()
+
+	reconciler := &WorkloadMonitorReconciler{Client: fakeClient, Scheme: s}
+	req := reconcile.Request{NamespacedName: types.NamespacedName{
+		Name:      "my-bucket",
+		Namespace: "tenant-demo",
+	}}
+
+	_, err := reconciler.Reconcile(context.TODO(), req)
+	if err != nil {
+		t.Fatalf("Reconcile returned error: %v", err)
+	}
+
+	workload := &cozyv1alpha1.Workload{}
+	err = fakeClient.Get(context.TODO(), types.NamespacedName{
+		Name:      "bucket-my-bucket",
+		Namespace: "tenant-demo",
+	}, workload)
+	if err != nil {
+		t.Fatalf("expected Workload to be created, got error: %v", err)
+	}
+
+	if workload.Status.Kind != "bucket" {
+		t.Errorf("expected Kind=bucket, got %q", workload.Status.Kind)
+	}
+	if workload.Status.Type != "s3" {
+		t.Errorf("expected Type=s3, got %q", workload.Status.Type)
+	}
+	if !workload.Status.Operational {
+		t.Error("expected Operational=true for ready BucketClaim")
+	}
+}
+
+func TestReconcileBucketClaimNotReady(t *testing.T) {
+	s := newTestScheme()
+
+	monitor := &cozyv1alpha1.WorkloadMonitor{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "my-bucket",
+			Namespace: "tenant-demo",
+		},
+		Spec: cozyv1alpha1.WorkloadMonitorSpec{
+			Kind: "bucket",
+			Type: "s3",
+			Selector: map[string]string{
+				"app.kubernetes.io/instance": "my-bucket",
+			},
+		},
+	}
+
+	bc := &cosiv1alpha1.BucketClaim{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "my-bucket",
+			Namespace: "tenant-demo",
+			Labels: map[string]string{
+				"app.kubernetes.io/instance": "my-bucket",
+			},
+		},
+		Spec: cosiv1alpha1.BucketClaimSpec{
+			BucketClassName: "seaweedfs",
+			Protocols:       []cosiv1alpha1.Protocol{cosiv1alpha1.ProtocolS3},
+		},
+		Status: cosiv1alpha1.BucketClaimStatus{
+			BucketReady: false,
+		},
+	}
+
+	fakeClient := fake.NewClientBuilder().
+		WithScheme(s).
+		WithObjects(monitor, bc).
+		WithStatusSubresource(monitor).
+		Build()
+
+	reconciler := &WorkloadMonitorReconciler{Client: fakeClient, Scheme: s}
+	req := reconcile.Request{NamespacedName: types.NamespacedName{
+		Name:      "my-bucket",
+		Namespace: "tenant-demo",
+	}}
+
+	_, err := reconciler.Reconcile(context.TODO(), req)
+	if err != nil {
+		t.Fatalf("Reconcile returned error: %v", err)
+	}
+
+	workload := &cozyv1alpha1.Workload{}
+	err = fakeClient.Get(context.TODO(), types.NamespacedName{
+		Name:      "bucket-my-bucket",
+		Namespace: "tenant-demo",
+	}, workload)
+	if err != nil {
+		t.Fatalf("expected Workload to be created, got error: %v", err)
+	}
+
+	if workload.Status.Operational {
+		t.Error("expected Operational=false for not-ready BucketClaim")
+	}
+}
+
+func TestReconcile_MonitorLabelsPropagatedToBucketClaimWorkload(t *testing.T) {
+	s := newTestScheme()
+
+	monitor := &cozyv1alpha1.WorkloadMonitor{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "my-bucket",
+			Namespace: "tenant-demo",
+			Labels: map[string]string{
+				"workloads.cozystack.io/resource-preset": "medium",
+				"app.kubernetes.io/name":                 "ignored-not-propagated",
+			},
+		},
+		Spec: cozyv1alpha1.WorkloadMonitorSpec{
+			Kind: "bucket",
+			Type: "s3",
+			Selector: map[string]string{
+				"app.kubernetes.io/instance": "my-bucket",
+			},
+		},
+	}
+
+	bc := &cosiv1alpha1.BucketClaim{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "my-bucket",
+			Namespace: "tenant-demo",
+			Labels: map[string]string{
+				"app.kubernetes.io/instance": "my-bucket",
+				"app.kubernetes.io/name":     "bucket-wins-on-conflict",
+			},
+		},
+		Spec: cosiv1alpha1.BucketClaimSpec{
+			BucketClassName: "seaweedfs",
+			Protocols:       []cosiv1alpha1.Protocol{cosiv1alpha1.ProtocolS3},
+		},
+		Status: cosiv1alpha1.BucketClaimStatus{
+			BucketReady: true,
+			BucketName:  "cosi-abc123",
+		},
+	}
+
+	fakeClient := fake.NewClientBuilder().
+		WithScheme(s).
+		WithObjects(monitor, bc).
+		WithStatusSubresource(monitor).
+		Build()
+
+	reconciler := &WorkloadMonitorReconciler{Client: fakeClient, Scheme: s}
+	req := reconcile.Request{NamespacedName: types.NamespacedName{
+		Name:      "my-bucket",
+		Namespace: "tenant-demo",
+	}}
+	if _, err := reconciler.Reconcile(context.TODO(), req); err != nil {
+		t.Fatalf("Reconcile returned error: %v", err)
+	}
+
+	workload := &cozyv1alpha1.Workload{}
+	if err := fakeClient.Get(context.TODO(), types.NamespacedName{
+		Name:      "bucket-my-bucket",
+		Namespace: "tenant-demo",
+	}, workload); err != nil {
+		t.Fatalf("Failed to get Workload: %v", err)
+	}
+
+	if got := workload.Labels["workloads.cozystack.io/resource-preset"]; got != "medium" {
+		t.Errorf("expected monitor label propagated, got %q", got)
+	}
+	// Source-object label takes precedence on conflict
+	if got := workload.Labels["app.kubernetes.io/name"]; got != "bucket-wins-on-conflict" {
+		t.Errorf("expected bucket claim label to win on conflict, got %q", got)
+	}
+	// Reserved monitor label is always set from the monitor name
+	if got := workload.Labels["workloads.cozystack.io/monitor"]; got != "my-bucket" {
+		t.Errorf("expected monitor-name label, got %q", got)
+	}
+}
+
+func TestReconcileNoBucketClaimSkips(t *testing.T) {
+	s := newTestScheme()
+
+	monitor := &cozyv1alpha1.WorkloadMonitor{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "my-postgres",
+			Namespace: "tenant-demo",
+		},
+		Spec: cozyv1alpha1.WorkloadMonitorSpec{
+			Kind: "postgres",
+			Type: "postgres",
+			Selector: map[string]string{
+				"app.kubernetes.io/instance": "my-postgres",
+			},
+		},
+	}
+
+	fakeClient := fake.NewClientBuilder().
+		WithScheme(s).
+		WithObjects(monitor).
+		WithStatusSubresource(monitor).
+		Build()
+
+	reconciler := &WorkloadMonitorReconciler{Client: fakeClient, Scheme: s}
+	req := reconcile.Request{NamespacedName: types.NamespacedName{
+		Name:      "my-postgres",
+		Namespace: "tenant-demo",
+	}}
+
+	_, err := reconciler.Reconcile(context.TODO(), req)
+	if err != nil {
+		t.Fatalf("Reconcile returned error: %v", err)
+	}
+
+	workloadList := &cozyv1alpha1.WorkloadList{}
+	err = fakeClient.List(context.TODO(), workloadList)
+	if err != nil {
+		t.Fatalf("List returned error: %v", err)
+	}
+
+	for _, w := range workloadList.Items {
+		if w.Status.Kind == "bucket" {
+			t.Error("expected no bucket workloads to be created for postgres monitor")
+		}
+	}
+}
+
+func TestQueryAllBucketMetrics(t *testing.T) {
+	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		fmt.Fprint(w, `{"status":"success","data":{"resultType":"vector","result":[
+			{"metric":{"__name__":"SeaweedFS_s3_bucket_size_bytes","bucket":"bucket-aaa"},"value":[1713000000,"10485864"]},
+			{"metric":{"__name__":"SeaweedFS_s3_bucket_physical_size_bytes","bucket":"bucket-aaa"},"value":[1713000000,"20971728"]},
+			{"metric":{"__name__":"SeaweedFS_s3_bucket_size_bytes","bucket":"bucket-bbb"},"value":[1713000000,"0"]}
+		]}}`)
+	}))
+	defer srv.Close()
+
+	reconciler := &WorkloadMonitorReconciler{}
+	metrics := reconciler.queryAllBucketMetrics(context.TODO(), srv.URL, []string{"bucket-aaa", "bucket-bbb"})
+
+	bm, ok := metrics["bucket-aaa"]
+	if !ok {
+		t.Fatal("expected bucket-aaa in metrics")
+	}
+	if !bm.HasLogical || bm.LogicalSize != 10485864 {
+		t.Errorf("expected logical=10485864, got %d", bm.LogicalSize)
+	}
+	if !bm.HasPhysical || bm.PhysicalSize != 20971728 {
+		t.Errorf("expected physical=20971728, got %d", bm.PhysicalSize)
+	}
+
+	bm2, ok := metrics["bucket-bbb"]
+	if !ok {
+		t.Fatal("expected bucket-bbb in metrics")
+	}
+	if !bm2.HasLogical || bm2.LogicalSize != 0 {
+		t.Errorf("expected logical=0 for empty bucket, got %d", bm2.LogicalSize)
+	}
+	if bm2.HasPhysical {
+		t.Error("expected no physical size for bucket-bbb")
+	}
+}
+
+func TestQueryAllBucketMetricsEmpty(t *testing.T) {
+	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		fmt.Fprint(w, `{"status":"success","data":{"resultType":"vector","result":[]}}`)
+	}))
+	defer srv.Close()
+
+	reconciler := &WorkloadMonitorReconciler{}
+	metrics := reconciler.queryAllBucketMetrics(context.TODO(), srv.URL, []string{"bucket-aaa", "bucket-bbb"})
+	if len(metrics) != 0 {
+		t.Errorf("expected empty metrics, got %d", len(metrics))
+	}
+}
+
+func TestQueryAllBucketMetricsServerError(t *testing.T) {
+	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		w.WriteHeader(http.StatusInternalServerError)
+	}))
+	defer srv.Close()
+
+	reconciler := &WorkloadMonitorReconciler{}
+	metrics := reconciler.queryAllBucketMetrics(context.TODO(), srv.URL, []string{"bucket-aaa", "bucket-bbb"})
+	if len(metrics) != 0 {
+		t.Errorf("expected empty metrics on error, got %d", len(metrics))
+	}
+}
+
+func TestQueryAllBucketMetricsNoURL(t *testing.T) {
+	reconciler := &WorkloadMonitorReconciler{}
+	metrics := reconciler.queryAllBucketMetrics(context.TODO(), "", nil)
+	if len(metrics) != 0 {
+		t.Errorf("expected empty metrics when URL is empty, got %d", len(metrics))
+	}
+}
+
+func TestResolvePrometheusURL(t *testing.T) {
+	s := newTestScheme()
+
+	ns := &corev1.Namespace{
+		ObjectMeta: metav1.ObjectMeta{
+			Name: "tenant-demo",
+			Labels: map[string]string{
+				"namespace.cozystack.io/monitoring": "tenant-root",
+			},
+		},
+	}
+
+	fakeClient := fake.NewClientBuilder().
+		WithScheme(s).
+		WithObjects(ns).
+		Build()
+
+	reconciler := &WorkloadMonitorReconciler{Client: fakeClient, Scheme: s}
+	url := reconciler.resolvePrometheusURL(context.TODO(), "tenant-demo")
+
+	expected := "http://vmselect-shortterm.tenant-root.svc:8481/select/0/prometheus"
+	if url != expected {
+		t.Errorf("expected %q, got %q", expected, url)
+	}
+}
+
+func TestResolvePrometheusURLNoLabel(t *testing.T) {
+	s := newTestScheme()
+
+	ns := &corev1.Namespace{
+		ObjectMeta: metav1.ObjectMeta{
+			Name: "tenant-demo",
+		},
+	}
+
+	fakeClient := fake.NewClientBuilder().
+		WithScheme(s).
+		WithObjects(ns).
+		Build()
+
+	reconciler := &WorkloadMonitorReconciler{Client: fakeClient, Scheme: s}
+	url := reconciler.resolvePrometheusURL(context.TODO(), "tenant-demo")
+
+	if url != "" {
+		t.Errorf("expected empty URL when no monitoring label, got %q", url)
+	}
+}
+
+func TestReconcileBucketClaimRequeuesWhenBucketsExist(t *testing.T) {
+	s := newTestScheme()
+
+	ns := &corev1.Namespace{
+		ObjectMeta: metav1.ObjectMeta{
+			Name: "tenant-demo",
+		},
+	}
+
+	monitor := &cozyv1alpha1.WorkloadMonitor{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "my-bucket",
+			Namespace: "tenant-demo",
+		},
+		Spec: cozyv1alpha1.WorkloadMonitorSpec{
+			Kind: "bucket",
+			Type: "s3",
+			Selector: map[string]string{
+				"app.kubernetes.io/instance": "my-bucket",
+			},
+		},
+	}
+
+	bc := &cosiv1alpha1.BucketClaim{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "my-bucket",
+			Namespace: "tenant-demo",
+			Labels: map[string]string{
+				"app.kubernetes.io/instance": "my-bucket",
+			},
+		},
+		Spec: cosiv1alpha1.BucketClaimSpec{
+			BucketClassName: "seaweedfs",
+			Protocols:       []cosiv1alpha1.Protocol{cosiv1alpha1.ProtocolS3},
+		},
+		Status: cosiv1alpha1.BucketClaimStatus{
+			BucketReady: true,
+			BucketName:  "cosi-abc123",
+		},
+	}
+
+	fakeClient := fake.NewClientBuilder().
+		WithScheme(s).
+		WithObjects(ns, monitor, bc).
+		WithStatusSubresource(monitor).
+		Build()
+
+	reconciler := &WorkloadMonitorReconciler{Client: fakeClient, Scheme: s}
+	req := reconcile.Request{NamespacedName: types.NamespacedName{
+		Name:      "my-bucket",
+		Namespace: "tenant-demo",
+	}}
+
+	result, err := reconciler.Reconcile(context.TODO(), req)
+	if err != nil {
+		t.Fatalf("Reconcile returned error: %v", err)
+	}
+
+	if result.RequeueAfter == 0 {
+		t.Error("expected RequeueAfter > 0 when buckets exist")
+	}
+
+	workload := &cozyv1alpha1.Workload{}
+	err = fakeClient.Get(context.TODO(), types.NamespacedName{
+		Name:      "bucket-my-bucket",
+		Namespace: "tenant-demo",
+	}, workload)
+	if err != nil {
+		t.Fatalf("expected Workload to be created, got error: %v", err)
+	}
+
+	// Without monitoring label on namespace, no size metrics should be set
+	if _, ok := workload.Status.Resources["s3-storage-bytes"]; ok {
+		t.Error("expected no s3-storage-bytes when monitoring is not configured")
+	}
+	if len(workload.Status.Resources) != 0 {
+		t.Errorf("expected empty resources without monitoring, got %v", workload.Status.Resources)
+	}
+}
diff --git a/internal/crdinstall/manifests/cozystack.io_packagesources.yaml b/internal/crdinstall/manifests/cozystack.io_packagesources.yaml
index 0acfcdd2..1d0037df 100644
--- a/internal/crdinstall/manifests/cozystack.io_packagesources.yaml
+++ b/internal/crdinstall/manifests/cozystack.io_packagesources.yaml
@@ -118,6 +118,19 @@ spec:
                                   ReleaseName is the name of the HelmRelease resource that will be created
                                   If not specified, defaults to the component Name field
                                 type: string
+                              upgradeCRDs:
+                                description: |-
+                                  UpgradeCRDs controls how CRDs from the chart's crds/ directory are
+                                  handled on HelmRelease upgrades. Maps to HelmRelease.Spec.Upgrade.CRDs.
+                                  Empty string (default) preserves the helm-controller default (Skip).
+                                  Use "CreateReplace" for operators that evolve their CRD set between
+                                  versions. Warning: CreateReplace overwrites CRDs and may cause data
+                                  loss if upstream drops fields from a CRD with live objects.
+                                enum:
+                                - Skip
+                                - Create
+                                - CreateReplace
+                                type: string
                             type: object
                           libraries:
                             description: |-
diff --git a/internal/operator/package_reconciler.go b/internal/operator/package_reconciler.go
index 32e667e4..0e724e49 100644
--- a/internal/operator/package_reconciler.go
+++ b/internal/operator/package_reconciler.go
@@ -45,6 +45,16 @@ const (
 	SecretCozystackValues = "cozystack-values"
 )
 
+// parseCRDPolicy maps ComponentInstall.UpgradeCRDs to a helmv2.CRDsPolicy.
+// Empty / nil preserves the helm-controller default (Skip on upgrade);
+// the CRD enum marker restricts the string to Skip/Create/CreateReplace.
+func parseCRDPolicy(install *cozyv1alpha1.ComponentInstall) helmv2.CRDsPolicy {
+	if install == nil || install.UpgradeCRDs == "" {
+		return ""
+	}
+	return helmv2.CRDsPolicy(install.UpgradeCRDs)
+}
+
 // PackageReconciler reconciles Package resources
 type PackageReconciler struct {
 	client.Client
@@ -221,6 +231,7 @@ func (r *PackageReconciler) Reconcile(ctx context.Context, req ctrl.Request) (ct
 					Remediation: &helmv2.UpgradeRemediation{
 						Retries: -1,
 					},
+					CRDs: parseCRDPolicy(component.Install),
 				},
 			},
 		}
diff --git a/internal/operator/package_reconciler_test.go b/internal/operator/package_reconciler_test.go
new file mode 100644
index 00000000..f0ee2c19
--- /dev/null
+++ b/internal/operator/package_reconciler_test.go
@@ -0,0 +1,138 @@
+/*
+Copyright 2025 The Cozystack Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/
+
+package operator
+
+import (
+	"encoding/json"
+	"os"
+	"path/filepath"
+	"testing"
+
+	cozyv1alpha1 "github.com/cozystack/cozystack/api/v1alpha1"
+	helmv2 "github.com/fluxcd/helm-controller/api/v2"
+	apiextensionsv1 "k8s.io/apiextensions-apiserver/pkg/apis/apiextensions/v1"
+	"sigs.k8s.io/yaml"
+)
+
+func TestParseCRDPolicy(t *testing.T) {
+	tests := []struct {
+		name    string
+		install *cozyv1alpha1.ComponentInstall
+		want    helmv2.CRDsPolicy
+	}{
+		{
+			name:    "nil install leaves flux default",
+			install: nil,
+			want:    "",
+		},
+		{
+			name:    "empty upgradeCRDs leaves flux default",
+			install: &cozyv1alpha1.ComponentInstall{},
+			want:    "",
+		},
+		{
+			name:    "Skip is passed through",
+			install: &cozyv1alpha1.ComponentInstall{UpgradeCRDs: "Skip"},
+			want:    helmv2.Skip,
+		},
+		{
+			name:    "Create is passed through",
+			install: &cozyv1alpha1.ComponentInstall{UpgradeCRDs: "Create"},
+			want:    helmv2.Create,
+		},
+		{
+			name:    "CreateReplace is passed through",
+			install: &cozyv1alpha1.ComponentInstall{UpgradeCRDs: "CreateReplace"},
+			want:    helmv2.CreateReplace,
+		},
+	}
+
+	for _, tc := range tests {
+		t.Run(tc.name, func(t *testing.T) {
+			got := parseCRDPolicy(tc.install)
+			if got != tc.want {
+				t.Errorf("parseCRDPolicy() = %q, want %q", got, tc.want)
+			}
+		})
+	}
+}
+
+// TestPackageSourceCRDHasUpgradeCRDsEnum guards the generated CRD schema: the
+// invalid-value case from the spec is enforced at the API server via a
+// kubebuilder enum marker, not in the reconciler. If someone drops the marker
+// and forgets to regenerate, this test catches it.
+func TestPackageSourceCRDHasUpgradeCRDsEnum(t *testing.T) {
+	path := filepath.Join("..", "crdinstall", "manifests", "cozystack.io_packagesources.yaml")
+	data, err := os.ReadFile(path)
+	if err != nil {
+		t.Fatalf("read %s: %v", path, err)
+	}
+
+	var crd apiextensionsv1.CustomResourceDefinition
+	if err := yaml.Unmarshal(data, &crd); err != nil {
+		t.Fatalf("unmarshal CRD: %v", err)
+	}
+
+	var field *apiextensionsv1.JSONSchemaProps
+	for i := range crd.Spec.Versions {
+		v := &crd.Spec.Versions[i]
+		if v.Schema == nil || v.Schema.OpenAPIV3Schema == nil {
+			continue
+		}
+		spec, ok := v.Schema.OpenAPIV3Schema.Properties["spec"]
+		if !ok {
+			continue
+		}
+		variants, ok := spec.Properties["variants"]
+		if !ok || variants.Items == nil || variants.Items.Schema == nil {
+			continue
+		}
+		components, ok := variants.Items.Schema.Properties["components"]
+		if !ok || components.Items == nil || components.Items.Schema == nil {
+			continue
+		}
+		install, ok := components.Items.Schema.Properties["install"]
+		if !ok {
+			continue
+		}
+		f, ok := install.Properties["upgradeCRDs"]
+		if !ok {
+			continue
+		}
+		field = &f
+		break
+	}
+
+	if field == nil {
+		t.Fatal("upgradeCRDs field not found in PackageSource CRD schema")
+	}
+
+	got := map[string]bool{}
+	for _, e := range field.Enum {
+		var s string
+		if err := json.Unmarshal(e.Raw, &s); err != nil {
+			t.Fatalf("unmarshal enum value %q: %v", e.Raw, err)
+		}
+		got[s] = true
+	}
+
+	for _, want := range []string{"Skip", "Create", "CreateReplace"} {
+		if !got[want] {
+			t.Errorf("enum value %q missing from upgradeCRDs; got %v", want, got)
+		}
+	}
+}
diff --git a/packages/apps/bucket/templates/bucketclaim.yaml b/packages/apps/bucket/templates/bucketclaim.yaml
index 2f57fd70..0639916f 100644
--- a/packages/apps/bucket/templates/bucketclaim.yaml
+++ b/packages/apps/bucket/templates/bucketclaim.yaml
@@ -4,6 +4,8 @@ apiVersion: objectstorage.k8s.io/v1alpha1
 kind: BucketClaim
 metadata:
   name: {{ .Release.Name }}
+  labels:
+    app.kubernetes.io/instance: {{ .Release.Name }}
 spec:
   bucketClassName: {{ $seaweedfs }}{{- if $pool }}-{{ $pool }}{{- end }}{{- if .Values.locking }}-lock{{- end }}
   protocols:
diff --git a/packages/apps/bucket/templates/workloadmonitor.yaml b/packages/apps/bucket/templates/workloadmonitor.yaml
new file mode 100644
index 00000000..a23f3147
--- /dev/null
+++ b/packages/apps/bucket/templates/workloadmonitor.yaml
@@ -0,0 +1,13 @@
+---
+apiVersion: cozystack.io/v1alpha1
+kind: WorkloadMonitor
+metadata:
+  name: {{ $.Release.Name }}
+spec:
+  replicas: 0
+  minReplicas: 0
+  kind: bucket
+  type: s3
+  selector:
+    app.kubernetes.io/instance: {{ $.Release.Name }}
+  version: {{ $.Chart.Version }}
diff --git a/packages/apps/clickhouse/templates/workloadmonitor.yaml b/packages/apps/clickhouse/templates/workloadmonitor.yaml
index 9020ad41..62c951c2 100644
--- a/packages/apps/clickhouse/templates/workloadmonitor.yaml
+++ b/packages/apps/clickhouse/templates/workloadmonitor.yaml
@@ -3,6 +3,8 @@ apiVersion: cozystack.io/v1alpha1
 kind: WorkloadMonitor
 metadata:
   name: {{ $.Release.Name }}
+  labels:
+    workloads.cozystack.io/resource-preset: {{ .Values.resourcesPreset | quote }}
 spec:
   replicas: {{ .Values.replicas }}
   minReplicas: 1
@@ -17,6 +19,8 @@ apiVersion: cozystack.io/v1alpha1
 kind: WorkloadMonitor
 metadata:
   name: {{ $.Release.Name }}-keeper
+  labels:
+    workloads.cozystack.io/resource-preset: {{ .Values.clickhouseKeeper.resourcesPreset | quote }}
 spec:
   replicas: {{ .Values.clickhouseKeeper.replicas }}
   minReplicas: 1
diff --git a/packages/apps/foundationdb/templates/workloadmonitor.yaml b/packages/apps/foundationdb/templates/workloadmonitor.yaml
index 306797f3..2e2a3d19 100644
--- a/packages/apps/foundationdb/templates/workloadmonitor.yaml
+++ b/packages/apps/foundationdb/templates/workloadmonitor.yaml
@@ -8,6 +8,7 @@ metadata:
     app.kubernetes.io/name: foundationdb
     app.kubernetes.io/instance: {{ .Release.Name }}
     app.kubernetes.io/managed-by: {{ .Release.Service }}
+    workloads.cozystack.io/resource-preset: {{ .Values.resourcesPreset | quote }}
 spec:
   replicas: {{ .Values.cluster.processCounts.storage }}
   minReplicas: {{ include "foundationdb.minReplicas" . }}
diff --git a/packages/apps/harbor/templates/harbor.yaml b/packages/apps/harbor/templates/harbor.yaml
index ff8e88ad..bf780967 100644
--- a/packages/apps/harbor/templates/harbor.yaml
+++ b/packages/apps/harbor/templates/harbor.yaml
@@ -158,6 +158,8 @@ apiVersion: cozystack.io/v1alpha1
 kind: WorkloadMonitor
 metadata:
   name: {{ $.Release.Name }}-core
+  labels:
+    workloads.cozystack.io/resource-preset: {{ .Values.core.resourcesPreset | quote }}
 spec:
   replicas: 1
   minReplicas: 1
@@ -174,6 +176,8 @@ apiVersion: cozystack.io/v1alpha1
 kind: WorkloadMonitor
 metadata:
   name: {{ $.Release.Name }}-registry
+  labels:
+    workloads.cozystack.io/resource-preset: {{ .Values.registry.resourcesPreset | quote }}
 spec:
   replicas: 1
   minReplicas: 1
diff --git a/packages/apps/harbor/templates/ingress.yaml b/packages/apps/harbor/templates/ingress.yaml
index bb6ef07c..70933f5f 100644
--- a/packages/apps/harbor/templates/ingress.yaml
+++ b/packages/apps/harbor/templates/ingress.yaml
@@ -15,7 +15,7 @@ metadata:
     nginx.ingress.kubernetes.io/ssl-redirect: "true"
     nginx.ingress.kubernetes.io/backend-protocol: "HTTP"
     {{- if eq $solver "http01" }}
-    acme.cert-manager.io/http01-ingress-class: {{ $ingress }}
+    acme.cert-manager.io/http01-ingress-ingressclassname: {{ $ingress }}
     {{- end }}
     cert-manager.io/cluster-issuer: {{ $clusterIssuer }}
 spec:
diff --git a/packages/apps/http-cache/images/nginx-cache.tag b/packages/apps/http-cache/images/nginx-cache.tag
index 1da929ec..b08a374e 100644
--- a/packages/apps/http-cache/images/nginx-cache.tag
+++ b/packages/apps/http-cache/images/nginx-cache.tag
@@ -1 +1 @@
-ghcr.io/cozystack/cozystack/nginx-cache:0.0.0@sha256:2f987017ff95d3c782e16ef0d99928831f520daf154d08a2fa38c2d68363e036
+ghcr.io/cozystack/cozystack/nginx-cache:0.0.0@sha256:d397781152ab9123b11b8191d92eba7a0d2faa376aa2c15ddeb67842a9b59bab
diff --git a/packages/apps/http-cache/templates/workloadmonitor.yaml b/packages/apps/http-cache/templates/workloadmonitor.yaml
index 150d8bbe..a38a395a 100644
--- a/packages/apps/http-cache/templates/workloadmonitor.yaml
+++ b/packages/apps/http-cache/templates/workloadmonitor.yaml
@@ -3,6 +3,8 @@ apiVersion: cozystack.io/v1alpha1
 kind: WorkloadMonitor
 metadata:
   name: {{ $.Release.Name }}-haproxy
+  labels:
+    workloads.cozystack.io/resource-preset: {{ .Values.haproxy.resourcesPreset | quote }}
 spec:
   replicas: {{ .Values.haproxy.replicas }}
   minReplicas: 1
@@ -16,6 +18,8 @@ apiVersion: cozystack.io/v1alpha1
 kind: WorkloadMonitor
 metadata:
   name: {{ $.Release.Name }}-nginx
+  labels:
+    workloads.cozystack.io/resource-preset: {{ .Values.nginx.resourcesPreset | quote }}
 spec:
   replicas: {{ .Values.nginx.replicas }}
   minReplicas: 1
diff --git a/packages/apps/kafka/templates/workloadmonitor.yaml b/packages/apps/kafka/templates/workloadmonitor.yaml
index 4b161b04..c31eb425 100644
--- a/packages/apps/kafka/templates/workloadmonitor.yaml
+++ b/packages/apps/kafka/templates/workloadmonitor.yaml
@@ -3,6 +3,8 @@ apiVersion: cozystack.io/v1alpha1
 kind: WorkloadMonitor
 metadata:
   name: {{ $.Release.Name }}
+  labels:
+    workloads.cozystack.io/resource-preset: {{ .Values.kafka.resourcesPreset | quote }}
 spec:
   replicas: {{ .Values.replicas }}
   minReplicas: 1
@@ -19,6 +21,8 @@ apiVersion: cozystack.io/v1alpha1
 kind: WorkloadMonitor
 metadata:
   name: {{ $.Release.Name }}-zookeeper
+  labels:
+    workloads.cozystack.io/resource-preset: {{ .Values.zookeeper.resourcesPreset | quote }}
 spec:
   replicas: {{ .Values.replicas }}
   minReplicas: 1
diff --git a/packages/apps/kubernetes/Makefile b/packages/apps/kubernetes/Makefile
index 01cf736d..4fea42ef 100644
--- a/packages/apps/kubernetes/Makefile
+++ b/packages/apps/kubernetes/Makefile
@@ -4,6 +4,9 @@ KUBERNETES_PKG_TAG = $(shell awk '$$1 == "version:" {print $$2}' Chart.yaml)
 include ../../../hack/common-envs.mk
 include ../../../hack/package.mk
 
+test:
+	helm unittest .
+
 generate:
 	cozyvalues-gen -m 'kubernetes' -v values.yaml -s values.schema.json -r README.md -g ../../../api/apps/v1alpha1/kubernetes/types.go
 	../../../hack/update-crd.sh
@@ -67,3 +70,4 @@ image-cluster-autoscaler:
 	echo "$(REGISTRY)/cluster-autoscaler:$(call settag,$(KUBERNETES_PKG_TAG))@$$(yq e '."containerimage.digest"' images/cluster-autoscaler.json -o json -r)" \
 		> images/cluster-autoscaler.tag
 	rm -f images/cluster-autoscaler.json
+
diff --git a/packages/apps/kubernetes/README.md b/packages/apps/kubernetes/README.md
index d62d3a39..bed1117a 100644
--- a/packages/apps/kubernetes/README.md
+++ b/packages/apps/kubernetes/README.md
@@ -128,6 +128,9 @@ See the reference for components utilized in this service:
 | `addons.gpuOperator`                          | NVIDIA GPU Operator.                                                        | `object`   | `{}`      |
 | `addons.gpuOperator.enabled`                  | Enable GPU Operator.                                                        | `bool`     | `false`   |
 | `addons.gpuOperator.valuesOverride`           | Custom Helm values overrides.                                               | `object`   | `{}`      |
+| `addons.hami`                                 | HAMi GPU virtualization middleware.                                         | `object`   | `{}`      |
+| `addons.hami.enabled`                         | Enable HAMi (requires GPU Operator).                                        | `bool`     | `false`   |
+| `addons.hami.valuesOverride`                  | Custom Helm values overrides.                                               | `object`   | `{}`      |
 | `addons.fluxcd`                               | FluxCD GitOps operator.                                                     | `object`   | `{}`      |
 | `addons.fluxcd.enabled`                       | Enable FluxCD.                                                              | `bool`     | `false`   |
 | `addons.fluxcd.valuesOverride`                | Custom Helm values overrides.                                               | `object`   | `{}`      |
@@ -145,31 +148,33 @@ See the reference for components utilized in this service:
 
 ### Kubernetes Control Plane Configuration
 
-| Name                                                | Description                                      | Type       | Value   |
-| --------------------------------------------------- | ------------------------------------------------ | ---------- | ------- |
-| `controlPlane`                                      | Kubernetes control-plane configuration.          | `object`   | `{}`    |
-| `controlPlane.replicas`                             | Number of control-plane replicas.                | `int`      | `2`     |
-| `controlPlane.apiServer`                            | API Server configuration.                        | `object`   | `{}`    |
-| `controlPlane.apiServer.resources`                  | CPU and memory resources for API Server.         | `object`   | `{}`    |
-| `controlPlane.apiServer.resources.cpu`              | CPU available.                                   | `quantity` | `""`    |
-| `controlPlane.apiServer.resources.memory`           | Memory (RAM) available.                          | `quantity` | `""`    |
-| `controlPlane.apiServer.resourcesPreset`            | Preset if `resources` omitted.                   | `string`   | `large` |
-| `controlPlane.controllerManager`                    | Controller Manager configuration.                | `object`   | `{}`    |
-| `controlPlane.controllerManager.resources`          | CPU and memory resources for Controller Manager. | `object`   | `{}`    |
-| `controlPlane.controllerManager.resources.cpu`      | CPU available.                                   | `quantity` | `""`    |
-| `controlPlane.controllerManager.resources.memory`   | Memory (RAM) available.                          | `quantity` | `""`    |
-| `controlPlane.controllerManager.resourcesPreset`    | Preset if `resources` omitted.                   | `string`   | `micro` |
-| `controlPlane.scheduler`                            | Scheduler configuration.                         | `object`   | `{}`    |
-| `controlPlane.scheduler.resources`                  | CPU and memory resources for Scheduler.          | `object`   | `{}`    |
-| `controlPlane.scheduler.resources.cpu`              | CPU available.                                   | `quantity` | `""`    |
-| `controlPlane.scheduler.resources.memory`           | Memory (RAM) available.                          | `quantity` | `""`    |
-| `controlPlane.scheduler.resourcesPreset`            | Preset if `resources` omitted.                   | `string`   | `micro` |
-| `controlPlane.konnectivity`                         | Konnectivity configuration.                      | `object`   | `{}`    |
-| `controlPlane.konnectivity.server`                  | Konnectivity Server configuration.               | `object`   | `{}`    |
-| `controlPlane.konnectivity.server.resources`        | CPU and memory resources for Konnectivity.       | `object`   | `{}`    |
-| `controlPlane.konnectivity.server.resources.cpu`    | CPU available.                                   | `quantity` | `""`    |
-| `controlPlane.konnectivity.server.resources.memory` | Memory (RAM) available.                          | `quantity` | `""`    |
-| `controlPlane.konnectivity.server.resourcesPreset`  | Preset if `resources` omitted.                   | `string`   | `micro` |
+| Name                                                | Description                                                                                   | Type       | Value   |
+| --------------------------------------------------- | --------------------------------------------------------------------------------------------- | ---------- | ------- |
+| `controlPlane`                                      | Kubernetes control-plane configuration.                                                       | `object`   | `{}`    |
+| `controlPlane.replicas`                             | Number of control-plane replicas.                                                             | `int`      | `2`     |
+| `controlPlane.apiServer`                            | API Server configuration.                                                                     | `object`   | `{}`    |
+| `controlPlane.apiServer.resources`                  | CPU and memory resources for API Server.                                                      | `object`   | `{}`    |
+| `controlPlane.apiServer.resources.cpu`              | CPU available.                                                                                | `quantity` | `""`    |
+| `controlPlane.apiServer.resources.memory`           | Memory (RAM) available.                                                                       | `quantity` | `""`    |
+| `controlPlane.apiServer.resourcesPreset`            | Preset if `resources` omitted.                                                                | `string`   | `large` |
+| `controlPlane.controllerManager`                    | Controller Manager configuration.                                                             | `object`   | `{}`    |
+| `controlPlane.controllerManager.resources`          | CPU and memory resources for Controller Manager.                                              | `object`   | `{}`    |
+| `controlPlane.controllerManager.resources.cpu`      | CPU available.                                                                                | `quantity` | `""`    |
+| `controlPlane.controllerManager.resources.memory`   | Memory (RAM) available.                                                                       | `quantity` | `""`    |
+| `controlPlane.controllerManager.resourcesPreset`    | Preset if `resources` omitted.                                                                | `string`   | `micro` |
+| `controlPlane.scheduler`                            | Scheduler configuration.                                                                      | `object`   | `{}`    |
+| `controlPlane.scheduler.resources`                  | CPU and memory resources for Scheduler.                                                       | `object`   | `{}`    |
+| `controlPlane.scheduler.resources.cpu`              | CPU available.                                                                                | `quantity` | `""`    |
+| `controlPlane.scheduler.resources.memory`           | Memory (RAM) available.                                                                       | `quantity` | `""`    |
+| `controlPlane.scheduler.resourcesPreset`            | Preset if `resources` omitted.                                                                | `string`   | `micro` |
+| `controlPlane.konnectivity`                         | Konnectivity configuration.                                                                   | `object`   | `{}`    |
+| `controlPlane.konnectivity.server`                  | Konnectivity Server configuration.                                                            | `object`   | `{}`    |
+| `controlPlane.konnectivity.server.resources`        | CPU and memory resources for Konnectivity.                                                    | `object`   | `{}`    |
+| `controlPlane.konnectivity.server.resources.cpu`    | CPU available.                                                                                | `quantity` | `""`    |
+| `controlPlane.konnectivity.server.resources.memory` | Memory (RAM) available.                                                                       | `quantity` | `""`    |
+| `controlPlane.konnectivity.server.resourcesPreset`  | Preset if `resources` omitted.                                                                | `string`   | `micro` |
+| `images`                                            | Optional image overrides for air-gapped or rate-limited registries.                           | `object`   | `{}`    |
+| `images.waitForKubeconfig`                          | Image used by the wait-for-kubeconfig init container. Empty falls back to images/busybox.tag. | `string`   | `""`    |
 
 
 ## Parameter examples and reference
diff --git a/packages/apps/kubernetes/images/busybox.tag b/packages/apps/kubernetes/images/busybox.tag
new file mode 100644
index 00000000..39de220a
--- /dev/null
+++ b/packages/apps/kubernetes/images/busybox.tag
@@ -0,0 +1 @@
+docker.io/library/busybox:1.37.0@sha256:1487d0af5f52b4ba31c7e465126ee2123fe3f2305d638e7827681e7cf6c83d5e
diff --git a/packages/apps/kubernetes/images/kubevirt-csi-driver.tag b/packages/apps/kubernetes/images/kubevirt-csi-driver.tag
index e623b778..00aa2caa 100644
--- a/packages/apps/kubernetes/images/kubevirt-csi-driver.tag
+++ b/packages/apps/kubernetes/images/kubevirt-csi-driver.tag
@@ -1 +1 @@
-ghcr.io/cozystack/cozystack/kubevirt-csi-driver:0.0.0@sha256:5b209248acba3d8cf0d999edd8a915915a2e565a1c8ac1f5dde0dffec20fa02e
+ghcr.io/cozystack/cozystack/kubevirt-csi-driver:0.0.0@sha256:72154a97054e16cdf3dea6129d962b8d7e86b55cf9386095e8ac2ce7c8b69172
diff --git a/packages/apps/kubernetes/images/ubuntu-container-disk-v1.30.tag b/packages/apps/kubernetes/images/ubuntu-container-disk-v1.30.tag
index 2d1e4d52..f49fffe6 100644
--- a/packages/apps/kubernetes/images/ubuntu-container-disk-v1.30.tag
+++ b/packages/apps/kubernetes/images/ubuntu-container-disk-v1.30.tag
@@ -1 +1 @@
-ghcr.io/cozystack/cozystack/ubuntu-container-disk:v1.30@sha256:51e272db23d64eb20bc44e38cdbb70199fd00a49a958b76c2a89d4a46fec9256
+ghcr.io/cozystack/cozystack/ubuntu-container-disk:v1.30@sha256:f460ed1fd5a721aed423e6c3b8be0105d7f693cd86ad992d9c15fd4b27e58cec
diff --git a/packages/apps/kubernetes/images/ubuntu-container-disk-v1.31.tag b/packages/apps/kubernetes/images/ubuntu-container-disk-v1.31.tag
index 9698df5c..b5bdc79a 100644
--- a/packages/apps/kubernetes/images/ubuntu-container-disk-v1.31.tag
+++ b/packages/apps/kubernetes/images/ubuntu-container-disk-v1.31.tag
@@ -1 +1 @@
-ghcr.io/cozystack/cozystack/ubuntu-container-disk:v1.31@sha256:31c3290305159fe484caf6c5780960cf071dc3939528295336da4abfa731a81b
+ghcr.io/cozystack/cozystack/ubuntu-container-disk:v1.31@sha256:aaa2dfa8ee53ae26295f44d2491330f51412457bbe4aa6ba256297cc0bd8a0da
diff --git a/packages/apps/kubernetes/images/ubuntu-container-disk-v1.32.tag b/packages/apps/kubernetes/images/ubuntu-container-disk-v1.32.tag
index 90545820..ca285644 100644
--- a/packages/apps/kubernetes/images/ubuntu-container-disk-v1.32.tag
+++ b/packages/apps/kubernetes/images/ubuntu-container-disk-v1.32.tag
@@ -1 +1 @@
-ghcr.io/cozystack/cozystack/ubuntu-container-disk:v1.32@sha256:9efb1fe18cffb045530b43ac4b39910f1802d3c16798d733165dcd1c67de9237
+ghcr.io/cozystack/cozystack/ubuntu-container-disk:v1.32@sha256:7acb4728ef6b9c0ff3344bf486bb00f2a083f2098452344a362242235013db01
diff --git a/packages/apps/kubernetes/images/ubuntu-container-disk-v1.33.tag b/packages/apps/kubernetes/images/ubuntu-container-disk-v1.33.tag
index 64345fab..bb58690a 100644
--- a/packages/apps/kubernetes/images/ubuntu-container-disk-v1.33.tag
+++ b/packages/apps/kubernetes/images/ubuntu-container-disk-v1.33.tag
@@ -1 +1 @@
-ghcr.io/cozystack/cozystack/ubuntu-container-disk:v1.33@sha256:a0afca1d6f720cfff764a1bc912bfe404b5e97ff731f6c126a7abdb51ed23f85
+ghcr.io/cozystack/cozystack/ubuntu-container-disk:v1.33@sha256:a2024339ab9edb980a96b43ad2744b7aa31afaef0983ddbc6a546068b4cfa87a
diff --git a/packages/apps/kubernetes/images/ubuntu-container-disk-v1.34.tag b/packages/apps/kubernetes/images/ubuntu-container-disk-v1.34.tag
index f6916755..36eed7a7 100644
--- a/packages/apps/kubernetes/images/ubuntu-container-disk-v1.34.tag
+++ b/packages/apps/kubernetes/images/ubuntu-container-disk-v1.34.tag
@@ -1 +1 @@
-ghcr.io/cozystack/cozystack/ubuntu-container-disk:v1.34@sha256:51dfe5ce3e2f8765fde9422ae7e9fba677016ec5e2be41559c5f11a56958b0e6
+ghcr.io/cozystack/cozystack/ubuntu-container-disk:v1.34@sha256:97bcf946a5687889c6a421d14e5adece11f0978151b2165dc87b28f77f7607db
diff --git a/packages/apps/kubernetes/images/ubuntu-container-disk-v1.35.tag b/packages/apps/kubernetes/images/ubuntu-container-disk-v1.35.tag
index b80ce821..faae83d2 100644
--- a/packages/apps/kubernetes/images/ubuntu-container-disk-v1.35.tag
+++ b/packages/apps/kubernetes/images/ubuntu-container-disk-v1.35.tag
@@ -1 +1 @@
-ghcr.io/cozystack/cozystack/ubuntu-container-disk:v1.35@sha256:c4ae418b2a2c139794cd9630c3b2c2171870cc7748ac200344d2c4ed4283cf0b
+ghcr.io/cozystack/cozystack/ubuntu-container-disk:v1.35@sha256:364c6d454891f1eb1a598fddb69cf328a14dbc451a8ac65812b038a7756da60a
diff --git a/packages/apps/kubernetes/templates/_helpers.tpl b/packages/apps/kubernetes/templates/_helpers.tpl
index 36c06b64..40a8ae7a 100644
--- a/packages/apps/kubernetes/templates/_helpers.tpl
+++ b/packages/apps/kubernetes/templates/_helpers.tpl
@@ -49,3 +49,52 @@ Selector labels
 app.kubernetes.io/name: {{ include "kubernetes.name" . }}
 app.kubernetes.io/instance: {{ .Release.Name }}
 {{- end }}
+
+{{/*
+wait-for-kubeconfig init container shared by the control-plane-side
+Deployments (cluster-autoscaler, kccm, kcsi-controller) that mount the
+*-admin-kubeconfig Secret provisioned asynchronously by Kamaji. The
+Secret volume is declared optional so kubelet does not FailedMount while
+Kamaji is still bootstrapping; this container polls the mounted path and
+exits only when super-admin.svc appears, which happens after kubelet's
+optional-Secret refresh cycle.
+
+The 10m deadline stays strictly below the 15m HelmRelease
+Install.Timeout set by cozystack-api for the Kubernetes kind (via the
+release.cozystack.io/helm-install-timeout annotation) so the
+CrashLoopBackOff surfaces before flux remediation fires and uninstalls
+the Cluster CR.
+
+The default image lives in images/busybox.tag and points directly at
+docker.io by digest (not mirrored to ghcr.io like the other .tag files
+here): the payload is a one-shot sh loop and the digest pin makes the
+pull immutable. Operators in air-gapped or rate-limited environments
+can override it via .Values.images.waitForKubeconfig (any registry
+reference kubelet can pull). When the value is empty the chart falls
+back to the bundled digest pin, preserving the prior default.
+
+Call site owns the surrounding volumes block; the kubeconfig volume
+must exist on the pod and mount at /etc/kubernetes/kubeconfig.
+*/}}
+{{- define "kubernetes.waitForAdminKubeconfig" -}}
+- name: wait-for-kubeconfig
+  image: "{{ default (.Files.Get "images/busybox.tag" | trim) .Values.images.waitForKubeconfig }}"
+  command:
+  - sh
+  - -c
+  - |
+    set -eu
+    deadline=$(( $(date +%s) + 600 ))
+    until [ -s /etc/kubernetes/kubeconfig/super-admin.svc ]; do
+      if [ "$(date +%s)" -ge "$deadline" ]; then
+        echo "admin kubeconfig was not provisioned within 10m; exiting so the pod goes CrashLoopBackOff and surfaces in dashboards" >&2
+        exit 1
+      fi
+      echo "waiting for admin kubeconfig (provisioned by Kamaji, visible after kubelet Secret refresh)..."
+      sleep 5
+    done
+  volumeMounts:
+  - name: kubeconfig
+    mountPath: /etc/kubernetes/kubeconfig
+    readOnly: true
+{{- end }}
diff --git a/packages/apps/kubernetes/templates/cluster-autoscaler/deployment.yaml b/packages/apps/kubernetes/templates/cluster-autoscaler/deployment.yaml
index a00e0155..298d86db 100644
--- a/packages/apps/kubernetes/templates/cluster-autoscaler/deployment.yaml
+++ b/packages/apps/kubernetes/templates/cluster-autoscaler/deployment.yaml
@@ -1,3 +1,14 @@
+{{- /*
+  Gate the control-plane-side workloads on the parent tenant having an etcd
+  DataStore. Without it no KamajiControlPlane is ever created, Kamaji never
+  provisions -admin-kubeconfig, and rendering these Deployments would cause
+  the wait-for-kubeconfig init to CrashLoopBackOff indefinitely, consuming
+  the parent HelmRelease install timeout and triggering the very uninstall
+  remediation cycle this chart is supposed to avoid. Rendering them only
+  when $etcd is set keeps the HelmRelease Ready while flux retries on its
+  interval and picks up the DataStore as soon as the Tenant chart finishes.
+*/}}
+{{- if .Values._namespace.etcd }}
 ---
 apiVersion: apps/v1
 kind: Deployment
@@ -23,6 +34,8 @@ spec:
         - key: node-role.kubernetes.io/control-plane
           operator: Exists
           effect: "NoSchedule"
+      initContainers:
+      {{- include "kubernetes.waitForAdminKubeconfig" $ | nindent 6 }}
       containers:
       - image: "{{ $.Files.Get "images/cluster-autoscaler.tag" | trim }}"
         name: cluster-autoscaler
@@ -56,6 +69,7 @@ spec:
         name: cloud-config
       - secret:
           secretName: {{ .Release.Name }}-admin-kubeconfig
+          optional: true
         name: kubeconfig
       serviceAccountName: {{ .Release.Name }}-cluster-autoscaler
       terminationGracePeriodSeconds: 10
@@ -105,3 +119,4 @@ rules:
     - list
     - update
     - watch
+{{- end }}
diff --git a/packages/apps/kubernetes/templates/cluster.yaml b/packages/apps/kubernetes/templates/cluster.yaml
index 10d6fd80..e1494e64 100644
--- a/packages/apps/kubernetes/templates/cluster.yaml
+++ b/packages/apps/kubernetes/templates/cluster.yaml
@@ -1,4 +1,15 @@
 {{- $etcd := .Values._namespace.etcd }}
+{{- /*
+  When $etcd is empty, the parent Tenant application has not populated
+  _namespace.etcd in cozystack-values yet - either the operator forgot to
+  set etcd: true on an ancestor Tenant, or the Tenant HelmRelease is still
+  reconciling. Either way, rendering a KamajiControlPlane with an empty
+  dataStoreName would be rejected by Kamaji's admission webhook and the
+  HelmRelease would fail to install, triggering remediation. Instead, emit
+  a single ConfigMap as a user-visible status beacon and skip the rest so
+  flux marks the HelmRelease Ready and retries its 5m reconcile loop until
+  the Tenant chart catches up.
+*/}}
 {{- $ingress := .Values._namespace.ingress }}
 {{- $host := .Values._namespace.host }}
 {{- $kubevirtmachinetemplateNames := list }}
@@ -84,6 +95,26 @@ spec:
           - name: default
             pod: {}
 {{- end }}
+{{- if not $etcd }}
+---
+apiVersion: v1
+kind: ConfigMap
+metadata:
+  name: {{ .Release.Name }}-awaiting-etcd
+  namespace: {{ .Release.Namespace }}
+  labels:
+    app.kubernetes.io/managed-by: {{ .Release.Service }}
+    app.kubernetes.io/name: kubernetes
+    app.kubernetes.io/instance: {{ .Release.Name }}
+data:
+  status: "awaiting-etcd"
+  message: |
+    No DataStore is available for this tenant Kubernetes cluster. The parent
+    Tenant application has not populated _namespace.etcd. Set spec.etcd: true
+    on an ancestor Tenant (usually tenant-root) and wait for its HelmRelease
+    to reconcile - this HelmRelease will pick up the DataStore on its next
+    5m reconcile loop and provision the cluster.
+{{- else }}
 ---
 apiVersion: cluster.x-k8s.io/v1beta1
 kind: Cluster
@@ -404,3 +435,4 @@ metadata:
 spec:
   {{- .spec | toYaml | nindent 2 }}
 {{- end }}
+{{- end }}
diff --git a/packages/apps/kubernetes/templates/csi/deploy.yaml b/packages/apps/kubernetes/templates/csi/deploy.yaml
index 938b6d67..de62104c 100644
--- a/packages/apps/kubernetes/templates/csi/deploy.yaml
+++ b/packages/apps/kubernetes/templates/csi/deploy.yaml
@@ -1,3 +1,4 @@
+{{- if .Values._namespace.etcd }}
 kind: Deployment
 apiVersion: apps/v1
 metadata:
@@ -24,6 +25,8 @@ spec:
         - key: node-role.kubernetes.io/control-plane
           operator: Exists
           effect: "NoSchedule"
+      initContainers:
+      {{- include "kubernetes.waitForAdminKubeconfig" $ | nindent 6 }}
       containers:
         - name: csi-driver
           imagePullPolicy: Always
@@ -234,4 +237,6 @@ spec:
           emptyDir: {}
         - secret:
             secretName: {{ .Release.Name }}-admin-kubeconfig
+            optional: true
           name: kubeconfig
+{{- end }}
diff --git a/packages/apps/kubernetes/templates/helmreleases/cert-manager-crds.yaml b/packages/apps/kubernetes/templates/helmreleases/cert-manager-crds.yaml
index be07a8b9..fd9dac7e 100644
--- a/packages/apps/kubernetes/templates/helmreleases/cert-manager-crds.yaml
+++ b/packages/apps/kubernetes/templates/helmreleases/cert-manager-crds.yaml
@@ -1,4 +1,4 @@
-{{- if .Values.addons.certManager.enabled }}
+{{- if and .Values.addons.certManager.enabled .Values._namespace.etcd }}
 apiVersion: helm.toolkit.fluxcd.io/v2
 kind: HelmRelease
 metadata:
diff --git a/packages/apps/kubernetes/templates/helmreleases/cert-manager.yaml b/packages/apps/kubernetes/templates/helmreleases/cert-manager.yaml
index 6857581a..700b666e 100644
--- a/packages/apps/kubernetes/templates/helmreleases/cert-manager.yaml
+++ b/packages/apps/kubernetes/templates/helmreleases/cert-manager.yaml
@@ -8,7 +8,7 @@ cert-manager:
 {{- end }}
 {{- end }}
 
-{{- if .Values.addons.certManager.enabled }}
+{{- if and .Values.addons.certManager.enabled .Values._namespace.etcd }}
 apiVersion: helm.toolkit.fluxcd.io/v2
 kind: HelmRelease
 metadata:
diff --git a/packages/apps/kubernetes/templates/helmreleases/cilium.yaml b/packages/apps/kubernetes/templates/helmreleases/cilium.yaml
index d032f5b6..d8c90cbf 100644
--- a/packages/apps/kubernetes/templates/helmreleases/cilium.yaml
+++ b/packages/apps/kubernetes/templates/helmreleases/cilium.yaml
@@ -14,6 +14,7 @@ cilium:
   {{- end }}
 {{- end }}
 
+{{- if .Values._namespace.etcd }}
 apiVersion: helm.toolkit.fluxcd.io/v2
 kind: HelmRelease
 metadata:
@@ -55,3 +56,4 @@ spec:
   - name: {{ .Release.Name }}-gateway-api-crds
     namespace: {{ .Release.Namespace }}
   {{- end }}
+{{- end }}
diff --git a/packages/apps/kubernetes/templates/helmreleases/coredns.yaml b/packages/apps/kubernetes/templates/helmreleases/coredns.yaml
index bdb6c682..0711a51d 100644
--- a/packages/apps/kubernetes/templates/helmreleases/coredns.yaml
+++ b/packages/apps/kubernetes/templates/helmreleases/coredns.yaml
@@ -4,6 +4,7 @@ coredns:
     clusterIP: "10.95.0.10"
 {{- end }}
 
+{{- if .Values._namespace.etcd }}
 apiVersion: helm.toolkit.fluxcd.io/v2
 kind: HelmRelease
 metadata:
@@ -42,3 +43,4 @@ spec:
   {{- end }}
   - name: {{ .Release.Name }}-cilium
     namespace: {{ .Release.Namespace }}
+{{- end }}
diff --git a/packages/apps/kubernetes/templates/helmreleases/csi.yaml b/packages/apps/kubernetes/templates/helmreleases/csi.yaml
index dd2c69a6..109d78e1 100644
--- a/packages/apps/kubernetes/templates/helmreleases/csi.yaml
+++ b/packages/apps/kubernetes/templates/helmreleases/csi.yaml
@@ -1,3 +1,4 @@
+{{- if .Values._namespace.etcd }}
 apiVersion: helm.toolkit.fluxcd.io/v2
 kind: HelmRelease
 metadata:
@@ -39,3 +40,4 @@ spec:
   - name: {{ .Release.Name }}
     namespace: {{ .Release.Namespace }}
   {{- end }}
+{{- end }}
diff --git a/packages/apps/kubernetes/templates/helmreleases/fluxcd.yaml b/packages/apps/kubernetes/templates/helmreleases/fluxcd.yaml
index 76499dfe..25fff01c 100644
--- a/packages/apps/kubernetes/templates/helmreleases/fluxcd.yaml
+++ b/packages/apps/kubernetes/templates/helmreleases/fluxcd.yaml
@@ -1,4 +1,4 @@
-{{- if .Values.addons.fluxcd.enabled }}
+{{- if and .Values.addons.fluxcd.enabled .Values._namespace.etcd }}
 apiVersion: helm.toolkit.fluxcd.io/v2
 kind: HelmRelease
 metadata:
diff --git a/packages/apps/kubernetes/templates/helmreleases/gateway-api-crds.yaml b/packages/apps/kubernetes/templates/helmreleases/gateway-api-crds.yaml
index 2bcc8d4d..b4172ed1 100644
--- a/packages/apps/kubernetes/templates/helmreleases/gateway-api-crds.yaml
+++ b/packages/apps/kubernetes/templates/helmreleases/gateway-api-crds.yaml
@@ -1,4 +1,4 @@
-{{- if $.Values.addons.gatewayAPI.enabled }}
+{{- if and $.Values.addons.gatewayAPI.enabled $.Values._namespace.etcd }}
 apiVersion: helm.toolkit.fluxcd.io/v2
 kind: HelmRelease
 metadata:
diff --git a/packages/apps/kubernetes/templates/helmreleases/gpu-operator.yaml b/packages/apps/kubernetes/templates/helmreleases/gpu-operator.yaml
index 5ef48912..655dc868 100644
--- a/packages/apps/kubernetes/templates/helmreleases/gpu-operator.yaml
+++ b/packages/apps/kubernetes/templates/helmreleases/gpu-operator.yaml
@@ -1,4 +1,12 @@
-{{- if .Values.addons.gpuOperator.enabled }}
+{{- define "cozystack.defaultGpuOperatorValues" -}}
+{{- if .Values.addons.hami.enabled }}
+gpu-operator:
+  devicePlugin:
+    enabled: false
+{{- end }}
+{{- end }}
+
+{{- if and .Values.addons.gpuOperator.enabled .Values._namespace.etcd }}
 apiVersion: helm.toolkit.fluxcd.io/v2
 kind: HelmRelease
 metadata:
@@ -29,9 +37,12 @@ spec:
     force: true
     remediation:
       retries: -1
-  {{- with .Values.addons.gpuOperator.valuesOverride }}
+  {{- $defaults := fromYaml (include "cozystack.defaultGpuOperatorValues" .) }}
+  {{- $overrides := deepCopy (default (dict) .Values.addons.gpuOperator.valuesOverride) }}
+  {{- $merged := mergeOverwrite (default (dict) $defaults) $overrides }}
+  {{- if $merged }}
   values:
-    {{- toYaml . | nindent 4 }}
+    {{- toYaml $merged | nindent 4 }}
   {{- end }}
 
   dependsOn:
diff --git a/packages/apps/kubernetes/templates/helmreleases/hami.yaml b/packages/apps/kubernetes/templates/helmreleases/hami.yaml
new file mode 100644
index 00000000..f1538c7c
--- /dev/null
+++ b/packages/apps/kubernetes/templates/helmreleases/hami.yaml
@@ -0,0 +1,49 @@
+{{- if and .Values.addons.hami.enabled .Values._namespace.etcd }}
+{{- if not .Values.addons.gpuOperator.enabled }}
+{{- fail "addons.hami requires addons.gpuOperator to be enabled" }}
+{{- end }}
+apiVersion: helm.toolkit.fluxcd.io/v2
+kind: HelmRelease
+metadata:
+  name: {{ .Release.Name }}-hami
+  labels:
+    cozystack.io/repository: system
+    cozystack.io/target-cluster-name: {{ .Release.Name }}
+    sharding.fluxcd.io/key: tenants
+spec:
+  releaseName: hami
+  chartRef:
+    kind: ExternalArtifact
+    name: cozystack-kubernetes-application-kubevirt-kubernetes-hami
+    namespace: cozy-system
+  kubeConfig:
+    secretRef:
+      name: {{ .Release.Name }}-admin-kubeconfig
+      key: super-admin.svc
+  targetNamespace: cozy-hami
+  storageNamespace: cozy-hami
+  interval: 5m
+  timeout: 10m
+  install:
+    createNamespace: true
+    remediation:
+      retries: -1
+  upgrade:
+    force: true
+    remediation:
+      retries: -1
+  {{- with .Values.addons.hami.valuesOverride }}
+  values:
+    {{- toYaml . | nindent 4 }}
+  {{- end }}
+
+  dependsOn:
+  {{- if lookup "helm.toolkit.fluxcd.io/v2" "HelmRelease" .Release.Namespace .Release.Name }}
+  - name: {{ .Release.Name }}
+    namespace: {{ .Release.Namespace }}
+  {{- end }}
+  - name: {{ .Release.Name }}-cilium
+    namespace: {{ .Release.Namespace }}
+  - name: {{ .Release.Name }}-gpu-operator
+    namespace: {{ .Release.Namespace }}
+{{- end }}
diff --git a/packages/apps/kubernetes/templates/helmreleases/ingress-nginx.yaml b/packages/apps/kubernetes/templates/helmreleases/ingress-nginx.yaml
index 5cafff90..6e2183d3 100644
--- a/packages/apps/kubernetes/templates/helmreleases/ingress-nginx.yaml
+++ b/packages/apps/kubernetes/templates/helmreleases/ingress-nginx.yaml
@@ -20,7 +20,7 @@ ingress-nginx:
     node-role.kubernetes.io/ingress-nginx: ""
 {{- end }}
 
-{{- if .Values.addons.ingressNginx.enabled }}
+{{- if and .Values.addons.ingressNginx.enabled .Values._namespace.etcd }}
 apiVersion: helm.toolkit.fluxcd.io/v2
 kind: HelmRelease
 metadata:
diff --git a/packages/apps/kubernetes/templates/helmreleases/metrics-server.yaml b/packages/apps/kubernetes/templates/helmreleases/metrics-server.yaml
index 3cc81a14..3e6f9660 100644
--- a/packages/apps/kubernetes/templates/helmreleases/metrics-server.yaml
+++ b/packages/apps/kubernetes/templates/helmreleases/metrics-server.yaml
@@ -1,3 +1,4 @@
+{{- if .Values._namespace.etcd }}
 apiVersion: helm.toolkit.fluxcd.io/v2
 kind: HelmRelease
 metadata:
@@ -36,3 +37,4 @@ spec:
     namespace: {{ .Release.Namespace }}
   - name: {{ .Release.Name }}-prometheus-operator-crds
     namespace: {{ .Release.Namespace }}
+{{- end }}
diff --git a/packages/apps/kubernetes/templates/helmreleases/monitoring-agents.yaml b/packages/apps/kubernetes/templates/helmreleases/monitoring-agents.yaml
index ea84dec0..a811f7dd 100644
--- a/packages/apps/kubernetes/templates/helmreleases/monitoring-agents.yaml
+++ b/packages/apps/kubernetes/templates/helmreleases/monitoring-agents.yaml
@@ -1,6 +1,6 @@
 {{- $targetTenant := .Values._namespace.monitoring }}
 {{- $clusterDomain := (index .Values._cluster "cluster-domain") | default "cozy.local" }}
-{{- if .Values.addons.monitoringAgents.enabled }}
+{{- if and .Values.addons.monitoringAgents.enabled .Values._namespace.etcd }}
 apiVersion: helm.toolkit.fluxcd.io/v2
 kind: HelmRelease
 metadata:
diff --git a/packages/apps/kubernetes/templates/helmreleases/prometheus-operator-crds.yaml b/packages/apps/kubernetes/templates/helmreleases/prometheus-operator-crds.yaml
index 600a7994..3038a058 100644
--- a/packages/apps/kubernetes/templates/helmreleases/prometheus-operator-crds.yaml
+++ b/packages/apps/kubernetes/templates/helmreleases/prometheus-operator-crds.yaml
@@ -1,3 +1,4 @@
+{{- if .Values._namespace.etcd }}
 apiVersion: helm.toolkit.fluxcd.io/v2
 kind: HelmRelease
 metadata:
@@ -31,3 +32,4 @@ spec:
   - name: {{ .Release.Name }}
     namespace: {{ .Release.Namespace }}
   {{- end }}
+{{- end }}
diff --git a/packages/apps/kubernetes/templates/helmreleases/velero.yaml b/packages/apps/kubernetes/templates/helmreleases/velero.yaml
index ad236d53..781b9c49 100644
--- a/packages/apps/kubernetes/templates/helmreleases/velero.yaml
+++ b/packages/apps/kubernetes/templates/helmreleases/velero.yaml
@@ -1,4 +1,4 @@
-{{- if .Values.addons.velero.enabled }}
+{{- if and .Values.addons.velero.enabled .Values._namespace.etcd }}
 apiVersion: helm.toolkit.fluxcd.io/v2
 kind: HelmRelease
 metadata:
diff --git a/packages/apps/kubernetes/templates/helmreleases/vertical-pod-autoscaler-crds.yaml b/packages/apps/kubernetes/templates/helmreleases/vertical-pod-autoscaler-crds.yaml
index a3b7a9b4..55a5faac 100644
--- a/packages/apps/kubernetes/templates/helmreleases/vertical-pod-autoscaler-crds.yaml
+++ b/packages/apps/kubernetes/templates/helmreleases/vertical-pod-autoscaler-crds.yaml
@@ -1,4 +1,4 @@
-{{- if .Values.addons.monitoringAgents.enabled }}
+{{- if and .Values.addons.monitoringAgents.enabled .Values._namespace.etcd }}
 apiVersion: helm.toolkit.fluxcd.io/v2
 kind: HelmRelease
 metadata:
diff --git a/packages/apps/kubernetes/templates/helmreleases/vertical-pod-autoscaler.yaml b/packages/apps/kubernetes/templates/helmreleases/vertical-pod-autoscaler.yaml
index 178df3e3..74fb5a39 100644
--- a/packages/apps/kubernetes/templates/helmreleases/vertical-pod-autoscaler.yaml
+++ b/packages/apps/kubernetes/templates/helmreleases/vertical-pod-autoscaler.yaml
@@ -24,7 +24,7 @@ vertical-pod-autoscaler:
         memory: 1600Mi
 {{- end }}
 
-{{- if .Values.addons.monitoringAgents.enabled }}
+{{- if and .Values.addons.monitoringAgents.enabled .Values._namespace.etcd }}
 apiVersion: helm.toolkit.fluxcd.io/v2
 kind: HelmRelease
 metadata:
diff --git a/packages/apps/kubernetes/templates/helmreleases/victoria-metrics-operator.yaml b/packages/apps/kubernetes/templates/helmreleases/victoria-metrics-operator.yaml
index 99744277..7302f8f5 100644
--- a/packages/apps/kubernetes/templates/helmreleases/victoria-metrics-operator.yaml
+++ b/packages/apps/kubernetes/templates/helmreleases/victoria-metrics-operator.yaml
@@ -1,4 +1,4 @@
-{{- if .Values.addons.monitoringAgents.enabled }}
+{{- if and .Values.addons.monitoringAgents.enabled .Values._namespace.etcd }}
 apiVersion: helm.toolkit.fluxcd.io/v2
 kind: HelmRelease
 metadata:
diff --git a/packages/apps/kubernetes/templates/helmreleases/volumesnapshot-crd.yaml b/packages/apps/kubernetes/templates/helmreleases/volumesnapshot-crd.yaml
index 025f01b7..d50fd93c 100644
--- a/packages/apps/kubernetes/templates/helmreleases/volumesnapshot-crd.yaml
+++ b/packages/apps/kubernetes/templates/helmreleases/volumesnapshot-crd.yaml
@@ -1,3 +1,4 @@
+{{- if .Values._namespace.etcd }}
 apiVersion: helm.toolkit.fluxcd.io/v2
 kind: HelmRelease
 metadata:
@@ -33,3 +34,4 @@ spec:
   - name: {{ .Release.Name }}
     namespace: {{ .Release.Namespace }}
   {{- end }}
+{{- end }}
diff --git a/packages/apps/kubernetes/templates/kccm/manager.yaml b/packages/apps/kubernetes/templates/kccm/manager.yaml
index 81426d4e..bd9e2798 100644
--- a/packages/apps/kubernetes/templates/kccm/manager.yaml
+++ b/packages/apps/kubernetes/templates/kccm/manager.yaml
@@ -1,3 +1,4 @@
+{{- if .Values._namespace.etcd }}
 apiVersion: apps/v1
 kind: Deployment
 metadata:
@@ -22,6 +23,8 @@ spec:
         - key: node-role.kubernetes.io/control-plane
           operator: Exists
           effect: "NoSchedule"
+      initContainers:
+      {{- include "kubernetes.waitForAdminKubeconfig" $ | nindent 6 }}
       containers:
       - name: kubevirt-cloud-controller-manager
         args:
@@ -55,5 +58,7 @@ spec:
         name: cloud-config
       - secret:
           secretName: {{ .Release.Name }}-admin-kubeconfig
+          optional: true
         name: kubeconfig
       serviceAccountName: {{ .Release.Name }}-kccm
+{{- end }}
diff --git a/packages/apps/kubernetes/tests/admin_kubeconfig_wait_test.yaml b/packages/apps/kubernetes/tests/admin_kubeconfig_wait_test.yaml
new file mode 100644
index 00000000..507e00af
--- /dev/null
+++ b/packages/apps/kubernetes/tests/admin_kubeconfig_wait_test.yaml
@@ -0,0 +1,155 @@
+suite: admin-kubeconfig wait guards
+
+release:
+  name: test
+  namespace: tenant-root
+
+values:
+  - values-ci.yaml
+
+tests:
+  - it: cluster-autoscaler mounts admin-kubeconfig as optional
+    template: templates/cluster-autoscaler/deployment.yaml
+    documentSelector:
+      path: kind
+      value: Deployment
+    asserts:
+      - equal:
+          path: spec.template.spec.volumes[?(@.name=="kubeconfig")].secret.secretName
+          value: test-admin-kubeconfig
+      - equal:
+          path: spec.template.spec.volumes[?(@.name=="kubeconfig")].secret.optional
+          value: true
+
+  - it: cluster-autoscaler waits for admin-kubeconfig via initContainer
+    template: templates/cluster-autoscaler/deployment.yaml
+    documentSelector:
+      path: kind
+      value: Deployment
+    asserts:
+      - equal:
+          path: spec.template.spec.initContainers[0].name
+          value: wait-for-kubeconfig
+      - contains:
+          path: spec.template.spec.initContainers[0].volumeMounts
+          content:
+            name: kubeconfig
+            mountPath: /etc/kubernetes/kubeconfig
+            readOnly: true
+
+  - it: kccm mounts admin-kubeconfig as optional
+    template: templates/kccm/manager.yaml
+    documentSelector:
+      path: kind
+      value: Deployment
+    asserts:
+      - equal:
+          path: spec.template.spec.volumes[?(@.name=="kubeconfig")].secret.secretName
+          value: test-admin-kubeconfig
+      - equal:
+          path: spec.template.spec.volumes[?(@.name=="kubeconfig")].secret.optional
+          value: true
+
+  - it: kccm waits for admin-kubeconfig via initContainer
+    template: templates/kccm/manager.yaml
+    documentSelector:
+      path: kind
+      value: Deployment
+    asserts:
+      - equal:
+          path: spec.template.spec.initContainers[0].name
+          value: wait-for-kubeconfig
+      - contains:
+          path: spec.template.spec.initContainers[0].volumeMounts
+          content:
+            name: kubeconfig
+            mountPath: /etc/kubernetes/kubeconfig
+            readOnly: true
+
+  - it: csi controller mounts admin-kubeconfig as optional
+    template: templates/csi/deploy.yaml
+    documentSelector:
+      path: kind
+      value: Deployment
+    asserts:
+      - equal:
+          path: spec.template.spec.volumes[?(@.name=="kubeconfig")].secret.secretName
+          value: test-admin-kubeconfig
+      - equal:
+          path: spec.template.spec.volumes[?(@.name=="kubeconfig")].secret.optional
+          value: true
+
+  - it: csi controller waits for admin-kubeconfig via initContainer
+    template: templates/csi/deploy.yaml
+    documentSelector:
+      path: kind
+      value: Deployment
+    asserts:
+      - equal:
+          path: spec.template.spec.initContainers[0].name
+          value: wait-for-kubeconfig
+      - contains:
+          path: spec.template.spec.initContainers[0].volumeMounts
+          content:
+            name: kubeconfig
+            mountPath: /etc/kubernetes/kubeconfig
+            readOnly: true
+
+  - it: wait-for-kubeconfig defaults to bundled busybox digest pin
+    template: templates/cluster-autoscaler/deployment.yaml
+    documentSelector:
+      path: kind
+      value: Deployment
+    asserts:
+      - matchRegex:
+          path: spec.template.spec.initContainers[0].image
+          pattern: '^docker\.io/library/busybox:[^@]+@sha256:[0-9a-f]{64}$'
+
+  - it: wait-for-kubeconfig honours images.waitForKubeconfig override
+    template: templates/cluster-autoscaler/deployment.yaml
+    documentSelector:
+      path: kind
+      value: Deployment
+    set:
+      images:
+        waitForKubeconfig: "registry.example.test/library/busybox:1.37.0@sha256:0000000000000000000000000000000000000000000000000000000000000000"
+    asserts:
+      - equal:
+          path: spec.template.spec.initContainers[0].image
+          value: "registry.example.test/library/busybox:1.37.0@sha256:0000000000000000000000000000000000000000000000000000000000000000"
+
+  - it: cluster.yaml renders and wires dataStoreName when tenant has etcd
+    template: templates/cluster.yaml
+    documentSelector:
+      path: kind
+      value: KamajiControlPlane
+    asserts:
+      - equal:
+          path: spec.dataStoreName
+          value: tenant-root
+
+  - it: cluster.yaml skips Cluster resources when tenant has no etcd DataStore
+    # Must NOT fail rendering - the parent Tenant chart populates
+    # _namespace.etcd asynchronously, so rendering failures here would cause
+    # flux install remediation on every cold bootstrap. Instead, emit only a
+    # ConfigMap status beacon so the HelmRelease reports Ready while flux
+    # retries on its interval until the DataStore appears.
+    template: templates/cluster.yaml
+    set:
+      _namespace:
+        etcd: ""
+        monitoring: ""
+        ingress: ""
+        seaweedfs: ""
+        host: ""
+    asserts:
+      - hasDocuments:
+          count: 1
+      - isKind:
+          of: ConfigMap
+      - equal:
+          path: metadata.name
+          value: test-awaiting-etcd
+      - equal:
+          path: data.status
+          value: awaiting-etcd
diff --git a/packages/apps/kubernetes/tests/gpu_operator_hami_test.yaml b/packages/apps/kubernetes/tests/gpu_operator_hami_test.yaml
new file mode 100644
index 00000000..44528470
--- /dev/null
+++ b/packages/apps/kubernetes/tests/gpu_operator_hami_test.yaml
@@ -0,0 +1,99 @@
+suite: GPU Operator HelmRelease HAMi integration tests
+templates:
+  - templates/helmreleases/gpu-operator.yaml
+values:
+  - values-ci.yaml
+tests:
+  - it: should disable devicePlugin when hami is enabled
+    set:
+      addons:
+        gpuOperator:
+          enabled: true
+          valuesOverride: {}
+        hami:
+          enabled: true
+          valuesOverride: {}
+    asserts:
+      - equal:
+          path: spec.values.gpu-operator.devicePlugin.enabled
+          value: false
+
+  - it: should not have values when hami is disabled and no overrides
+    set:
+      addons:
+        gpuOperator:
+          enabled: true
+          valuesOverride: {}
+        hami:
+          enabled: false
+          valuesOverride: {}
+    asserts:
+      - notExists:
+          path: spec.values
+
+  - it: should apply hami defaults when valuesOverride key is omitted
+    set:
+      addons:
+        gpuOperator:
+          enabled: true
+        hami:
+          enabled: true
+    asserts:
+      - equal:
+          path: spec.values.gpu-operator.devicePlugin.enabled
+          value: false
+
+  - it: should allow user overrides to merge with hami defaults
+    set:
+      addons:
+        gpuOperator:
+          enabled: true
+          valuesOverride:
+            gpu-operator:
+              driver:
+                enabled: false
+        hami:
+          enabled: true
+          valuesOverride: {}
+    asserts:
+      - equal:
+          path: spec.values.gpu-operator.devicePlugin.enabled
+          value: false
+      - equal:
+          path: spec.values.gpu-operator.driver.enabled
+          value: false
+
+  - it: should let user explicitly override devicePlugin.enabled to true with hami enabled
+    set:
+      addons:
+        gpuOperator:
+          enabled: true
+          valuesOverride:
+            gpu-operator:
+              devicePlugin:
+                enabled: true
+              driver:
+                enabled: false
+        hami:
+          enabled: true
+          valuesOverride: {}
+    asserts:
+      - equal:
+          path: spec.values.gpu-operator.devicePlugin.enabled
+          value: true
+      - equal:
+          path: spec.values.gpu-operator.driver.enabled
+          value: false
+
+  - it: should not render when gpuOperator is disabled
+    set:
+      addons:
+        gpuOperator:
+          enabled: false
+          valuesOverride: {}
+        hami:
+          enabled: false
+          valuesOverride: {}
+    asserts:
+      - hasDocuments:
+          count: 0
diff --git a/packages/apps/kubernetes/tests/hami_test.yaml b/packages/apps/kubernetes/tests/hami_test.yaml
new file mode 100644
index 00000000..27f7c75b
--- /dev/null
+++ b/packages/apps/kubernetes/tests/hami_test.yaml
@@ -0,0 +1,153 @@
+suite: HAMi HelmRelease tests
+templates:
+  - templates/helmreleases/hami.yaml
+values:
+  - values-ci.yaml
+tests:
+  - it: should not render when hami is disabled
+    set:
+      addons:
+        hami:
+          enabled: false
+          valuesOverride: {}
+        gpuOperator:
+          enabled: true
+          valuesOverride: {}
+    asserts:
+      - hasDocuments:
+          count: 0
+
+  - it: should render HelmRelease when hami is enabled
+    set:
+      addons:
+        hami:
+          enabled: true
+          valuesOverride: {}
+        gpuOperator:
+          enabled: true
+          valuesOverride: {}
+    asserts:
+      - hasDocuments:
+          count: 1
+      - isKind:
+          of: HelmRelease
+
+  - it: should fail when gpuOperator is not enabled
+    set:
+      addons:
+        hami:
+          enabled: true
+          valuesOverride: {}
+        gpuOperator:
+          enabled: false
+          valuesOverride: {}
+    asserts:
+      - failedTemplate:
+          errorMessage: "addons.hami requires addons.gpuOperator to be enabled"
+
+  - it: should have correct metadata labels
+    set:
+      addons:
+        hami:
+          enabled: true
+          valuesOverride: {}
+        gpuOperator:
+          enabled: true
+          valuesOverride: {}
+    asserts:
+      - equal:
+          path: metadata.labels["cozystack.io/repository"]
+          value: system
+      - equal:
+          path: metadata.labels["sharding.fluxcd.io/key"]
+          value: tenants
+
+  - it: should use ExternalArtifact chartRef
+    set:
+      addons:
+        hami:
+          enabled: true
+          valuesOverride: {}
+        gpuOperator:
+          enabled: true
+          valuesOverride: {}
+    asserts:
+      - equal:
+          path: spec.chartRef.kind
+          value: ExternalArtifact
+      - equal:
+          path: spec.chartRef.namespace
+          value: cozy-system
+
+  - it: should target cozy-hami namespace
+    set:
+      addons:
+        hami:
+          enabled: true
+          valuesOverride: {}
+        gpuOperator:
+          enabled: true
+          valuesOverride: {}
+    asserts:
+      - equal:
+          path: spec.targetNamespace
+          value: cozy-hami
+      - equal:
+          path: spec.storageNamespace
+          value: cozy-hami
+
+  - it: should depend on gpu-operator and cilium
+    release:
+      name: test
+      namespace: test-ns
+    set:
+      addons:
+        hami:
+          enabled: true
+          valuesOverride: {}
+        gpuOperator:
+          enabled: true
+          valuesOverride: {}
+    asserts:
+      - contains:
+          path: spec.dependsOn
+          content:
+            name: test-cilium
+            namespace: test-ns
+      - contains:
+          path: spec.dependsOn
+          content:
+            name: test-gpu-operator
+            namespace: test-ns
+
+  - it: should not render spec.values when valuesOverride is empty
+    set:
+      addons:
+        hami:
+          enabled: true
+          valuesOverride: {}
+        gpuOperator:
+          enabled: true
+          valuesOverride: {}
+    asserts:
+      - hasDocuments:
+          count: 1
+      - notExists:
+          path: spec.values
+
+  - it: should pass through valuesOverride
+    set:
+      addons:
+        hami:
+          enabled: true
+          valuesOverride:
+            hami:
+              devicePlugin:
+                deviceSplitCount: 5
+        gpuOperator:
+          enabled: true
+          valuesOverride: {}
+    asserts:
+      - equal:
+          path: spec.values.hami.devicePlugin.deviceSplitCount
+          value: 5
diff --git a/packages/apps/kubernetes/tests/values-ci-no-etcd.yaml b/packages/apps/kubernetes/tests/values-ci-no-etcd.yaml
new file mode 100644
index 00000000..c7c8196f
--- /dev/null
+++ b/packages/apps/kubernetes/tests/values-ci-no-etcd.yaml
@@ -0,0 +1,9 @@
+_namespace:
+  etcd: ""
+  monitoring: ""
+  ingress: ""
+  seaweedfs: ""
+  host: ""
+_cluster:
+  cluster-domain: cozy.local
+nodeGroups: null
diff --git a/packages/apps/kubernetes/tests/values-ci.yaml b/packages/apps/kubernetes/tests/values-ci.yaml
new file mode 100644
index 00000000..13365e8c
--- /dev/null
+++ b/packages/apps/kubernetes/tests/values-ci.yaml
@@ -0,0 +1,9 @@
+_namespace:
+  etcd: tenant-root
+  monitoring: ""
+  ingress: ""
+  seaweedfs: ""
+  host: ""
+_cluster:
+  cluster-domain: cozy.local
+nodeGroups: null
diff --git a/packages/apps/kubernetes/values.schema.json b/packages/apps/kubernetes/values.schema.json
index 1beeff9c..a43deaec 100644
--- a/packages/apps/kubernetes/values.schema.json
+++ b/packages/apps/kubernetes/values.schema.json
@@ -149,6 +149,7 @@
         "fluxcd",
         "gatewayAPI",
         "gpuOperator",
+        "hami",
         "ingressNginx",
         "monitoringAgents",
         "velero",
@@ -268,6 +269,28 @@
             }
           }
         },
+        "hami": {
+          "description": "HAMi GPU virtualization middleware.",
+          "type": "object",
+          "default": {},
+          "required": [
+            "enabled",
+            "valuesOverride"
+          ],
+          "properties": {
+            "enabled": {
+              "description": "Enable HAMi (requires GPU Operator).",
+              "type": "boolean",
+              "default": false
+            },
+            "valuesOverride": {
+              "description": "Custom Helm values overrides.",
+              "type": "object",
+              "default": {},
+              "x-kubernetes-preserve-unknown-fields": true
+            }
+          }
+        },
         "ingressNginx": {
           "description": "Ingress-NGINX controller.",
           "type": "object",
@@ -630,6 +653,18 @@
           }
         }
       }
+    },
+    "images": {
+      "description": "Optional image overrides for air-gapped or rate-limited registries.",
+      "type": "object",
+      "default": {},
+      "properties": {
+        "waitForKubeconfig": {
+          "description": "Image used by the wait-for-kubeconfig init container. Empty falls back to images/busybox.tag.",
+          "type": "string",
+          "default": ""
+        }
+      }
     }
   }
 }
diff --git a/packages/apps/kubernetes/values.yaml b/packages/apps/kubernetes/values.yaml
index a67b5d69..d609476c 100644
--- a/packages/apps/kubernetes/values.yaml
+++ b/packages/apps/kubernetes/values.yaml
@@ -94,6 +94,10 @@ host: ""
 ## @field {bool} enabled - Enable FluxCD.
 ## @field {object} valuesOverride - Custom Helm values overrides.
 
+## @typedef {struct} HAMiAddon - HAMi GPU virtualization middleware.
+## @field {bool} enabled - Enable HAMi (requires GPU Operator).
+## @field {object} valuesOverride - Custom Helm values overrides.
+
 ## @typedef {struct} MonitoringAgentsAddon - Monitoring agents (Fluent Bit, VMAgents).
 ## @field {bool} enabled - Enable monitoring agents.
 ## @field {object} valuesOverride - Custom Helm values overrides.
@@ -114,6 +118,7 @@ host: ""
 ## @field {GatewayAPIAddon} gatewayAPI - Gateway API addon.
 ## @field {IngressNginxAddon} ingressNginx - Ingress-NGINX controller.
 ## @field {GPUOperatorAddon} gpuOperator - NVIDIA GPU Operator.
+## @field {HAMiAddon} hami - HAMi GPU virtualization middleware.
 ## @field {FluxCDAddon} fluxcd - FluxCD GitOps operator.
 ## @field {MonitoringAgentsAddon} monitoringAgents - Monitoring agents.
 ## @field {VerticalPodAutoscalerAddon} verticalPodAutoscaler - Vertical Pod Autoscaler.
@@ -137,6 +142,9 @@ addons:
   gpuOperator:
     enabled: false
     valuesOverride: {}
+  hami:
+    enabled: false
+    valuesOverride: {}
   fluxcd:
     enabled: false
     valuesOverride: {}
@@ -197,3 +205,10 @@ controlPlane:
     server:
       resources: {}
       resourcesPreset: "micro"
+
+## @typedef {struct} Images - Optional image overrides for chart-internal helpers.
+## @field {string} [waitForKubeconfig] - Image used by the wait-for-kubeconfig init container. Empty falls back to images/busybox.tag.
+
+## @param {Images} images - Optional image overrides for air-gapped or rate-limited registries.
+images:
+  waitForKubeconfig: ""
diff --git a/packages/apps/mariadb/images/mariadb-backup.tag b/packages/apps/mariadb/images/mariadb-backup.tag
index 1e381661..6c830892 100644
--- a/packages/apps/mariadb/images/mariadb-backup.tag
+++ b/packages/apps/mariadb/images/mariadb-backup.tag
@@ -1 +1 @@
-ghcr.io/cozystack/cozystack/mariadb-backup:0.0.0@sha256:0ddbbec0568dcb9fbc317cd9cc654e826dbe88ba3f184fa9b6b58aacb93b4570
+ghcr.io/cozystack/cozystack/mariadb-backup:0.0.0@sha256:3841eb171416711977dea0cf8cd45d32344caac9727af760c37d5e1dd41ee4bb
diff --git a/packages/apps/mariadb/templates/workloadmonitor.yaml b/packages/apps/mariadb/templates/workloadmonitor.yaml
index b69139bf..36cb59f0 100644
--- a/packages/apps/mariadb/templates/workloadmonitor.yaml
+++ b/packages/apps/mariadb/templates/workloadmonitor.yaml
@@ -3,6 +3,8 @@ apiVersion: cozystack.io/v1alpha1
 kind: WorkloadMonitor
 metadata:
   name: {{ $.Release.Name }}
+  labels:
+    workloads.cozystack.io/resource-preset: {{ .Values.resourcesPreset | quote }}
 spec:
   replicas: {{ .Values.replicas }}
   minReplicas: 1
diff --git a/packages/apps/mongodb/templates/mongodb.yaml b/packages/apps/mongodb/templates/mongodb.yaml
index 67969758..9273e506 100644
--- a/packages/apps/mongodb/templates/mongodb.yaml
+++ b/packages/apps/mongodb/templates/mongodb.yaml
@@ -169,6 +169,8 @@ apiVersion: cozystack.io/v1alpha1
 kind: WorkloadMonitor
 metadata:
   name: {{ .Release.Name }}
+  labels:
+    workloads.cozystack.io/resource-preset: {{ .Values.resourcesPreset | quote }}
 spec:
   {{- if .Values.sharding }}
   {{- $totalReplicas := 0 }}
diff --git a/packages/apps/nats/templates/workloadmonitor.yaml b/packages/apps/nats/templates/workloadmonitor.yaml
index 43d64a46..cb20b4a6 100644
--- a/packages/apps/nats/templates/workloadmonitor.yaml
+++ b/packages/apps/nats/templates/workloadmonitor.yaml
@@ -3,6 +3,8 @@ apiVersion: cozystack.io/v1alpha1
 kind: WorkloadMonitor
 metadata:
   name: {{ $.Release.Name }}
+  labels:
+    workloads.cozystack.io/resource-preset: {{ .Values.resourcesPreset | quote }}
 spec:
   replicas: {{ .Values.replicas }}
   minReplicas: 1
diff --git a/packages/apps/openbao/templates/workloadmonitor.yaml b/packages/apps/openbao/templates/workloadmonitor.yaml
index 0a9acf76..56f9ec8e 100644
--- a/packages/apps/openbao/templates/workloadmonitor.yaml
+++ b/packages/apps/openbao/templates/workloadmonitor.yaml
@@ -3,6 +3,8 @@ apiVersion: cozystack.io/v1alpha1
 kind: WorkloadMonitor
 metadata:
   name: {{ $.Release.Name }}
+  labels:
+    workloads.cozystack.io/resource-preset: {{ .Values.resourcesPreset | quote }}
 spec:
   replicas: {{ .Values.replicas }}
   minReplicas: 1
diff --git a/packages/apps/opensearch/templates/opensearch.yaml b/packages/apps/opensearch/templates/opensearch.yaml
index fcf431ac..836a3d4b 100644
--- a/packages/apps/opensearch/templates/opensearch.yaml
+++ b/packages/apps/opensearch/templates/opensearch.yaml
@@ -93,6 +93,8 @@ apiVersion: cozystack.io/v1alpha1
 kind: WorkloadMonitor
 metadata:
   name: {{ .Release.Name }}
+  labels:
+    workloads.cozystack.io/resource-preset: {{ .Values.resourcesPreset | quote }}
 spec:
   replicas: {{ .Values.replicas }}
   minReplicas: 1
diff --git a/packages/apps/postgres/README.md b/packages/apps/postgres/README.md
index 5a550f7a..4cda7284 100644
--- a/packages/apps/postgres/README.md
+++ b/packages/apps/postgres/README.md
@@ -133,12 +133,13 @@ See:
 
 ### Bootstrap (recovery) parameters
 
-| Name                     | Description                                                         | Type     | Value   |
-| ------------------------ | ------------------------------------------------------------------- | -------- | ------- |
-| `bootstrap`              | Bootstrap configuration.                                            | `object` | `{}`    |
-| `bootstrap.enabled`      | Whether to restore from a backup.                                   | `bool`   | `false` |
-| `bootstrap.recoveryTime` | Timestamp (RFC3339) for point-in-time recovery; empty means latest. | `string` | `""`    |
-| `bootstrap.oldName`      | Previous cluster name before deletion.                              | `string` | `""`    |
+| Name                     | Description                                                                                                                                                                                                                  | Type     | Value   |
+| ------------------------ | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | -------- | ------- |
+| `bootstrap`              | Bootstrap configuration.                                                                                                                                                                                                     | `object` | `{}`    |
+| `bootstrap.enabled`      | Whether to restore from a backup.                                                                                                                                                                                            | `bool`   | `false` |
+| `bootstrap.recoveryTime` | Timestamp (RFC3339) for point-in-time recovery; empty means latest.                                                                                                                                                          | `string` | `""`    |
+| `bootstrap.oldName`      | Previous cluster name before deletion.                                                                                                                                                                                       | `string` | `""`    |
+| `bootstrap.serverName`   | Barman server name (S3 path prefix) used by the original cluster when writing backups. Set this only when the original cluster had an explicit barmanObjectStore.serverName that differed from its Kubernetes resource name. | `string` | `""`    |
 
 
 ## Parameter examples and reference
diff --git a/packages/apps/postgres/templates/db.yaml b/packages/apps/postgres/templates/db.yaml
index 7557c436..2cdaec1d 100644
--- a/packages/apps/postgres/templates/db.yaml
+++ b/packages/apps/postgres/templates/db.yaml
@@ -32,6 +32,9 @@ spec:
     - name: {{ .Values.bootstrap.oldName }}
       barmanObjectStore:
         destinationPath: {{ .Values.backup.destinationPath }}
+        {{- if .Values.bootstrap.serverName }}
+        serverName: {{ .Values.bootstrap.serverName }}
+        {{- end }}
         endpointURL: {{ .Values.backup.endpointURL }}
         s3Credentials:
           accessKeyId:
@@ -84,6 +87,8 @@ apiVersion: cozystack.io/v1alpha1
 kind: WorkloadMonitor
 metadata:
   name: {{ $.Release.Name }}
+  labels:
+    workloads.cozystack.io/resource-preset: {{ .Values.resourcesPreset | quote }}
 spec:
   replicas: {{ .Values.replicas }}
   minReplicas: 1
diff --git a/packages/apps/postgres/values.schema.json b/packages/apps/postgres/values.schema.json
index b2a4aeba..98e29822 100644
--- a/packages/apps/postgres/values.schema.json
+++ b/packages/apps/postgres/values.schema.json
@@ -254,6 +254,11 @@
           "description": "Timestamp (RFC3339) for point-in-time recovery; empty means latest.",
           "type": "string",
           "default": ""
+        },
+        "serverName": {
+          "description": "Barman server name (S3 path prefix) used by the original cluster when writing backups. Set this only when the original cluster had an explicit barmanObjectStore.serverName that differed from its Kubernetes resource name.",
+          "type": "string",
+          "default": ""
         }
       }
     }
diff --git a/packages/apps/postgres/values.yaml b/packages/apps/postgres/values.yaml
index b8f07f63..2ceaa9cf 100644
--- a/packages/apps/postgres/values.yaml
+++ b/packages/apps/postgres/values.yaml
@@ -154,6 +154,7 @@ backup:
 ## @field {bool} enabled - Whether to restore from a backup.
 ## @field {string} [recoveryTime] - Timestamp (RFC3339) for point-in-time recovery; empty means latest.
 ## @field {string} oldName - Previous cluster name before deletion.
+## @field {string} [serverName] - Barman server name (S3 path prefix) used by the original cluster when writing backups. Set this only when the original cluster had an explicit barmanObjectStore.serverName that differed from its Kubernetes resource name.
 
 ## @param {Bootstrap} bootstrap - Bootstrap configuration.
 bootstrap:
@@ -161,3 +162,4 @@ bootstrap:
   # example: 2020-11-26 15:22:00.00000+00
   recoveryTime: ""
   oldName: ""
+  serverName: ""
diff --git a/packages/apps/qdrant/templates/dashboard-resourcemap.yaml b/packages/apps/qdrant/templates/dashboard-resourcemap.yaml
index b0a85bcc..0986951f 100644
--- a/packages/apps/qdrant/templates/dashboard-resourcemap.yaml
+++ b/packages/apps/qdrant/templates/dashboard-resourcemap.yaml
@@ -41,6 +41,8 @@ apiVersion: cozystack.io/v1alpha1
 kind: WorkloadMonitor
 metadata:
   name: {{ .Release.Name }}
+  labels:
+    workloads.cozystack.io/resource-preset: {{ .Values.resourcesPreset | quote }}
 spec:
   minReplicas: 1
   replicas: {{ .Values.replicas }}
diff --git a/packages/apps/rabbitmq/templates/workloadmonitor.yaml b/packages/apps/rabbitmq/templates/workloadmonitor.yaml
index 0f7462c7..66941153 100644
--- a/packages/apps/rabbitmq/templates/workloadmonitor.yaml
+++ b/packages/apps/rabbitmq/templates/workloadmonitor.yaml
@@ -3,6 +3,8 @@ apiVersion: cozystack.io/v1alpha1
 kind: WorkloadMonitor
 metadata:
   name: {{ $.Release.Name }}
+  labels:
+    workloads.cozystack.io/resource-preset: {{ .Values.resourcesPreset | quote }}
 spec:
   replicas: {{ .Values.replicas }}
   minReplicas: 1
diff --git a/packages/apps/redis/templates/redisfailover.yaml b/packages/apps/redis/templates/redisfailover.yaml
index 936217a3..160e030d 100644
--- a/packages/apps/redis/templates/redisfailover.yaml
+++ b/packages/apps/redis/templates/redisfailover.yaml
@@ -75,6 +75,8 @@ kind: WorkloadMonitor
 metadata:
   name: {{ $.Release.Name }}-redis
   namespace: {{ $.Release.Namespace }}
+  labels:
+    workloads.cozystack.io/resource-preset: {{ .Values.resourcesPreset | quote }}
 spec:
   minReplicas: 1
   replicas: {{ .Values.replicas }}
@@ -90,6 +92,8 @@ kind: WorkloadMonitor
 metadata:
   name: {{ $.Release.Name }}-sentinel
   namespace: {{ $.Release.Namespace }}
+  labels:
+    workloads.cozystack.io/resource-preset: {{ .Values.resourcesPreset | quote }}
 spec:
   minReplicas: 2
   replicas: 3
diff --git a/packages/apps/tcp-balancer/templates/workloadmonitor.yaml b/packages/apps/tcp-balancer/templates/workloadmonitor.yaml
index 41dce2ab..3478826b 100644
--- a/packages/apps/tcp-balancer/templates/workloadmonitor.yaml
+++ b/packages/apps/tcp-balancer/templates/workloadmonitor.yaml
@@ -3,6 +3,8 @@ apiVersion: cozystack.io/v1alpha1
 kind: WorkloadMonitor
 metadata:
   name: {{ $.Release.Name }}
+  labels:
+    workloads.cozystack.io/resource-preset: {{ .Values.resourcesPreset | quote }}
 spec:
   replicas: {{ .Values.replicas }}
   minReplicas: 1
diff --git a/packages/apps/tenant/Makefile b/packages/apps/tenant/Makefile
index 3fe3810d..2f51d836 100644
--- a/packages/apps/tenant/Makefile
+++ b/packages/apps/tenant/Makefile
@@ -3,3 +3,6 @@ include ../../../hack/package.mk
 generate:
 	cozyvalues-gen -m 'tenant' -v values.yaml -s values.schema.json -r README.md -g ../../../api/apps/v1alpha1/tenant/types.go
 	../../../hack/update-crd.sh
+
+test:
+	helm unittest .
diff --git a/packages/apps/tenant/README.md b/packages/apps/tenant/README.md
index 987e2c99..96cf5c6e 100644
--- a/packages/apps/tenant/README.md
+++ b/packages/apps/tenant/README.md
@@ -6,15 +6,15 @@ Tenants can be created recursively and are subject to the following rules:
 
 ### Tenant naming
 
-Tenant names must follow DNS-1035 naming rules:
--   Must start with a lowercase letter (`a-z`)
--   Can only contain lowercase letters, numbers, and hyphens (`a-z`, `0-9`, `-`)
--   Must end with a letter or number (not a hyphen)
+Tenant names must be alphanumeric:
+
+-   Lowercase letters (`a-z`) and digits (`0-9`) only
+-   Must start with a lowercase letter
+-   Dashes (`-`) are **not allowed**, unlike with other services
 -   Maximum length depends on the cluster configuration (Helm release prefix and root domain)
 
-**Note:** Using dashes (`-`) in tenant names is **allowed but discouraged**, unlike with other services.
-This is to keep consistent naming in tenants, nested tenants, and services deployed in them.
-Names with dashes (e.g., `foo-bar`) may lead to ambiguous parsing of internal resource names like `tenant-foo-bar`.
+This restriction exists to keep consistent naming in tenants, nested tenants, and services deployed in them.
+A tenant cannot be named `foo-bar` because parsing internal resource names like `tenant-foo-bar` would be ambiguous.
 
 For example:
 
diff --git a/packages/apps/tenant/templates/cilium-lb-pool.yaml b/packages/apps/tenant/templates/cilium-lb-pool.yaml
new file mode 100644
index 00000000..63dbcaca
--- /dev/null
+++ b/packages/apps/tenant/templates/cilium-lb-pool.yaml
@@ -0,0 +1,28 @@
+{{- $exposeMode := (index .Values._cluster "expose-mode") | default "externalIPs" }}
+{{- $exposeIngress := (index .Values._cluster "expose-ingress") | default "tenant-root" }}
+{{- $exposeIPs := (index .Values._cluster "expose-external-ips") | default "" | nospace }}
+{{- $ipsList := list }}
+{{- range splitList "," $exposeIPs }}
+  {{- $ip := . | trim }}
+  {{- if $ip }}
+    {{- $ipsList = append $ipsList $ip }}
+  {{- end }}
+{{- end }}
+{{- $isPublishingIngressLB := and
+      (eq $exposeMode "loadBalancer")
+      (eq $exposeIngress .Release.Namespace)
+      .Values.ingress }}
+{{- if and $isPublishingIngressLB $ipsList }}
+apiVersion: cilium.io/v2
+kind: CiliumLoadBalancerIPPool
+metadata:
+  name: {{ trimPrefix "tenant-" .Release.Namespace }}-exposure
+spec:
+  blocks:
+  {{- range $ipsList }}
+  - cidr: {{ . }}{{ if not (contains "/" .) }}/{{ if contains ":" . }}128{{ else }}32{{ end }}{{ end }}
+  {{- end }}
+  serviceSelector:
+    matchLabels:
+      "io.kubernetes.service.namespace": {{ .Release.Namespace | quote }}
+{{- end }}
diff --git a/packages/apps/tenant/templates/networkpolicy.yaml b/packages/apps/tenant/templates/networkpolicy.yaml
index ce3e33ab..4bd3ba02 100644
--- a/packages/apps/tenant/templates/networkpolicy.yaml
+++ b/packages/apps/tenant/templates/networkpolicy.yaml
@@ -186,6 +186,23 @@ spec:
 ---
 apiVersion: cilium.io/v2
 kind: CiliumNetworkPolicy
+metadata:
+  # Allow the tenant's ingress-nginx to reach both the linstor-gui
+  # gatekeeper pods (for /oauth2/* and app traffic) AND the transient
+  # ACME http-01 solver pods that cert-manager spins up in the same
+  # namespace during certificate issuance/renewal. Matching the whole
+  # namespace mirrors allow-to-dashboard / allow-to-keycloak.
+  name: allow-to-linstor-gui
+  namespace: {{ include "tenant.name" . }}
+spec:
+  endpointSelector: {}
+  egress:
+  - toEndpoints:
+      - matchLabels:
+          "k8s:io.kubernetes.pod.namespace": cozy-linstor
+---
+apiVersion: cilium.io/v2
+kind: CiliumNetworkPolicy
 metadata:
   name: allow-to-keycloak
   namespace: {{ include "tenant.name" . }}
diff --git a/packages/apps/tenant/tests/exposure_test.yaml b/packages/apps/tenant/tests/exposure_test.yaml
new file mode 100644
index 00000000..e86fab5b
--- /dev/null
+++ b/packages/apps/tenant/tests/exposure_test.yaml
@@ -0,0 +1,141 @@
+suite: tenant CiliumLoadBalancerIPPool rendering for publishing.exposure=loadBalancer
+templates:
+  - templates/cilium-lb-pool.yaml
+
+release:
+  name: tenant-root
+  namespace: tenant-root
+
+tests:
+  - it: default exposure (externalIPs) renders no pool
+    set:
+      ingress: true
+      _cluster:
+        expose-ingress: tenant-root
+        expose-external-ips: "192.0.2.10,192.0.2.11"
+    asserts:
+      - hasDocuments:
+          count: 0
+
+  - it: loadBalancer mode in publishing tenant with ingress=true renders v2 pool with namespace-only selector
+    set:
+      ingress: true
+      _cluster:
+        expose-ingress: tenant-root
+        expose-external-ips: "192.0.2.10,192.0.2.11"
+        expose-mode: loadBalancer
+    asserts:
+      - hasDocuments:
+          count: 1
+      - equal:
+          path: apiVersion
+          value: cilium.io/v2
+      - equal:
+          path: kind
+          value: CiliumLoadBalancerIPPool
+      - equal:
+          path: metadata.name
+          value: root-exposure
+      - equal:
+          path: spec.blocks
+          value:
+            - cidr: 192.0.2.10/32
+            - cidr: 192.0.2.11/32
+      - equal:
+          path: spec.serviceSelector.matchLabels["io.kubernetes.service.namespace"]
+          value: tenant-root
+      - notExists:
+          path: spec.serviceSelector.matchLabels["app.kubernetes.io/name"]
+
+  - it: loadBalancer mode with IPv6 emits /128 CIDR
+    set:
+      ingress: true
+      _cluster:
+        expose-ingress: tenant-root
+        expose-external-ips: "2001:db8::1"
+        expose-mode: loadBalancer
+    asserts:
+      - equal:
+          path: spec.blocks
+          value:
+            - cidr: 2001:db8::1/128
+
+  - it: loadBalancer mode with mixed IPv4 and IPv6 emits correct CIDR per family
+    set:
+      ingress: true
+      _cluster:
+        expose-ingress: tenant-root
+        expose-external-ips: "192.0.2.10,2001:db8::1"
+        expose-mode: loadBalancer
+    asserts:
+      - equal:
+          path: spec.blocks
+          value:
+            - cidr: 192.0.2.10/32
+            - cidr: 2001:db8::1/128
+
+  - it: loadBalancer mode accepts pre-CIDR input without double-suffixing
+    set:
+      ingress: true
+      _cluster:
+        expose-ingress: tenant-root
+        expose-external-ips: "192.0.2.10/32,2001:db8::1/128"
+        expose-mode: loadBalancer
+    asserts:
+      - equal:
+          path: spec.blocks
+          value:
+            - cidr: 192.0.2.10/32
+            - cidr: 2001:db8::1/128
+
+  - it: loadBalancer mode filters out empty entries from externalIPs (trailing, leading, repeated commas)
+    set:
+      ingress: true
+      _cluster:
+        expose-ingress: tenant-root
+        expose-external-ips: "192.0.2.10,,192.0.2.11,"
+        expose-mode: loadBalancer
+    asserts:
+      - hasDocuments:
+          count: 1
+      - equal:
+          path: spec.blocks
+          value:
+            - cidr: 192.0.2.10/32
+            - cidr: 192.0.2.11/32
+
+  - it: loadBalancer mode with ingress=false in publishing tenant renders no pool
+    set:
+      ingress: false
+      _cluster:
+        expose-ingress: tenant-root
+        expose-external-ips: "192.0.2.10"
+        expose-mode: loadBalancer
+    asserts:
+      - hasDocuments:
+          count: 0
+
+  - it: loadBalancer mode in a non-publishing tenant renders no pool
+    set:
+      ingress: true
+      _cluster:
+        expose-ingress: tenant-root
+        expose-external-ips: "192.0.2.10"
+        expose-mode: loadBalancer
+    release:
+      name: tenant-u1
+      namespace: tenant-u1
+    asserts:
+      - hasDocuments:
+          count: 0
+
+  - it: loadBalancer mode in publishing tenant with empty externalIPs renders no pool
+    set:
+      ingress: true
+      _cluster:
+        expose-ingress: tenant-root
+        expose-external-ips: ""
+        expose-mode: loadBalancer
+    asserts:
+      - hasDocuments:
+          count: 0
diff --git a/packages/apps/vm-disk/README.md b/packages/apps/vm-disk/README.md
index b8a4cdac..3727d0f5 100644
--- a/packages/apps/vm-disk/README.md
+++ b/packages/apps/vm-disk/README.md
@@ -6,15 +6,17 @@ A Virtual Machine Disk
 
 ### Common parameters
 
-| Name                | Description                                                                                                              | Type       | Value        |
-| ------------------- | ------------------------------------------------------------------------------------------------------------------------ | ---------- | ------------ |
-| `source`            | The source image location used to create a disk.                                                                         | `object`   | `{}`         |
-| `source.image`      | Use image by name.                                                                                                       | `*object`  | `null`       |
-| `source.image.name` | Name of the image to use (uploaded as "golden image" or from the list: `ubuntu`, `fedora`, `cirros`, `alpine`, `talos`). | `string`   | `""`         |
-| `source.upload`     | Upload local image.                                                                                                      | `*object`  | `null`       |
-| `source.http`       | Download image from an HTTP source.                                                                                      | `*object`  | `null`       |
-| `source.http.url`   | URL to download the image.                                                                                               | `string`   | `""`         |
-| `optical`           | Defines if disk should be considered optical.                                                                            | `bool`     | `false`      |
-| `storage`           | The size of the disk allocated for the virtual machine.                                                                  | `quantity` | `5Gi`        |
-| `storageClass`      | StorageClass used to store the data.                                                                                     | `string`   | `replicated` |
+| Name                | Description                                             | Type       | Value        |
+| ------------------- | ------------------------------------------------------- | ---------- | ------------ |
+| `source`            | The source image location used to create a disk.        | `object`   | `{}`         |
+| `source.image`      | Use image by name from default collection.              | `*object`  | `null`       |
+| `source.image.name` | Name of the image to use.                               | `string`   | `""`         |
+| `source.upload`     | Upload local image.                                     | `*object`  | `null`       |
+| `source.http`       | Download image from an HTTP source.                     | `*object`  | `null`       |
+| `source.http.url`   | URL to download the image.                              | `string`   | `""`         |
+| `source.disk`       | Clone an existing vm-disk.                              | `*object`  | `null`       |
+| `source.disk.name`  | Name of the vm-disk to clone.                           | `string`   | `""`         |
+| `optical`           | Defines if disk should be considered optical.           | `bool`     | `false`      |
+| `storage`           | The size of the disk allocated for the virtual machine. | `quantity` | `5Gi`        |
+| `storageClass`      | StorageClass used to store the data.                    | `string`   | `replicated` |
 
diff --git a/packages/apps/vm-disk/templates/dv.yaml b/packages/apps/vm-disk/templates/dv.yaml
index 3d68e639..521a82b8 100644
--- a/packages/apps/vm-disk/templates/dv.yaml
+++ b/packages/apps/vm-disk/templates/dv.yaml
@@ -21,15 +21,18 @@ spec:
   {{- end }}
   source:
     {{- if hasKey .Values.source "image" }}
-    {{- $dv := lookup "cdi.kubevirt.io/v1beta1" "DataVolume" "cozy-public" (printf "vm-image-%s" .Values.source.image.name) }}
     pvc:
-      name: vm-image-{{ required "A valid .Values.source.image.name entry required!" .Values.source.image.name }}
+      name: vm-default-images-{{ required "A valid .Values.source.image.name entry required!" .Values.source.image.name }}
       namespace: cozy-public
     {{- else if hasKey .Values.source "http" }}
     http:
       url: {{ required "A valid .Values.source.http.url entry required!" .Values.source.http.url }}
     {{- else if hasKey .Values.source "upload" }}
     upload: {}
+    {{- else if hasKey .Values.source "disk" }}
+    pvc:
+      name: vm-disk-{{ required "A valid .Values.source.disk.name entry required!" .Values.source.disk.name }}
+      namespace: {{ $.Release.Namespace }}
     {{- end }}
   {{- else }}
   source:
diff --git a/packages/apps/vm-disk/values.schema.json b/packages/apps/vm-disk/values.schema.json
index aec36a47..d61fc359 100644
--- a/packages/apps/vm-disk/values.schema.json
+++ b/packages/apps/vm-disk/values.schema.json
@@ -7,6 +7,19 @@
       "type": "object",
       "default": {},
       "properties": {
+        "disk": {
+          "description": "Clone an existing vm-disk.",
+          "type": "object",
+          "required": [
+            "name"
+          ],
+          "properties": {
+            "name": {
+              "description": "Name of the vm-disk to clone.",
+              "type": "string"
+            }
+          }
+        },
         "http": {
           "description": "Download image from an HTTP source.",
           "type": "object",
@@ -21,14 +34,14 @@
           }
         },
         "image": {
-          "description": "Use image by name.",
+          "description": "Use image by name from default collection.",
           "type": "object",
           "required": [
             "name"
           ],
           "properties": {
             "name": {
-              "description": "Name of the image to use (uploaded as \"golden image\" or from the list: `ubuntu`, `fedora`, `cirros`, `alpine`, `talos`).",
+              "description": "Name of the image to use.",
               "type": "string"
             }
           }
diff --git a/packages/apps/vm-disk/values.yaml b/packages/apps/vm-disk/values.yaml
index 33bd174f..b9a20d68 100644
--- a/packages/apps/vm-disk/values.yaml
+++ b/packages/apps/vm-disk/values.yaml
@@ -3,17 +3,21 @@
 ##
 
 ## @typedef {struct} SourceImage - Use image by name.
-## @field {string} name - Name of the image to use (uploaded as "golden image" or from the list: `ubuntu`, `fedora`, `cirros`, `alpine`, `talos`).
+## @field {string} name - Name of the image to use.
 
 ## @typedef {struct} SourceUpload - Upload local image.
 
 ## @typedef {struct} SourceHTTP - Download image from an HTTP source.
 ## @field {string} url - URL to download the image.
 
+## @typedef {struct} SourceDisk - Clone an existing vm-disk.
+## @field {string} name - Name of the vm-disk to clone.
+
 ## @typedef {struct} Source - The source image location used to create a disk.
-## @field {*SourceImage} [image] - Use image by name.
+## @field {*SourceImage} [image] - Use image by name from default collection.
 ## @field {*SourceUpload} [upload] - Upload local image.
 ## @field {*SourceHTTP} [http] - Download image from an HTTP source.
+## @field {*SourceDisk} [disk] - Clone an existing vm-disk.
 
 ## @param {Source} source - The source image location used to create a disk.
 source: {}
diff --git a/packages/apps/vm-instance/README.md b/packages/apps/vm-instance/README.md
index a2b6603e..9d6a52b2 100644
--- a/packages/apps/vm-instance/README.md
+++ b/packages/apps/vm-instance/README.md
@@ -36,31 +36,32 @@ virtctl ssh <user>@<vm>
 
 ### Common parameters
 
-| Name                | Description                                                                                                                       | Type       | Value       |
-| ------------------- | --------------------------------------------------------------------------------------------------------------------------------- | ---------- | ----------- |
-| `external`          | Enable external access from outside the cluster.                                                                                  | `bool`     | `false`     |
-| `externalMethod`    | Method to pass through traffic to the VM.                                                                                         | `string`   | `PortList`  |
-| `externalPorts`     | Ports to forward from outside the cluster.                                                                                        | `[]int`    | `[22]`      |
-| `runStrategy`       | Requested running state of the VirtualMachineInstance                                                                             | `string`   | `Always`    |
-| `instanceType`      | Virtual Machine instance type.                                                                                                    | `string`   | `u1.medium` |
-| `instanceProfile`   | Virtual Machine preferences profile.                                                                                              | `string`   | `ubuntu`    |
-| `disks`             | List of disks to attach.                                                                                                          | `[]object` | `[]`        |
-| `disks[i].name`     | Disk name.                                                                                                                        | `string`   | `""`        |
-| `disks[i].bus`      | Disk bus type (e.g. "sata").                                                                                                      | `string`   | `""`        |
-| `networks`          | Networks to attach the VM to.                                                                                                     | `[]object` | `[]`        |
-| `networks[i].name`  | Network attachment name.                                                                                                          | `string`   | `""`        |
-| `subnets`           | Deprecated: use networks instead.                                                                                                 | `[]object` | `[]`        |
-| `subnets[i].name`   | Network attachment name.                                                                                                          | `string`   | `""`        |
-| `gpus`              | List of GPUs to attach (NVIDIA driver requires at least 4 GiB RAM).                                                               | `[]object` | `[]`        |
-| `gpus[i].name`      | The name of the GPU resource to attach.                                                                                           | `string`   | `""`        |
-| `cpuModel`          | Model specifies the CPU model inside the VMI. List of available models https://github.com/libvirt/libvirt/tree/master/src/cpu_map | `string`   | `""`        |
-| `resources`         | Resource configuration for the virtual machine.                                                                                   | `object`   | `{}`        |
-| `resources.cpu`     | Number of CPU cores allocated.                                                                                                    | `quantity` | `""`        |
-| `resources.memory`  | Amount of memory allocated.                                                                                                       | `quantity` | `""`        |
-| `resources.sockets` | Number of CPU sockets (vCPU topology).                                                                                            | `quantity` | `""`        |
-| `sshKeys`           | List of SSH public keys for authentication.                                                                                       | `[]string` | `[]`        |
-| `cloudInit`         | Cloud-init user data.                                                                                                             | `string`   | `""`        |
-| `cloudInitSeed`     | Seed string to generate SMBIOS UUID for the VM.                                                                                   | `string`   | `""`        |
+| Name                | Description                                                                                                                                                                                                   | Type       | Value       |
+| ------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ---------- | ----------- |
+| `external`          | Enable external access from outside the cluster.                                                                                                                                                              | `bool`     | `false`     |
+| `externalMethod`    | Method to pass through traffic to the VM.                                                                                                                                                                     | `string`   | `PortList`  |
+| `externalPorts`     | Ports to forward from outside the cluster.                                                                                                                                                                    | `[]int`    | `[22]`      |
+| `externalAllowICMP` | Whether to accept ICMP traffic to the VM in PortList mode (preserves ping and PMTU discovery). No effect in WholeIP mode. Default true so ping behaves as users expect even when port filtering is in effect. | `bool`     | `true`      |
+| `runStrategy`       | Requested running state of the VirtualMachineInstance                                                                                                                                                         | `string`   | `Always`    |
+| `instanceType`      | Virtual Machine instance type.                                                                                                                                                                                | `string`   | `u1.medium` |
+| `instanceProfile`   | Virtual Machine preferences profile.                                                                                                                                                                          | `string`   | `ubuntu`    |
+| `disks`             | List of disks to attach.                                                                                                                                                                                      | `[]object` | `[]`        |
+| `disks[i].name`     | Disk name.                                                                                                                                                                                                    | `string`   | `""`        |
+| `disks[i].bus`      | Disk bus type (e.g. "sata").                                                                                                                                                                                  | `string`   | `""`        |
+| `networks`          | Networks to attach the VM to.                                                                                                                                                                                 | `[]object` | `[]`        |
+| `networks[i].name`  | Network attachment name.                                                                                                                                                                                      | `string`   | `""`        |
+| `subnets`           | Deprecated: use networks instead.                                                                                                                                                                             | `[]object` | `[]`        |
+| `subnets[i].name`   | Network attachment name.                                                                                                                                                                                      | `string`   | `""`        |
+| `gpus`              | List of GPUs to attach (NVIDIA driver requires at least 4 GiB RAM).                                                                                                                                           | `[]object` | `[]`        |
+| `gpus[i].name`      | The name of the GPU resource to attach.                                                                                                                                                                       | `string`   | `""`        |
+| `cpuModel`          | Model specifies the CPU model inside the VMI. List of available models https://github.com/libvirt/libvirt/tree/master/src/cpu_map                                                                             | `string`   | `""`        |
+| `resources`         | Resource configuration for the virtual machine.                                                                                                                                                               | `object`   | `{}`        |
+| `resources.cpu`     | Number of CPU cores allocated.                                                                                                                                                                                | `quantity` | `""`        |
+| `resources.memory`  | Amount of memory allocated.                                                                                                                                                                                   | `quantity` | `""`        |
+| `resources.sockets` | Number of CPU sockets (vCPU topology).                                                                                                                                                                        | `quantity` | `""`        |
+| `sshKeys`           | List of SSH public keys for authentication.                                                                                                                                                                   | `[]string` | `[]`        |
+| `cloudInit`         | Cloud-init user data.                                                                                                                                                                                         | `string`   | `""`        |
+| `cloudInitSeed`     | Seed string to generate SMBIOS UUID for the VM.                                                                                                                                                               | `string`   | `""`        |
 
 
 ## U Series
diff --git a/packages/apps/vm-instance/templates/service.yaml b/packages/apps/vm-instance/templates/service.yaml
index b7e3a420..cfeffa81 100644
--- a/packages/apps/vm-instance/templates/service.yaml
+++ b/packages/apps/vm-instance/templates/service.yaml
@@ -9,7 +9,10 @@ metadata:
 {{- if .Values.external }}
     service.kubernetes.io/service-proxy-name: "cozy-proxy"
   annotations:
-    networking.cozystack.io/wholeIP: "true"
+    networking.cozystack.io/wholeIP: {{ ternary "true" "false" (eq .Values.externalMethod "WholeIP") | quote }}
+    {{- if eq .Values.externalMethod "PortList" }}
+    networking.cozystack.io/allowICMP: {{ ternary "true" "false" (ne .Values.externalAllowICMP false) | quote }}
+    {{- end }}
 {{- end }}
 spec:
   type: {{ ternary "LoadBalancer" "ClusterIP" .Values.external }}
diff --git a/packages/apps/vm-instance/values.schema.json b/packages/apps/vm-instance/values.schema.json
index 34f7f634..01bd30a9 100644
--- a/packages/apps/vm-instance/values.schema.json
+++ b/packages/apps/vm-instance/values.schema.json
@@ -26,6 +26,11 @@
         "type": "integer"
       }
     },
+    "externalAllowICMP": {
+      "description": "Whether to accept ICMP traffic to the VM in PortList mode (preserves ping and PMTU discovery). No effect in WholeIP mode. Default true so ping behaves as users expect even when port filtering is in effect.",
+      "type": "boolean",
+      "default": true
+    },
     "runStrategy": {
       "description": "Requested running state of the VirtualMachineInstance",
       "type": "string",
diff --git a/packages/apps/vm-instance/values.yaml b/packages/apps/vm-instance/values.yaml
index f07b75d3..92e399c2 100644
--- a/packages/apps/vm-instance/values.yaml
+++ b/packages/apps/vm-instance/values.yaml
@@ -31,6 +31,9 @@ externalMethod: PortList
 externalPorts:
 - 22
 
+## @param {bool} externalAllowICMP - Whether to accept ICMP traffic to the VM in PortList mode (preserves ping and PMTU discovery). No effect in WholeIP mode. Default true so ping behaves as users expect even when port filtering is in effect.
+externalAllowICMP: true
+
 ## @enum {string} RunStrategy - Requested running state of the VirtualMachineInstance
 ## @value Always - VMI should always be running
 ## @value Halted - VMI should never be running
diff --git a/packages/apps/vpn/templates/workloadmonitor.yaml b/packages/apps/vpn/templates/workloadmonitor.yaml
index a75f7940..7fc1ec7a 100644
--- a/packages/apps/vpn/templates/workloadmonitor.yaml
+++ b/packages/apps/vpn/templates/workloadmonitor.yaml
@@ -2,6 +2,8 @@ apiVersion: cozystack.io/v1alpha1
 kind: WorkloadMonitor
 metadata:
   name: {{ $.Release.Name }}
+  labels:
+    workloads.cozystack.io/resource-preset: {{ .Values.resourcesPreset | quote }}
 spec:
   replicas: {{ .Values.replicas }}
   minReplicas: 1
diff --git a/packages/core/installer/values.yaml b/packages/core/installer/values.yaml
index 98443211..eef691ea 100644
--- a/packages/core/installer/values.yaml
+++ b/packages/core/installer/values.yaml
@@ -1,9 +1,9 @@
 cozystackOperator:
   # Deployment variant: talos, generic, hosted
   variant: talos
-  image: ghcr.io/cozystack/cozystack/cozystack-operator:v1.2.1@sha256:c89352808022944c4791d63cf82cc95d78124ce799d355b60427f317087e8909
+  image: ghcr.io/cozystack/cozystack/cozystack-operator:v1.3.0@sha256:62574f12486bb40c901cf5ed484cca264405ce5810196d86555cbb27cce1ba48
   platformSourceUrl: 'oci://ghcr.io/cozystack/cozystack/cozystack-packages'
-  platformSourceRef: 'digest=sha256:d02f211f79d4912f04c21856de078b5b41bb3976d84504d37cf8d9108f7cbec6'
+  platformSourceRef: 'digest=sha256:a0b9ef938446b3132d3d22ad2262beb1027c48c9037b6c2346fdc2f19acd3036'
 # Generic variant configuration (only used when cozystackOperator.variant=generic)
 cozystack:
   # Kubernetes API server host (IP only, no protocol/port)
diff --git a/packages/core/platform/images/migrations/migrations/37 b/packages/core/platform/images/migrations/migrations/37
index b8ff3147..856da262 100755
--- a/packages/core/platform/images/migrations/migrations/37
+++ b/packages/core/platform/images/migrations/migrations/37
@@ -1,42 +1,77 @@
-#!/bin/sh
+#!/bin/bash
 # Migration 37 --> 38
-# Backfill spec.version on postgreses.apps.cozystack.io resources.
+# Pin PostgreSQL image to 17.7-standard-trixie for system databases.
 #
-# Before this migration PostgreSQL had a default version field set to v18.
-# This migration sets spec.version to "v17" for any postgres app resource that
-# does not already have it set, to ensure compatibility with monitoring
-# configurations that are hardcoded to PostgreSQL 17.
+# This migration updates the imageName for all CNPG clusters belonging to
+# system components (keycloak-db, grafana-db, alerta-db, seaweedfs-db, and
+# harbor's dynamically-named DB):
+# - If imageName is not set: pins to 17.7-standard-trixie
+# - If imageName has any PG 17 tag: forces 17.7-standard-trixie
+# - If imageName has a bare version tag (e.g. :18.1): appends -standard-trixie
+#
+# NOTE: This migration ONLY updates system CNPG Cluster resources (clusters.postgresql.cnpg.io),
+# NOT user-created Postgres applications.
 
 set -euo pipefail
 
-DEFAULT_VERSION="v17"
+TARGET_IMAGE="ghcr.io/cloudnative-pg/postgresql:17.7-standard-trixie"
 
-# Skip if the CRD does not exist (postgres was never installed)
-if ! kubectl api-resources --api-group=apps.cozystack.io -o name 2>/dev/null | grep -q '^postgreses\.'; then
-  echo "CRD postgreses.apps.cozystack.io not found, skipping migration"
-  kubectl create configmap -n cozy-system cozystack-version \
-    --from-literal=version=38 --dry-run=client -o yaml | kubectl apply -f-
-  exit 0
-fi
+# Static system database names
+STATIC_DB_NAMES="keycloak-db grafana-db alerta-db seaweedfs-db"
 
-POSTGRESES=$(kubectl get postgreses.apps.cozystack.io -A -o jsonpath='{range .items[*]}{.metadata.namespace}/{.metadata.name}{"\n"}{end}')
-for resource in $POSTGRESES; do
-  NS="${resource%%/*}"
-  APP_NAME="${resource##*/}"
+echo "=== Updating PostgreSQL images for system databases ==="
+echo "Target image: $TARGET_IMAGE"
 
-  # Skip if spec.version is already set
-  CURRENT_VER=$(kubectl get postgreses.apps.cozystack.io -n "$NS" "$APP_NAME" \
-    -o jsonpath='{.spec.version}')
-  if [ -n "$CURRENT_VER" ]; then
-    echo "SKIP $NS/$APP_NAME: spec.version already set to '$CURRENT_VER'"
+# Fetch all CNPG clusters with their name, namespace, imageName, and Helm release annotation in one call
+ALL_CLUSTERS=$(kubectl get clusters.postgresql.cnpg.io -A \
+  -o jsonpath='{range .items[*]}{.metadata.namespace}{"\t"}{.metadata.name}{"\t"}{.spec.imageName}{"\t"}{.metadata.annotations.meta\.helm\.sh/release-name}{"\n"}{end}' 2>/dev/null || true)
+
+while IFS=$'\t' read -r NAMESPACE CLUSTER_NAME CURRENT_IMAGE HELM_RELEASE; do
+  [ -z "$NAMESPACE" ] && continue
+
+  # Check if cluster name matches one of the static system databases
+  MATCH=false
+  for db_name in $STATIC_DB_NAMES; do
+    if [ "$CLUSTER_NAME" = "$db_name" ]; then
+      MATCH=true
+      break
+    fi
+  done
+
+  # For harbor: match clusters ending with -db whose Helm release ends with -system
+  if [ "$MATCH" = false ] && [[ "$CLUSTER_NAME" == *-db ]] && [[ "$HELM_RELEASE" == *-system ]]; then
+    MATCH=true
+  fi
+
+  if [ "$MATCH" = false ]; then
     continue
   fi
 
-  echo "Patching postgres/$APP_NAME in $NS: setting version=$DEFAULT_VERSION"
+  PATCH_IMAGE=""
 
-  kubectl patch postgreses.apps.cozystack.io -n "$NS" "$APP_NAME" --type=merge \
-    --patch "{\"spec\":{\"version\":\"${DEFAULT_VERSION}\"}}"
-done
+  if [ -z "$CURRENT_IMAGE" ]; then
+    # imageName not set — pin to target
+    PATCH_IMAGE="$TARGET_IMAGE"
+    echo "PATCH $NAMESPACE/$CLUSTER_NAME: imageName not set, setting to $PATCH_IMAGE"
+  elif [[ "$CURRENT_IMAGE" =~ :17\. ]]; then
+    # Any PG 17 image — force to the pinned 17.7-standard-trixie
+    PATCH_IMAGE="$TARGET_IMAGE"
+    echo "PATCH $NAMESPACE/$CLUSTER_NAME: PG 17 detected, $CURRENT_IMAGE -> $PATCH_IMAGE"
+  elif [[ "$CURRENT_IMAGE" =~ :[0-9]+\.[0-9]+$ ]]; then
+    # Bare version tag for other majors (e.g. :18.1) — append -standard-trixie suffix
+    PATCH_IMAGE="${CURRENT_IMAGE}-standard-trixie"
+    echo "PATCH $NAMESPACE/$CLUSTER_NAME: bare tag detected, $CURRENT_IMAGE -> $PATCH_IMAGE"
+  else
+    echo "SKIP $NAMESPACE/$CLUSTER_NAME: imageName already set to $CURRENT_IMAGE"
+    continue
+  fi
+
+  kubectl patch clusters.postgresql.cnpg.io -n "$NAMESPACE" "$CLUSTER_NAME" \
+    --type=merge \
+    --patch "{\"spec\":{\"imageName\":\"${PATCH_IMAGE}\"}}"
+done <<< "$ALL_CLUSTERS"
+
+echo "=== PostgreSQL image update completed ==="
 
 # Stamp version
 kubectl create configmap -n cozy-system cozystack-version \
diff --git a/packages/core/platform/images/migrations/migrations/38 b/packages/core/platform/images/migrations/migrations/38
new file mode 100755
index 00000000..59536cb1
--- /dev/null
+++ b/packages/core/platform/images/migrations/migrations/38
@@ -0,0 +1,52 @@
+#!/bin/sh
+# Migration 38 --> 39
+# Rename DataVolumes/PVCs in cozy-public from vm-image-<name> to vm-default-images-<name>.
+#
+# The vm-disk package previously looked up golden images with the prefix "vm-image-".
+# The new vm-default-images package uses the prefix "vm-default-images-".
+# This migration renames any existing DataVolumes so that live vm-disk resources
+# continue to resolve their source PVC after upgrade.
+
+set -euo pipefail
+
+NAMESPACE="cozy-public"
+
+# Skip if CDI DataVolume CRD is not installed
+if ! kubectl api-resources --api-group=cdi.kubevirt.io -o name 2>/dev/null | grep -q '^datavolumes\.'; then
+  echo "CDI DataVolume CRD not found, skipping rename"
+  kubectl create configmap -n cozy-system cozystack-version \
+    --from-literal=version=39 --dry-run=client -o yaml | kubectl apply -f-
+  exit 0
+fi
+
+# Find DataVolumes whose name starts with "vm-image-"
+DVNAMES=$(kubectl get datavolumes.cdi.kubevirt.io -n "$NAMESPACE" \
+  -o jsonpath='{range .items[*]}{.metadata.name}{"\n"}{end}' 2>/dev/null \
+  | grep '^vm-image-' || true)
+
+if [ -z "$DVNAMES" ]; then
+  echo "No vm-image-* DataVolumes found in $NAMESPACE, nothing to rename"
+else
+  for DVNAME in $DVNAMES; do
+    NEWNAME="vm-default-images-${DVNAME#vm-image-}"
+    echo "Renaming DataVolume $DVNAME -> $NEWNAME in namespace $NAMESPACE"
+
+    # Export existing DV spec, swap name, and apply as new object
+    kubectl get datavolumes.cdi.kubevirt.io -n "$NAMESPACE" "$DVNAME" -o json \
+      | jq --arg new "$NEWNAME" '
+          del(.metadata.resourceVersion, .metadata.uid, .metadata.creationTimestamp,
+              .metadata.generation, .metadata.selfLink, .metadata.managedFields,
+              .status)
+          | .metadata.name = $new
+        ' \
+      | kubectl apply -f -
+
+    # Delete the old DataVolume
+    kubectl delete datavolumes.cdi.kubevirt.io -n "$NAMESPACE" "$DVNAME" --ignore-not-found=true
+    echo "Renamed $DVNAME -> $NEWNAME"
+  done
+fi
+
+# Stamp version
+kubectl create configmap -n cozy-system cozystack-version \
+  --from-literal=version=39 --dry-run=client -o yaml | kubectl apply -f-
diff --git a/packages/core/platform/sources/cozystack-scheduler.yaml b/packages/core/platform/sources/cozystack-scheduler.yaml
index af4c8338..52e85a4f 100644
--- a/packages/core/platform/sources/cozystack-scheduler.yaml
+++ b/packages/core/platform/sources/cozystack-scheduler.yaml
@@ -17,3 +17,14 @@ spec:
       install:
         namespace: kube-system
         releaseName: cozystack-scheduler
+  - name: linstor
+    dependsOn:
+    - cozystack.linstor-scheduler
+    components:
+    - name: cozystack-scheduler
+      path: system/cozystack-scheduler
+      valuesFiles:
+      - values-linstor.yaml
+      install:
+        namespace: kube-system
+        releaseName: cozystack-scheduler
diff --git a/packages/core/platform/sources/hami.yaml b/packages/core/platform/sources/hami.yaml
new file mode 100644
index 00000000..0184e405
--- /dev/null
+++ b/packages/core/platform/sources/hami.yaml
@@ -0,0 +1,24 @@
+---
+apiVersion: cozystack.io/v1alpha1
+kind: PackageSource
+metadata:
+  name: cozystack.hami
+spec:
+  sourceRef:
+    kind: OCIRepository
+    name: cozystack-packages
+    namespace: cozy-system
+    path: /
+  variants:
+  - name: default
+    dependsOn:
+    - cozystack.gpu-operator
+    components:
+    - name: hami
+      path: system/hami
+      valuesFiles:
+      - values.yaml
+      install:
+        privileged: true
+        namespace: cozy-hami
+        releaseName: hami
diff --git a/packages/core/platform/sources/kubernetes-application.yaml b/packages/core/platform/sources/kubernetes-application.yaml
index 9787cb19..088383ad 100644
--- a/packages/core/platform/sources/kubernetes-application.yaml
+++ b/packages/core/platform/sources/kubernetes-application.yaml
@@ -52,6 +52,8 @@ spec:
           path: system/cilium
         - name: kubernetes-gpu-operator
           path: system/gpu-operator
+        - name: kubernetes-hami
+          path: system/hami
         - name: kubernetes-vertical-pod-autoscaler
           path: system/vertical-pod-autoscaler
         - name: kubernetes-prometheus-operator-crds
diff --git a/packages/core/platform/sources/linstor-gui.yaml b/packages/core/platform/sources/linstor-gui.yaml
new file mode 100644
index 00000000..005e1fb7
--- /dev/null
+++ b/packages/core/platform/sources/linstor-gui.yaml
@@ -0,0 +1,21 @@
+---
+apiVersion: cozystack.io/v1alpha1
+kind: PackageSource
+metadata:
+  name: cozystack.linstor-gui
+spec:
+  sourceRef:
+    kind: OCIRepository
+    name: cozystack-packages
+    namespace: cozy-system
+    path: /
+  variants:
+  - name: default
+    dependsOn:
+    - cozystack.linstor
+    components:
+    - name: linstor-gui
+      path: system/linstor-gui
+      install:
+        namespace: cozy-linstor
+        releaseName: linstor-gui
diff --git a/packages/core/platform/sources/networking.yaml b/packages/core/platform/sources/networking.yaml
index 8e8485b3..bece76f8 100644
--- a/packages/core/platform/sources/networking.yaml
+++ b/packages/core/platform/sources/networking.yaml
@@ -66,6 +66,7 @@ spec:
       path: system/cilium
       valuesFiles:
       - values.yaml
+      - values-apparmor.yaml
       install:
         privileged: true
         namespace: cozy-cilium
@@ -117,6 +118,7 @@ spec:
       path: system/cilium
       valuesFiles:
       - values.yaml
+      - values-apparmor.yaml
       - values-kubeovn.yaml
       install:
         privileged: true
diff --git a/packages/core/platform/sources/vm-default-images.yaml b/packages/core/platform/sources/vm-default-images.yaml
new file mode 100644
index 00000000..16cd6a94
--- /dev/null
+++ b/packages/core/platform/sources/vm-default-images.yaml
@@ -0,0 +1,22 @@
+---
+apiVersion: cozystack.io/v1alpha1
+kind: PackageSource
+metadata:
+  name: cozystack.vm-default-images
+spec:
+  sourceRef:
+    kind: OCIRepository
+    name: cozystack-packages
+    namespace: cozy-system
+    path: /
+  variants:
+    - name: default
+      dependsOn:
+        - cozystack.networking
+        - cozystack.kubevirt-cdi
+      components:
+        - name: vm-default-images
+          path: system/vm-default-images
+          install:
+            namespace: cozy-system
+            releaseName: vm-default-images
diff --git a/packages/core/platform/templates/apps.yaml b/packages/core/platform/templates/apps.yaml
index 7ee0de52..6e1ae25a 100644
--- a/packages/core/platform/templates/apps.yaml
+++ b/packages/core/platform/templates/apps.yaml
@@ -30,6 +30,7 @@ stringData:
       expose-services: {{ .Values.publishing.exposedServices | join "," | quote }}
       expose-ingress: {{ .Values.publishing.ingressName | quote }}
       expose-external-ips: {{ .Values.publishing.externalIPs | join "," | quote }}
+      expose-mode: {{ .Values.publishing.exposure | default "externalIPs" | quote }}
       cluster-domain: {{ .Values.networking.clusterDomain | quote }}
       api-server-endpoint: {{ .Values.publishing.apiServerEndpoint | quote }}
       {{- with .Values.branding }}
diff --git a/packages/core/platform/templates/bundles/iaas.yaml b/packages/core/platform/templates/bundles/iaas.yaml
index bd002c16..66ec98b2 100644
--- a/packages/core/platform/templates/bundles/iaas.yaml
+++ b/packages/core/platform/templates/bundles/iaas.yaml
@@ -8,7 +8,9 @@
 {{- end -}}
 {{include "cozystack.platform.package" (list "cozystack.kubevirt" "default" $ $kubevirtComponents) }}
 {{include "cozystack.platform.package.default" (list "cozystack.kubevirt-cdi" $) }}
+{{include "cozystack.platform.package.optional.default" (list "cozystack.vm-default-images" $) }}
 {{include "cozystack.platform.package.optional.default" (list "cozystack.gpu-operator" $) }}
+{{include "cozystack.platform.package.optional.default" (list "cozystack.hami" $) }}
 {{include "cozystack.platform.package.default" (list "cozystack.kamaji" $) }}
 {{include "cozystack.platform.package.default" (list "cozystack.capi-operator" $) }}
 {{include "cozystack.platform.package.default" (list "cozystack.capi-provider-bootstrap-kubeadm" $) }}
diff --git a/packages/core/platform/templates/bundles/system.yaml b/packages/core/platform/templates/bundles/system.yaml
index 795b523f..6c587c82 100644
--- a/packages/core/platform/templates/bundles/system.yaml
+++ b/packages/core/platform/templates/bundles/system.yaml
@@ -25,10 +25,12 @@
 {{include "cozystack.platform.package" (list "cozystack.networking" "kubeovn-cilium" $ $networkingComponents) }}
 {{include "cozystack.platform.system.common-packages" $ }}
 {{include "cozystack.platform.package.default" (list "cozystack.linstor" $) }}
+{{include "cozystack.platform.package" (list "cozystack.cozystack-scheduler" "linstor" $) }}
 {{- end }}
 
 {{- if eq .Values.bundles.system.variant "isp-hosted" }}
 {{include "cozystack.platform.package" (list "cozystack.networking" "noop" $) }}
+{{include "cozystack.platform.package.default" (list "cozystack.cozystack-scheduler" $) }}
 {{- end }}
 
 {{- if eq .Values.bundles.system.variant "isp-full-generic" }}
@@ -92,6 +94,7 @@
 {{- /* Pass talos.enabled: false to linstor for generic Linux */ -}}
 {{- $linstorComponents := dict "linstor" (dict "values" (dict "talos" (dict "enabled" false))) -}}
 {{include "cozystack.platform.package" (list "cozystack.linstor" "default" $ $linstorComponents) }}
+{{include "cozystack.platform.package" (list "cozystack.cozystack-scheduler" "linstor" $) }}
 {{- end }}
 
 # Cozystack Engine
@@ -102,9 +105,6 @@
 {{- /* cozystack-api DaemonSet */ -}}
 {{- $apiValues := dict "cozystackAPI" (dict "nodeSelector" $genericNodeSelector) -}}
 {{- $_ := set $cozystackEngineComponents "cozystack-api" (dict "values" $apiValues) -}}
-{{- /* lineage-controller-webhook DaemonSet */ -}}
-{{- $lineageValues := dict "lineageControllerWebhook" (dict "nodeSelector" $genericNodeSelector) -}}
-{{- $_ := set $cozystackEngineComponents "lineage-controller-webhook" (dict "values" $lineageValues) -}}
 {{- end -}}
 {{- if .Values.authentication.oidc.enabled }}
 {{include "cozystack.platform.package" (list "cozystack.cozystack-engine" "oidc" $ $cozystackEngineComponents) }}
@@ -130,7 +130,6 @@
 {{include "cozystack.platform.package.default" (list "cozystack.cozystack-basics" $) }}
 {{include "cozystack.platform.package.default" (list "cozystack.backupstrategy-controller" $) }}
 {{include "cozystack.platform.package.default" (list "cozystack.backup-controller" $) }}
-{{include "cozystack.platform.package.default" (list "cozystack.cozystack-scheduler" $) }}
 {{include "cozystack.platform.package.default" (list "cozystack.vertical-pod-autoscaler" $) }}
 {{include "cozystack.platform.package.default" (list "cozystack.metrics-server" $) }}
 {{- $monitoringAgentsComponents := dict "monitoring-agents" (dict "values" (dict "global" (dict "target" "tenant-root"))) -}}
@@ -149,6 +148,7 @@
 {{include "cozystack.platform.package.optional.default" (list "cozystack.external-dns-application" $) }}
 {{include "cozystack.platform.package.optional.default" (list "cozystack.external-secrets-operator" $) }}
 {{include "cozystack.platform.package.optional.default" (list "cozystack.velero" $) }}
+{{include "cozystack.platform.package.optional.default" (list "cozystack.linstor-gui" $) }}
 {{- if has "cozystack.bootbox" (default (list) .Values.bundles.enabledPackages) }}
 {{include "cozystack.platform.package.default" (list "cozystack.bootbox-application" $) }}
 {{include "cozystack.platform.package.default" (list "cozystack.bootbox" $) }}
diff --git a/packages/core/platform/values.yaml b/packages/core/platform/values.yaml
index a2c97d8c..605e23bc 100644
--- a/packages/core/platform/values.yaml
+++ b/packages/core/platform/values.yaml
@@ -5,8 +5,8 @@ sourceRef:
   path: /
 migrations:
   enabled: false
-  image: ghcr.io/cozystack/cozystack/platform-migrations:v1.2.1@sha256:e8fcf006a4451fc0e961455e9b27a61b7103ee49b1a81fe5e4662ffed093fad6
-  targetVersion: 37
+  image: ghcr.io/cozystack/cozystack/platform-migrations:v1.3.0@sha256:555e4b76421361805a84bc9088b01b23a9c4a9430bd8ebd2db82ef9677d7008c
+  targetVersion: 39
 # Bundle deployment configuration
 bundles:
   system:
@@ -45,6 +45,41 @@ publishing:
     - cdi-uploadproxy
   apiServerEndpoint: "" # example: "https://api.example.org"
   externalIPs: []
+  # Exposure mode for the ingress-nginx Service. When "externalIPs" (current
+  # default) is selected, the Service is created as ClusterIP with
+  # Service.spec.externalIPs set from publishing.externalIPs. When
+  # "loadBalancer" is selected, the Service is type: LoadBalancer and a
+  # CiliumLoadBalancerIPPool makes those same addresses allocatable via LB IPAM.
+  #
+  # Service.spec.externalIPs is deprecated upstream in Kubernetes v1.36
+  # (KEP-5707). The AllowServiceExternalIPs feature gate is expected to default
+  # to false around v1.40 and the implementation removed around v1.43 — switch
+  # to "loadBalancer" before upgrading past v1.40.
+  #
+  # Caveats for the "loadBalancer" mode:
+  #   - publishing.externalIPs must contain at least one non-empty address,
+  #     otherwise the chart render fails with an explicit error (a LoadBalancer
+  #     Service without a pool would sit in <pending> forever).
+  #   - The ingress-nginx Service is created with externalTrafficPolicy: Local
+  #     to preserve the client source IP. Traffic arriving on a node that does
+  #     not host an ingress-nginx pod is dropped, so the external IP must be
+  #     routed to a node that runs the ingress pod (floating IP / keepalived /
+  #     upstream router / podAntiAffinity).
+  #   - Cilium does NOT announce the IP on its own unless L2 announcements or
+  #     BGP are enabled in the Cilium values (disabled by default in Cozystack).
+  #     This mode assumes the operator already routes the externalIPs to a
+  #     cluster node; enabling announcements is out of scope for this setting.
+  #   - Switching this value on a running cluster causes the ingress-nginx
+  #     Service to be recreated (the HelmRelease has upgrade.force: true and
+  #     the Service kind changes between ClusterIP and LoadBalancer). Expect a
+  #     brief interruption of ingress traffic during the flip.
+  #
+  # Scope: this setting only controls the ingress-nginx Service. Other
+  # cozystack components that currently write Service.spec.externalIPs directly
+  # (e.g. the vpn app at packages/apps/vpn/templates/service.yaml) are NOT
+  # migrated by flipping this value and must be addressed separately before
+  # the AllowServiceExternalIPs feature gate flips to off in ~v1.40.
+  exposure: externalIPs # "externalIPs" or "loadBalancer"
   certificates:
     solver: http01 # "http01" or "dns01"
     issuerName: letsencrypt-prod
diff --git a/packages/core/testing/values.yaml b/packages/core/testing/values.yaml
index b17da614..9d4bdd10 100644
--- a/packages/core/testing/values.yaml
+++ b/packages/core/testing/values.yaml
@@ -1,2 +1,2 @@
 e2e:
-  image: ghcr.io/cozystack/cozystack/e2e-sandbox:v1.2.1@sha256:7964a3e4b11053887be201d62f94138c3a7836c861036c26fb833aa1fcb934e1
+  image: ghcr.io/cozystack/cozystack/e2e-sandbox:v1.3.0@sha256:0367a03b981df2a3ea13f411d4cb7869c2bf2c89c07d3d5c8971b9a28921ccef
diff --git a/packages/extra/bootbox/images/matchbox.tag b/packages/extra/bootbox/images/matchbox.tag
index 0ba57d47..f51c7675 100644
--- a/packages/extra/bootbox/images/matchbox.tag
+++ b/packages/extra/bootbox/images/matchbox.tag
@@ -1 +1 @@
-ghcr.io/cozystack/cozystack/matchbox:v1.2.1@sha256:64685523ba693964f22ce7d7dc545bf7bf96dea34e5ae525a2315614adef4b3d
+ghcr.io/cozystack/cozystack/matchbox:v1.3.0@sha256:85b8e04bf6f0690612dd63e80475df269f4a436d16680f8a40f2860cf16e2f74
diff --git a/packages/extra/bootbox/templates/matchbox/ingress.yaml b/packages/extra/bootbox/templates/matchbox/ingress.yaml
index 4262cd10..5eef3489 100644
--- a/packages/extra/bootbox/templates/matchbox/ingress.yaml
+++ b/packages/extra/bootbox/templates/matchbox/ingress.yaml
@@ -10,7 +10,7 @@ metadata:
     app: bootbox
   annotations:
     {{- if eq $solver "http01" }}
-    acme.cert-manager.io/http01-ingress-class: {{ $ingress }}
+    acme.cert-manager.io/http01-ingress-ingressclassname: {{ $ingress }}
     {{- end }}
     cert-manager.io/cluster-issuer: {{ $clusterIssuer }}
     {{- if .Values.whitelistHTTP }}
diff --git a/packages/extra/etcd/Makefile b/packages/extra/etcd/Makefile
index f37d6e1e..2b3ed61e 100644
--- a/packages/extra/etcd/Makefile
+++ b/packages/extra/etcd/Makefile
@@ -5,3 +5,6 @@ include ../../../hack/package.mk
 generate:
 	cozyvalues-gen -v values.yaml -s values.schema.json -r README.md
 	../../../hack/update-crd.sh
+
+test:
+	helm unittest .
diff --git a/packages/extra/etcd/templates/hook/job.yaml b/packages/extra/etcd/templates/hook/job.yaml
deleted file mode 100644
index 3bf2f84b..00000000
--- a/packages/extra/etcd/templates/hook/job.yaml
+++ /dev/null
@@ -1,39 +0,0 @@
-{{- $shouldUpdateCerts := true }}
-{{- $configMap := lookup "v1" "ConfigMap" .Release.Namespace "etcd-deployed-version" }}
-{{- if $configMap }}
-  {{- $deployedVersion := index $configMap "data" "version" }}
-  {{- if $deployedVersion | semverCompare ">= 2.6.1" }}
-    {{- $shouldUpdateCerts = false }}
-  {{- end }}
-{{- end }}
-
-{{- if $shouldUpdateCerts }}
----
-apiVersion: batch/v1
-kind: Job
-metadata:
-  name: etcd-hook
-  annotations:
-    helm.sh/hook: post-upgrade
-    helm.sh/hook-delete-policy: before-hook-creation,hook-succeeded
-spec:
-  template:
-    metadata:
-      labels:
-        policy.cozystack.io/allow-to-apiserver: "true"
-    spec:
-      serviceAccountName: etcd-hook
-      containers:
-        - name: kubectl
-          image: docker.io/alpine/k8s:1.33.4
-          command:
-          - sh
-          args:
-            - -exc
-            - |-
-              kubectl --namespace={{ .Release.Namespace }} delete secrets etcd-ca-tls etcd-peer-ca-tls 
-              sleep 10
-              kubectl --namespace={{ .Release.Namespace }} delete secrets etcd-client-tls etcd-peer-tls etcd-server-tls
-              kubectl --namespace={{ .Release.Namespace }} delete pods --selector=app.kubernetes.io/instance=etcd,app.kubernetes.io/managed-by=etcd-operator,app.kubernetes.io/name=etcd,cozystack.io/service=etcd
-      restartPolicy: Never
-{{- end }}
diff --git a/packages/extra/etcd/templates/hook/role.yaml b/packages/extra/etcd/templates/hook/role.yaml
deleted file mode 100644
index 327eeadb..00000000
--- a/packages/extra/etcd/templates/hook/role.yaml
+++ /dev/null
@@ -1,26 +0,0 @@
-apiVersion: rbac.authorization.k8s.io/v1
-kind: Role
-metadata:
-  annotations:
-    helm.sh/hook: post-upgrade
-    helm.sh/hook-delete-policy: before-hook-creation,hook-succeeded
-  name: etcd-hook
-rules:
-- apiGroups:
-  - ""
-  resources:
-  - secrets
-  - pods
-  verbs:
-  - get
-  - list
-  - watch
-  - delete
-- apiGroups:
-  - ""
-  resources:
-  - configmaps
-  verbs:
-  - get
-  - list
-  - watch
diff --git a/packages/extra/etcd/templates/hook/rolebinding.yaml b/packages/extra/etcd/templates/hook/rolebinding.yaml
deleted file mode 100644
index 0ee0ffd1..00000000
--- a/packages/extra/etcd/templates/hook/rolebinding.yaml
+++ /dev/null
@@ -1,15 +0,0 @@
-apiVersion: rbac.authorization.k8s.io/v1
-kind: RoleBinding
-metadata:
-  name: etcd-hook
-  annotations:
-    helm.sh/hook: post-upgrade
-    helm.sh/hook-delete-policy: before-hook-creation,hook-succeeded
-roleRef:
-  apiGroup: rbac.authorization.k8s.io
-  kind: Role
-  name: etcd-hook
-subjects:
-  - kind: ServiceAccount
-    name: etcd-hook
-    namespace: {{ .Release.Namespace | quote }}
diff --git a/packages/extra/etcd/templates/hook/serviceaccount.yaml b/packages/extra/etcd/templates/hook/serviceaccount.yaml
deleted file mode 100644
index 552fb5fc..00000000
--- a/packages/extra/etcd/templates/hook/serviceaccount.yaml
+++ /dev/null
@@ -1,7 +0,0 @@
-apiVersion: v1
-kind: ServiceAccount
-metadata:
-  name: etcd-hook
-  annotations:
-    helm.sh/hook: post-upgrade
-    helm.sh/hook-delete-policy: before-hook-creation,hook-succeeded
diff --git a/packages/extra/etcd/templates/version.yaml b/packages/extra/etcd/templates/version.yaml
deleted file mode 100644
index cc9375bb..00000000
--- a/packages/extra/etcd/templates/version.yaml
+++ /dev/null
@@ -1,6 +0,0 @@
-apiVersion: v1
-kind: ConfigMap
-metadata:
-  name: etcd-deployed-version
-data:
-  version: {{ .Chart.Version }}
diff --git a/packages/extra/etcd/tests/no-post-upgrade-hook_test.yaml b/packages/extra/etcd/tests/no-post-upgrade-hook_test.yaml
new file mode 100644
index 00000000..0e4f5aa3
--- /dev/null
+++ b/packages/extra/etcd/tests/no-post-upgrade-hook_test.yaml
@@ -0,0 +1,34 @@
+suite: etcd chart does not ship a destructive post-upgrade cert-regeneration hook
+
+release:
+  name: etcd
+  namespace: tenant-root
+
+templates:
+  - templates/check-release-name.yaml
+  - templates/dashboard-resourcemap.yaml
+  - templates/datastore.yaml
+  - templates/etcd-defrag.yaml
+  - templates/hook/job.yaml
+  - templates/podscrape.yaml
+  - templates/prometheus-rules.yaml
+  - templates/version.yaml
+
+tests:
+  - it: renders no Job named etcd-hook
+    documentSelector:
+      path: metadata.name
+      value: etcd-hook
+      skipEmptyTemplates: true
+    asserts:
+      - hasDocuments:
+          count: 0
+
+  - it: renders no ConfigMap named etcd-deployed-version
+    documentSelector:
+      path: metadata.name
+      value: etcd-deployed-version
+      skipEmptyTemplates: true
+    asserts:
+      - hasDocuments:
+          count: 0
diff --git a/packages/extra/ingress/Makefile b/packages/extra/ingress/Makefile
index 958ce484..65b53c03 100644
--- a/packages/extra/ingress/Makefile
+++ b/packages/extra/ingress/Makefile
@@ -10,3 +10,6 @@ get-cloudflare-ips:
 generate:
 	cozyvalues-gen -v values.yaml -s values.schema.json -r README.md
 	../../../hack/update-crd.sh
+
+test:
+	helm unittest .
diff --git a/packages/extra/ingress/README.md b/packages/extra/ingress/README.md
index 0e786dfb..c541d8e9 100644
--- a/packages/extra/ingress/README.md
+++ b/packages/extra/ingress/README.md
@@ -14,3 +14,17 @@
 | `resources.memory` | Memory (RAM) available to each replica.                                                                                                 | `quantity` | `""`    |
 | `resourcesPreset`  | Default sizing preset used when `resources` is omitted.                                                                                 | `string`   | `micro` |
 
+
+## Exposure mode
+
+The ingress Service type is driven by the cluster-wide `publishing.exposure` value in the platform chart, not by any key in this package. Two modes exist:
+
+- `externalIPs` (default) has three rendered shapes:
+  - Release namespace matches `publishing.ingressName` AND `publishing.externalIPs` is non-empty → Service is `ClusterIP` with `Service.spec.externalIPs` set from that list and `externalTrafficPolicy: Cluster`.
+  - Release namespace matches `publishing.ingressName` but `publishing.externalIPs` is empty → Service falls back to `type: LoadBalancer` with `externalTrafficPolicy: Local`.
+  - Release namespace does not match `publishing.ingressName` (non-root tenants) → Service is `type: LoadBalancer` with `externalTrafficPolicy: Local`.
+  `Service.spec.externalIPs` is deprecated upstream in Kubernetes v1.36 (KEP-5707); plan migration before v1.40.
+- `loadBalancer` — Service is `type: LoadBalancer` with `externalTrafficPolicy: Local`, and a `CiliumLoadBalancerIPPool` makes the addresses in `publishing.externalIPs` allocatable via Cilium LB IPAM. Requires `publishing.externalIPs` to contain at least one non-empty address (render fails otherwise) and assumes the addresses are already routed to a cluster node (floating IP / upstream router). See the inline comment on `publishing.exposure` in the platform chart for full caveats, including the note that switching the value on a running cluster causes the ingress Service to be recreated.
+
+This setting only migrates ingress-nginx away from `Service.spec.externalIPs`. Other cozystack components that use the same deprecated field (e.g. the `vpn` app) must be migrated separately before Kubernetes v1.40 flips the `AllowServiceExternalIPs` feature gate off.
+
diff --git a/packages/extra/ingress/templates/nginx-ingress.yaml b/packages/extra/ingress/templates/nginx-ingress.yaml
index ca50d276..8e1f00fb 100644
--- a/packages/extra/ingress/templates/nginx-ingress.yaml
+++ b/packages/extra/ingress/templates/nginx-ingress.yaml
@@ -1,5 +1,19 @@
 {{- $exposeIngress := (index .Values._cluster "expose-ingress") | default "tenant-root" }}
 {{- $exposeExternalIPs := (index .Values._cluster "expose-external-ips") | default "" | nospace }}
+{{- $exposeMode := (index .Values._cluster "expose-mode") | default "externalIPs" }}
+{{- $exposeIPsList := list }}
+{{- range splitList "," $exposeExternalIPs }}
+  {{- $ip := . | trim }}
+  {{- if $ip }}
+    {{- $exposeIPsList = append $exposeIPsList $ip }}
+  {{- end }}
+{{- end }}
+{{- if not (has $exposeMode (list "externalIPs" "loadBalancer")) }}
+{{- fail (printf "unknown publishing.exposure mode %q: must be \"externalIPs\" or \"loadBalancer\"" $exposeMode) }}
+{{- end }}
+{{- if and (eq $exposeMode "loadBalancer") (eq $exposeIngress .Release.Namespace) (not $exposeIPsList) }}
+{{- fail "publishing.exposure=loadBalancer requires publishing.externalIPs to contain at least one non-empty address: the CiliumLoadBalancerIPPool has nothing to advertise and the Service would stay in <pending>." }}
+{{- end }}
 apiVersion: helm.toolkit.fluxcd.io/v2
 kind: HelmRelease
 metadata:
@@ -41,9 +55,12 @@ spec:
           enabled: false
         {{- end }}
         service:
-          {{- if and (eq $exposeIngress .Release.Namespace) $exposeExternalIPs }}
+          {{- if and (eq $exposeIngress .Release.Namespace) (eq $exposeMode "loadBalancer") }}
+          type: LoadBalancer
+          externalTrafficPolicy: Local
+          {{- else if and (eq $exposeIngress .Release.Namespace) $exposeIPsList }}
           externalIPs:
-            {{- toYaml (splitList "," $exposeExternalIPs) | nindent 12 }}
+            {{- toYaml $exposeIPsList | nindent 12 }}
           type: ClusterIP
           externalTrafficPolicy: Cluster
           {{- else }}
diff --git a/packages/extra/ingress/tests/exposure_test.yaml b/packages/extra/ingress/tests/exposure_test.yaml
new file mode 100644
index 00000000..1c76b3da
--- /dev/null
+++ b/packages/extra/ingress/tests/exposure_test.yaml
@@ -0,0 +1,165 @@
+suite: ingress exposure modes
+templates:
+  - templates/nginx-ingress.yaml
+
+release:
+  name: ingress
+  namespace: tenant-root
+
+tests:
+  - it: default exposure (externalIPs) renders ClusterIP Service with spec.externalIPs
+    set:
+      _cluster:
+        expose-ingress: tenant-root
+        expose-external-ips: "192.0.2.10,192.0.2.11"
+    asserts:
+      - template: templates/nginx-ingress.yaml
+        equal:
+          path: spec.values.ingress-nginx.controller.service.type
+          value: ClusterIP
+      - template: templates/nginx-ingress.yaml
+        equal:
+          path: spec.values.ingress-nginx.controller.service.externalTrafficPolicy
+          value: Cluster
+      - template: templates/nginx-ingress.yaml
+        equal:
+          path: spec.values.ingress-nginx.controller.service.externalIPs
+          value:
+            - 192.0.2.10
+            - 192.0.2.11
+      - template: templates/nginx-ingress.yaml
+        notExists:
+          path: spec.values.ingress-nginx.controller.service.labels
+
+  - it: legacy config without expose-mode falls back to externalIPs behavior
+    set:
+      _cluster:
+        expose-ingress: tenant-root
+        expose-external-ips: "192.0.2.10"
+    asserts:
+      - template: templates/nginx-ingress.yaml
+        equal:
+          path: spec.values.ingress-nginx.controller.service.type
+          value: ClusterIP
+
+  - it: externalIPs mode in a namespace other than publishing.ingressName renders LoadBalancer fallback without externalIPs
+    set:
+      _cluster:
+        expose-ingress: tenant-root
+        expose-external-ips: "192.0.2.10"
+    release:
+      namespace: tenant-other
+    asserts:
+      - template: templates/nginx-ingress.yaml
+        equal:
+          path: spec.values.ingress-nginx.controller.service.type
+          value: LoadBalancer
+      - template: templates/nginx-ingress.yaml
+        equal:
+          path: spec.values.ingress-nginx.controller.service.externalTrafficPolicy
+          value: Local
+      - template: templates/nginx-ingress.yaml
+        notExists:
+          path: spec.values.ingress-nginx.controller.service.externalIPs
+
+  - it: loadBalancer mode renders LoadBalancer Service without externalIPs on the Service
+    set:
+      _cluster:
+        expose-ingress: tenant-root
+        expose-external-ips: "192.0.2.10,192.0.2.11"
+        expose-mode: loadBalancer
+    asserts:
+      - template: templates/nginx-ingress.yaml
+        equal:
+          path: spec.values.ingress-nginx.controller.service.type
+          value: LoadBalancer
+      - template: templates/nginx-ingress.yaml
+        equal:
+          path: spec.values.ingress-nginx.controller.service.externalTrafficPolicy
+          value: Local
+      - template: templates/nginx-ingress.yaml
+        notExists:
+          path: spec.values.ingress-nginx.controller.service.labels
+      - template: templates/nginx-ingress.yaml
+        notExists:
+          path: spec.values.ingress-nginx.controller.service.externalIPs
+
+  - it: loadBalancer mode without externalIPs fails chart render with explicit message
+    set:
+      _cluster:
+        expose-ingress: tenant-root
+        expose-external-ips: ""
+        expose-mode: loadBalancer
+    asserts:
+      - template: templates/nginx-ingress.yaml
+        failedTemplate:
+          errorMessage: "publishing.exposure=loadBalancer requires publishing.externalIPs to contain at least one non-empty address: the CiliumLoadBalancerIPPool has nothing to advertise and the Service would stay in <pending>."
+
+  - it: unknown exposure mode is rejected with a clear error (case-sensitive enum)
+    set:
+      _cluster:
+        expose-ingress: tenant-root
+        expose-external-ips: "192.0.2.10"
+        expose-mode: LoadBalancer
+    asserts:
+      - template: templates/nginx-ingress.yaml
+        failedTemplate:
+          errorMessage: "unknown publishing.exposure mode \"LoadBalancer\": must be \"externalIPs\" or \"loadBalancer\""
+
+  - it: another typo in exposure mode also fails
+    set:
+      _cluster:
+        expose-ingress: tenant-root
+        expose-external-ips: "192.0.2.10"
+        expose-mode: loadbalancer
+    asserts:
+      - template: templates/nginx-ingress.yaml
+        failedTemplate:
+          errorMessage: "unknown publishing.exposure mode \"loadbalancer\": must be \"externalIPs\" or \"loadBalancer\""
+
+  - it: loadBalancer mode in a namespace other than publishing.ingressName falls back to LoadBalancer Service without pool
+    set:
+      _cluster:
+        expose-ingress: tenant-root
+        expose-external-ips: "192.0.2.10"
+        expose-mode: loadBalancer
+    release:
+      namespace: tenant-other
+    asserts:
+      - template: templates/nginx-ingress.yaml
+        equal:
+          path: spec.values.ingress-nginx.controller.service.type
+          value: LoadBalancer
+      - template: templates/nginx-ingress.yaml
+        equal:
+          path: spec.values.ingress-nginx.controller.service.externalTrafficPolicy
+          value: Local
+      - template: templates/nginx-ingress.yaml
+        notExists:
+          path: spec.values.ingress-nginx.controller.service.labels
+      - template: templates/nginx-ingress.yaml
+        notExists:
+          path: spec.values.ingress-nginx.controller.service.externalIPs
+
+  - it: loadBalancer mode with only-empty externalIPs fails chart render (comma-only input)
+    set:
+      _cluster:
+        expose-ingress: tenant-root
+        expose-external-ips: ",,"
+        expose-mode: loadBalancer
+    asserts:
+      - template: templates/nginx-ingress.yaml
+        failedTemplate:
+          errorMessage: "publishing.exposure=loadBalancer requires publishing.externalIPs to contain at least one non-empty address: the CiliumLoadBalancerIPPool has nothing to advertise and the Service would stay in <pending>."
+
+  - it: externalIPs mode also filters out empty entries (trailing comma)
+    set:
+      _cluster:
+        expose-ingress: tenant-root
+        expose-external-ips: "192.0.2.10,"
+    asserts:
+      - template: templates/nginx-ingress.yaml
+        equal:
+          path: spec.values.ingress-nginx.controller.service.externalIPs
+          value:
+            - 192.0.2.10
diff --git a/packages/extra/seaweedfs/images/objectstorage-sidecar.tag b/packages/extra/seaweedfs/images/objectstorage-sidecar.tag
index bb0702e8..9b413da5 100644
--- a/packages/extra/seaweedfs/images/objectstorage-sidecar.tag
+++ b/packages/extra/seaweedfs/images/objectstorage-sidecar.tag
@@ -1 +1 @@
-ghcr.io/cozystack/cozystack/objectstorage-sidecar:v1.2.1@sha256:1ad07fb9e96477e5c322f240f6023ceedd8762261d8379ec784dde92797ee206
+ghcr.io/cozystack/cozystack/objectstorage-sidecar:v1.3.0@sha256:5bcccbdb13979a16cee535eb5fbcdf0d87973689010a10b7b25e55c5be3edaa6
diff --git a/packages/extra/seaweedfs/templates/seaweedfs.yaml b/packages/extra/seaweedfs/templates/seaweedfs.yaml
index 42441b26..2f5720ca 100644
--- a/packages/extra/seaweedfs/templates/seaweedfs.yaml
+++ b/packages/extra/seaweedfs/templates/seaweedfs.yaml
@@ -243,7 +243,7 @@ spec:
             nginx.ingress.kubernetes.io/proxy-body-size: "0"
             nginx.ingress.kubernetes.io/backend-protocol: "HTTPS"
             {{- if eq $solver "http01" }}
-            acme.cert-manager.io/http01-ingress-class: {{ $ingress }}
+            acme.cert-manager.io/http01-ingress-ingressclassname: {{ $ingress }}
             {{- end }}
             cert-manager.io/cluster-issuer: {{ $clusterIssuer }}
           tls:
diff --git a/packages/system/backup-controller/definitions/backups.cozystack.io_backups.yaml b/packages/system/backup-controller/definitions/backups.cozystack.io_backups.yaml
index 6e031a84..d48df533 100644
--- a/packages/system/backup-controller/definitions/backups.cozystack.io_backups.yaml
+++ b/packages/system/backup-controller/definitions/backups.cozystack.io_backups.yaml
@@ -142,14 +142,6 @@ spec:
                 required:
                 - uri
                 type: object
-              underlyingResources:
-                description: |-
-                  Holds application-specific resource metadata discovered during backup.
-                  The payload is a self-typed JSON object carrying an inlined TypeMeta
-                  (kind/apiVersion) so the consuming controller can dispatch on the
-                  application kind.
-                type: object
-                x-kubernetes-preserve-unknown-fields: true
               conditions:
                 description: Conditions represents the latest available observations
                   of a Backup's state.
@@ -213,6 +205,14 @@ spec:
                   Phase is a simple, high-level summary of the backup's state.
                   Typical values are: Pending, Ready, Failed.
                 type: string
+              underlyingResources:
+                description: |-
+                  UnderlyingResources holds application-specific resource metadata discovered
+                  during backup (e.g., VM disks, network configuration). The payload is a
+                  self-typed JSON object carrying an inlined TypeMeta (kind/apiVersion) so
+                  the consuming controller can dispatch on the application kind.
+                type: object
+                x-kubernetes-preserve-unknown-fields: true
             type: object
         type: object
     selectableFields:
diff --git a/packages/system/backup-controller/values.yaml b/packages/system/backup-controller/values.yaml
index ff5fe774..34e033ce 100644
--- a/packages/system/backup-controller/values.yaml
+++ b/packages/system/backup-controller/values.yaml
@@ -1,5 +1,5 @@
 backupController:
-  image: "ghcr.io/cozystack/cozystack/backup-controller:v1.2.1@sha256:6e515cdfe302bb079dafa0f7f302c6ff7b7e9c02a75574880a45e35d5acb1f59"
+  image: "ghcr.io/cozystack/cozystack/backup-controller:v1.3.0@sha256:e1a083dc92f26dfef004f47c1cd20a6357174aad835004f58e751c494b76649a"
   replicas: 2
   debug: false
   metrics:
diff --git a/packages/system/backupstrategy-controller/templates/rbac-bind.yaml b/packages/system/backupstrategy-controller/templates/rbac-bind.yaml
index 4ecacce0..03578cc2 100644
--- a/packages/system/backupstrategy-controller/templates/rbac-bind.yaml
+++ b/packages/system/backupstrategy-controller/templates/rbac-bind.yaml
@@ -10,17 +10,3 @@ subjects:
 - kind: ServiceAccount
   name: backupstrategy-controller
   namespace: {{ .Release.Namespace }}
----
-apiVersion: rbac.authorization.k8s.io/v1
-kind: RoleBinding
-metadata:
-  name: backups.cozystack.io:strategy-controller:velero-configmaps
-  namespace: cozy-velero
-roleRef:
-  apiGroup: rbac.authorization.k8s.io
-  kind: Role
-  name: backups.cozystack.io:strategy-controller:velero-configmaps
-subjects:
-- kind: ServiceAccount
-  name: backupstrategy-controller
-  namespace: {{ .Release.Namespace }}
diff --git a/packages/system/backupstrategy-controller/templates/rbac.yaml b/packages/system/backupstrategy-controller/templates/rbac.yaml
index bbdb74f8..634ea88b 100644
--- a/packages/system/backupstrategy-controller/templates/rbac.yaml
+++ b/packages/system/backupstrategy-controller/templates/rbac.yaml
@@ -27,7 +27,9 @@ rules:
   resources: ["pods"]
   verbs: ["get", "list", "watch"]
 # ConfigMaps: controller-runtime cache requires cluster-scoped list/watch;
-# create/update/delete is scoped to cozy-velero via the Role below.
+# create/update/delete is scoped to cozy-velero via a Role shipped by the
+# velero chart (packages/system/velero/templates/backupstrategy-controller-rbac.yaml)
+# so it is only created when velero is installed.
 - apiGroups: [""]
   resources: ["configmaps"]
   verbs: ["get", "list", "watch"]
@@ -78,14 +80,3 @@ rules:
 - apiGroups: ["coordination.k8s.io"]
   resources: ["leases"]
   verbs: ["get", "list", "watch", "create", "update", "patch"]
----
-# To create ResourceModifiers in ConfigMaps for Restore in Velero install namespace.
-apiVersion: rbac.authorization.k8s.io/v1
-kind: Role
-metadata:
-  name: backups.cozystack.io:strategy-controller:velero-configmaps
-  namespace: cozy-velero
-rules:
-- apiGroups: [""]
-  resources: ["configmaps"]
-  verbs: ["create", "get", "list", "delete", "deletecollection", "patch", "update"]
diff --git a/packages/system/backupstrategy-controller/values.yaml b/packages/system/backupstrategy-controller/values.yaml
index bd3cad8c..d6453124 100644
--- a/packages/system/backupstrategy-controller/values.yaml
+++ b/packages/system/backupstrategy-controller/values.yaml
@@ -1,5 +1,5 @@
 backupStrategyController:
-  image: "ghcr.io/cozystack/cozystack/backupstrategy-controller:v1.2.1@sha256:3544a08bfac2df6dcb1842f73747768b24625c6d6b97230be9a00f80637f0d32"
+  image: "ghcr.io/cozystack/cozystack/backupstrategy-controller:v1.3.0@sha256:be0a9ec1f4307064b16388a24628aee46e06252738338add80b99ea1e04e62bf"
   replicas: 2
   debug: false
   metrics:
diff --git a/packages/system/bucket/images/s3manager.tag b/packages/system/bucket/images/s3manager.tag
index 6c3d1449..889db0af 100644
--- a/packages/system/bucket/images/s3manager.tag
+++ b/packages/system/bucket/images/s3manager.tag
@@ -1 +1 @@
-ghcr.io/cozystack/cozystack/s3manager:v0.5.0@sha256:73382d7911274e6528cc2272ca9bfdb095c59e78845a3cdffd39d6c8ac29d920
+ghcr.io/cozystack/cozystack/s3manager:v0.5.0@sha256:fb65e734bbdbdaef2238769cc2ecfb54dbddaf1e0952ad438a9b3e26b9dbb4b5
diff --git a/packages/system/bucket/templates/ingress.yaml b/packages/system/bucket/templates/ingress.yaml
index 6e5028fc..b7ffaf8b 100644
--- a/packages/system/bucket/templates/ingress.yaml
+++ b/packages/system/bucket/templates/ingress.yaml
@@ -12,7 +12,7 @@ metadata:
     nginx.ingress.kubernetes.io/proxy-read-timeout: "99999"
     nginx.ingress.kubernetes.io/proxy-send-timeout: "99999"
     {{- if eq $solver "http01" }}
-    acme.cert-manager.io/http01-ingress-class: {{ $ingress }}
+    acme.cert-manager.io/http01-ingress-ingressclassname: {{ $ingress }}
     {{- end }}
     cert-manager.io/cluster-issuer: {{ $clusterIssuer }}
 spec:
diff --git a/packages/system/cert-manager-issuers/templates/cluster-issuers.yaml b/packages/system/cert-manager-issuers/templates/cluster-issuers.yaml
index 442f1fb3..3b582082 100644
--- a/packages/system/cert-manager-issuers/templates/cluster-issuers.yaml
+++ b/packages/system/cert-manager-issuers/templates/cluster-issuers.yaml
@@ -1,6 +1,7 @@
 {{- $solver := (index .Values._cluster "solver") | default "http01" }}
+{{- $exposeIngress := (index .Values._cluster "expose-ingress") | default "tenant-root" }}
 
-apiVersion: cert-manager.io/v1 
+apiVersion: cert-manager.io/v1
 kind: ClusterIssuer  
 metadata:  
   name: letsencrypt-prod  
@@ -17,9 +18,9 @@ spec:
               name: cloudflare-api-token-secret  
               key: api-token  
       {{- else }}  
-        http01:  
-          ingress:  
-            class: nginx  
+        http01:
+          ingress:
+            ingressClassName: {{ $exposeIngress }}
       {{- end }}  
 
 ---  
@@ -41,9 +42,9 @@ spec:
               name: cloudflare-api-token-secret  
               key: api-token  
       {{- else }}  
-        http01:  
-          ingress:  
-            class: nginx  
+        http01:
+          ingress:
+            ingressClassName: {{ $exposeIngress }}
       {{- end }}  
 
 ---  
diff --git a/packages/system/cilium/Makefile b/packages/system/cilium/Makefile
index 07d4ca97..71b47a77 100644
--- a/packages/system/cilium/Makefile
+++ b/packages/system/cilium/Makefile
@@ -12,6 +12,15 @@ update:
 	helm repo update cilium
 	helm pull cilium/cilium --untar --untardir charts --version 1.19
 	$(SED_INPLACE) -e '/Used in iptables/d' -e '/SYS_MODULE/d' charts/cilium/values.yaml
+	# Drop the whole k8s<1.30 AppArmor annotations block from the cilium-agent
+	# DaemonSet. Cozystack manages these annotations through cilium.podAnnotations
+	# in values.yaml on every k8s version, so keeping the upstream block would
+	# produce duplicate mapping keys on k8s<1.30.
+	perl -i -0pe 's|\n        \{\{- if not \.Values\.securityContext\.privileged \}\}\n        \{\{- if semverCompare "<1\.30\.0".*?\n        \{\{- end \}\}\n        \{\{- end \}\}\n        \{\{- end \}\}||s' \
+		charts/cilium/templates/cilium-agent/daemonset.yaml
+	@! grep -q 'container\.apparmor\.security\.beta\.kubernetes\.io' \
+		charts/cilium/templates/cilium-agent/daemonset.yaml || \
+		{ echo 'ERROR: perl patch did not remove the upstream AppArmor block from cilium-agent/daemonset.yaml (upstream template format may have changed)' >&2; exit 1; }
 	version=$$(awk '$$1 == "version:" {print $$2}' charts/cilium/Chart.yaml) && \
 	$(SED_INPLACE) "s/ARG VERSION=.*/ARG VERSION=v$${version}/" images/cilium/Dockerfile
 
diff --git a/packages/system/cilium/charts/cilium/Chart.yaml b/packages/system/cilium/charts/cilium/Chart.yaml
index 0bb34451..5fe2baed 100644
--- a/packages/system/cilium/charts/cilium/Chart.yaml
+++ b/packages/system/cilium/charts/cilium/Chart.yaml
@@ -76,7 +76,7 @@ annotations:
     Cilium Gateway Class Config\n  description: |\n    CiliumGatewayClassConfig defines
     a configuration for Gateway API GatewayClass.\n"
 apiVersion: v2
-appVersion: 1.19.1
+appVersion: 1.19.3
 description: eBPF-based Networking, Security, and Observability
 home: https://cilium.io/
 icon: https://cdn.jsdelivr.net/gh/cilium/cilium@main/Documentation/images/logo-solo.svg
@@ -92,4 +92,4 @@ kubeVersion: '>= 1.21.0-0'
 name: cilium
 sources:
 - https://github.com/cilium/cilium
-version: 1.19.1
+version: 1.19.3
diff --git a/packages/system/cilium/charts/cilium/README.md b/packages/system/cilium/charts/cilium/README.md
index fe8aea3f..d7697c56 100644
--- a/packages/system/cilium/charts/cilium/README.md
+++ b/packages/system/cilium/charts/cilium/README.md
@@ -1,6 +1,6 @@
 # cilium
 
-![Version: 1.19.1](https://img.shields.io/badge/Version-1.19.1-informational?style=flat-square) ![AppVersion: 1.19.1](https://img.shields.io/badge/AppVersion-1.19.1-informational?style=flat-square)
+![Version: 1.19.3](https://img.shields.io/badge/Version-1.19.3-informational?style=flat-square) ![AppVersion: 1.19.3](https://img.shields.io/badge/AppVersion-1.19.3-informational?style=flat-square)
 
 Cilium is open source software for providing and transparently securing
 network connectivity and loadbalancing between application workloads such as
@@ -89,7 +89,7 @@ contributors across the globe, there is almost always someone available to help.
 | authentication.mutual.spire.install.agent.tolerations | list | `[{"effect":"NoSchedule","key":"node.kubernetes.io/not-ready"},{"effect":"NoSchedule","key":"node-role.kubernetes.io/master"},{"effect":"NoSchedule","key":"node-role.kubernetes.io/control-plane"},{"effect":"NoSchedule","key":"node.cloudprovider.kubernetes.io/uninitialized","value":"true"},{"key":"CriticalAddonsOnly","operator":"Exists"}]` | SPIRE agent tolerations configuration By default it follows the same tolerations as the agent itself to allow the Cilium agent on this node to connect to SPIRE. ref: https://kubernetes.io/docs/concepts/scheduling-eviction/taint-and-toleration/ |
 | authentication.mutual.spire.install.enabled | bool | `true` | Enable SPIRE installation. This will only take effect only if authentication.mutual.spire.enabled is true |
 | authentication.mutual.spire.install.existingNamespace | bool | `false` | SPIRE namespace already exists. Set to true if Helm should not create, manage, and import the SPIRE namespace. |
-| authentication.mutual.spire.install.initImage | object | `{"digest":"sha256:b3255e7dfbcd10cb367af0d409747d511aeb66dfac98cf30e97e87e4207dd76f","override":null,"pullPolicy":"IfNotPresent","repository":"docker.io/library/busybox","tag":"1.37.0","useDigest":true}` | init container image of SPIRE agent and server |
+| authentication.mutual.spire.install.initImage | object | `{"digest":"sha256:1487d0af5f52b4ba31c7e465126ee2123fe3f2305d638e7827681e7cf6c83d5e","override":null,"pullPolicy":"IfNotPresent","repository":"docker.io/library/busybox","tag":"1.37.0","useDigest":true}` | init container image of SPIRE agent and server |
 | authentication.mutual.spire.install.namespace | string | `"cilium-spire"` | SPIRE namespace to install into |
 | authentication.mutual.spire.install.server.affinity | object | `{}` | SPIRE server affinity configuration |
 | authentication.mutual.spire.install.server.annotations | object | `{}` | SPIRE server annotations |
@@ -175,7 +175,7 @@ contributors across the globe, there is almost always someone available to help.
 | bpf.tproxy | bool | `false` | Configure the eBPF-based TPROXY (beta) to reduce reliance on iptables rules for implementing Layer 7 policy. Note this is incompatible with netkit (`bpf.datapathMode=netkit`, `bpf.datapathMode=netkit-l2`). |
 | bpf.vlanBypass | list | `[]` | Configure explicitly allowed VLAN id's for bpf logic bypass. [0] will allow all VLAN id's without any filtering. |
 | bpfClockProbe | bool | `false` | Enable BPF clock source probing for more efficient tick retrieval. |
-| certgen | object | `{"affinity":{},"annotations":{"cronJob":{},"job":{}},"cronJob":{"failedJobsHistoryLimit":1,"successfulJobsHistoryLimit":3},"extraVolumeMounts":[],"extraVolumes":[],"generateCA":true,"image":{"digest":"sha256:19921f48ee7e2295ea4dca955878a6cd8d70e6d4219d08f688e866ece9d95d4d","override":null,"pullPolicy":"IfNotPresent","repository":"quay.io/cilium/certgen","tag":"v0.3.2","useDigest":true},"nodeSelector":{},"podLabels":{},"priorityClassName":"","resources":{},"tolerations":[],"ttlSecondsAfterFinished":null}` | Configure certificate generation for Hubble integration. If hubble.tls.auto.method=cronJob, these values are used for the Kubernetes CronJob which will be scheduled regularly to (re)generate any certificates not provided manually. |
+| certgen | object | `{"affinity":{},"annotations":{"cronJob":{},"job":{}},"cronJob":{"failedJobsHistoryLimit":1,"successfulJobsHistoryLimit":3},"extraVolumeMounts":[],"extraVolumes":[],"generateCA":true,"image":{"digest":"sha256:f0c656830e856d26b24b0e144df1f8b327d3b46748d76a630514111fc365b697","override":null,"pullPolicy":"IfNotPresent","repository":"quay.io/cilium/certgen","tag":"v0.4.1","useDigest":true},"nodeSelector":{},"podLabels":{},"priorityClassName":"","resources":{},"tolerations":[],"ttlSecondsAfterFinished":null}` | Configure certificate generation for Hubble integration. If hubble.tls.auto.method=cronJob, these values are used for the Kubernetes CronJob which will be scheduled regularly to (re)generate any certificates not provided manually. |
 | certgen.affinity | object | `{}` | Affinity for certgen |
 | certgen.annotations | object | `{"cronJob":{},"job":{}}` | Annotations to be added to the hubble-certgen initial Job and CronJob |
 | certgen.cronJob.failedJobsHistoryLimit | int | `1` | The number of failed finished jobs to keep |
@@ -214,7 +214,7 @@ contributors across the globe, there is almost always someone available to help.
 | clustermesh.apiserver.extraVolumeMounts | list | `[]` | Additional clustermesh-apiserver volumeMounts. |
 | clustermesh.apiserver.extraVolumes | list | `[]` | Additional clustermesh-apiserver volumes. |
 | clustermesh.apiserver.healthPort | int | `9880` | TCP port for the clustermesh-apiserver health API. |
-| clustermesh.apiserver.image | object | `{"digest":"sha256:56d6c3dc13b50126b80ecb571707a0ea97f6db694182b9d61efd386d04e5bb28","override":null,"pullPolicy":"IfNotPresent","repository":"quay.io/cilium/clustermesh-apiserver","tag":"v1.19.1","useDigest":true}` | Clustermesh API server image. |
+| clustermesh.apiserver.image | object | `{"digest":"sha256:a8136a7615d6c6041d3aa6f2674d17beaec238170d669507ccc05328a778e2b7","override":null,"pullPolicy":"IfNotPresent","repository":"quay.io/cilium/clustermesh-apiserver","tag":"v1.19.3","useDigest":true}` | Clustermesh API server image. |
 | clustermesh.apiserver.kvstoremesh.enabled | bool | `true` | Enable KVStoreMesh. KVStoreMesh caches the information retrieved from the remote clusters in the local etcd instance (deprecated - KVStoreMesh will always be enabled once the option is removed). |
 | clustermesh.apiserver.kvstoremesh.extraArgs | list | `[]` | Additional KVStoreMesh arguments. |
 | clustermesh.apiserver.kvstoremesh.extraEnv | list | `[]` | Additional KVStoreMesh environment variables. |
@@ -340,6 +340,10 @@ contributors across the globe, there is almost always someone available to help.
 | cni.resources | object | `{"limits":{"cpu":1,"memory":"1Gi"},"requests":{"cpu":"100m","memory":"10Mi"}}` | Specifies the resources for the cni initContainer |
 | cni.uninstall | bool | `false` | Remove the CNI configuration and binary files on agent shutdown. Enable this if you're removing Cilium from the cluster. Disable this to prevent the CNI configuration file from being removed during agent upgrade, which can cause nodes to go unmanageable. |
 | commonLabels | object | `{}` | commonLabels allows users to add common labels for all Cilium resources. |
+| configDriftDetection | object | `{"driftChecker":true,"enabled":true,"ignoredKeys":[]}` | Configuration for the ConfigMap drift detection feature. When enabled, the agent continuously watches the cilium-config ConfigMap and exposes a cilium_drift_checker_config_delta Prometheus metric reporting the number of keys that differ between the ConfigMap and the agent's active settings. A non-zero value indicates that the agent has not yet applied all current ConfigMap changes and needs to be restarted. |
+| configDriftDetection.driftChecker | bool | `true` | Enable the drift checker which compares the DynamicConfig table against the agent's active settings and publishes the cilium_drift_checker_config_delta metric. |
+| configDriftDetection.enabled | bool | `true` | Enable watching of the cilium-config ConfigMap and reflecting its contents into the agent's internal DynamicConfig table. |
+| configDriftDetection.ignoredKeys | list | `[]` | List of config-map keys to ignore when computing the drift delta. |
 | connectivityProbeFrequencyRatio | float64 | `0.5` | Ratio of the connectivity probe frequency vs resource usage, a float in [0, 1]. 0 will give more frequent probing, 1 will give less frequent probing. Probing frequency is dynamically adjusted based on the cluster size. |
 | conntrackGCInterval | string | `"0s"` | Configure how frequently garbage collection should occur for the datapath connection tracking table. |
 | conntrackGCMaxInterval | string | `""` | Configure the maximum frequency for the garbage collection of the connection tracking table. Only affects the automatic computation for the frequency and has no effect when 'conntrackGCInterval' is set. This can be set to more frequently clean up unused identities created from ToFQDN policies. |
@@ -380,7 +384,6 @@ contributors across the globe, there is almost always someone available to help.
 | enableMasqueradeRouteSource | bool | `false` | Enables masquerading to the source of the route for traffic leaving the node from endpoints. |
 | enableNoServiceEndpointsRoutable | bool | `true` | Enable routing to a service that has zero endpoints |
 | enableNonDefaultDenyPolicies | bool | `true` | Enable Non-Default-Deny policies |
-| enableTunnelBIGTCP | bool | `false` | Enable BIG TCP in tunneling mode and increase maximum GRO/GSO limits for VXLAN/GENEVE tunnels |
 | enableXTSocketFallback | bool | `true` | Enables the fallback compatibility solution for when the xt_socket kernel module is missing and it is needed for the datapath L7 redirection to work properly. See documentation for details on when this can be disabled: https://docs.cilium.io/en/stable/operations/system_requirements/#linux-kernel. |
 | encryption.enabled | bool | `false` | Enable transparent network encryption. |
 | encryption.ipsec.encryptedOverlay | bool | `false` | Enable IPsec encrypted overlay |
@@ -401,8 +404,29 @@ contributors across the globe, there is almost always someone available to help.
 | encryption.strictMode.ingress.enabled | bool | `false` | Enable strict ingress encryption. When enabled, all unencrypted overlay ingress traffic will be dropped. This option is only applicable when WireGuard and tunneling are enabled. |
 | encryption.type | string | `"ipsec"` | Encryption method. Can be one of ipsec, wireguard or ztunnel. |
 | encryption.wireguard.persistentKeepalive | string | `"0s"` | Controls WireGuard PersistentKeepalive option. Set 0s to disable. |
+| encryption.ztunnel | object | `{"affinity":{},"annotations":{},"caAddress":"https://localhost:15012","extraEnv":[],"extraVolumeMounts":[],"extraVolumes":[],"healthPort":15021,"image":{"digest":null,"override":null,"pullPolicy":"IfNotPresent","repository":"docker.io/istio/ztunnel","tag":"1.28.0-distroless","useDigest":false},"nodeSelector":{"kubernetes.io/os":"linux"},"podAnnotations":{},"podLabels":{},"priorityClassName":null,"readinessProbe":{"failureThreshold":3,"initialDelaySeconds":0,"periodSeconds":10},"resources":{"requests":{"cpu":"200m","memory":"512Mi"}},"secrets":{"bootstrapRootCert":null},"terminationGracePeriodSeconds":30,"tolerations":[{"effect":"NoSchedule","operator":"Exists"},{"key":"CriticalAddonsOnly","operator":"Exists"},{"effect":"NoExecute","operator":"Exists"}],"updateStrategy":{"rollingUpdate":{"maxSurge":1,"maxUnavailable":0},"type":"RollingUpdate"}}` | ztunnel encryption configuration. ztunnel is Istio's purpose-built, per-node proxy for handling L4 traffic in ambient mesh mode. These settings only apply when encryption.type is set to "ztunnel". |
+| encryption.ztunnel.affinity | object | `{}` | Affinity for ztunnel pods. |
+| encryption.ztunnel.annotations | object | `{}` | Annotations to be added to all ztunnel resources. |
+| encryption.ztunnel.caAddress | string | `"https://localhost:15012"` | CA server address for certificate requests. |
+| encryption.ztunnel.extraEnv | list | `[]` | Additional ztunnel container environment variables. |
+| encryption.ztunnel.extraVolumeMounts | list | `[]` | Additional ztunnel volumeMounts. |
+| encryption.ztunnel.extraVolumes | list | `[]` | Additional ztunnel volumes. |
+| encryption.ztunnel.healthPort | int | `15021` | TCP port for the health API. |
+| encryption.ztunnel.image | object | `{"digest":null,"override":null,"pullPolicy":"IfNotPresent","repository":"docker.io/istio/ztunnel","tag":"1.28.0-distroless","useDigest":false}` | ztunnel container image. |
+| encryption.ztunnel.nodeSelector | object | `{"kubernetes.io/os":"linux"}` | Node selector for ztunnel pods. |
+| encryption.ztunnel.podAnnotations | object | `{}` | Annotations to be added to ztunnel pods. |
+| encryption.ztunnel.podLabels | object | `{}` | Labels to be added to ztunnel pods. |
+| encryption.ztunnel.priorityClassName | string | `nil` | The priority class to use for ztunnel pods. |
+| encryption.ztunnel.readinessProbe | object | `{"failureThreshold":3,"initialDelaySeconds":0,"periodSeconds":10}` | Readiness probe configuration. |
+| encryption.ztunnel.resources | object | `{"requests":{"cpu":"200m","memory":"512Mi"}}` | ztunnel resource limits & requests. |
+| encryption.ztunnel.secrets | object | `{"bootstrapRootCert":null}` | ztunnel secrets configuration. |
+| encryption.ztunnel.secrets.bootstrapRootCert | string | `nil` | Base64-encoded bootstrap root certificate content. If not provided, the secret must be created manually before deploying. @schema type: [null, string] @schema |
+| encryption.ztunnel.terminationGracePeriodSeconds | int | `30` | Configure termination grace period for ztunnel DaemonSet. |
+| encryption.ztunnel.tolerations | list | `[{"effect":"NoSchedule","operator":"Exists"},{"key":"CriticalAddonsOnly","operator":"Exists"},{"effect":"NoExecute","operator":"Exists"}]` | Node tolerations for ztunnel scheduling. |
+| encryption.ztunnel.updateStrategy | object | `{"rollingUpdate":{"maxSurge":1,"maxUnavailable":0},"type":"RollingUpdate"}` | ztunnel update strategy. |
 | endpointHealthChecking.enabled | bool | `true` | Enable connectivity health checking between virtual endpoints. |
 | endpointLockdownOnMapOverflow | bool | `false` | Enable endpoint lockdown on policy map overflow. |
+| endpointPolicyUpdateTimeoutDuration | string | `nil` | Max duration to wait for envoy to respond to configuration changes. Default "10s". |
 | endpointRoutes.enabled | bool | `false` | Enable use of per endpoint routes instead of routing via the cilium_host interface. |
 | eni.awsEnablePrefixDelegation | bool | `false` | Enable ENI prefix delegation |
 | eni.awsReleaseExcessIPs | bool | `false` | Release IPs not used from the ENI |
@@ -446,7 +470,7 @@ contributors across the globe, there is almost always someone available to help.
 | envoy.httpRetryCount | int | `3` | Maximum number of retries for each HTTP request |
 | envoy.httpUpstreamLingerTimeout | string | `nil` | Time in seconds to block Envoy worker thread while an upstream HTTP connection is closing. If set to 0, the connection is closed immediately (with TCP RST). If set to -1, the connection is closed asynchronously in the background. |
 | envoy.idleTimeoutDurationSeconds | int | `60` | Set Envoy upstream HTTP idle connection timeout seconds. Does not apply to connections with pending requests. Default 60s |
-| envoy.image | object | `{"digest":"sha256:8188114a2768b5f49d6ce58e168b20d765e0fbc64eee0d83241aa2b150ccd788","override":null,"pullPolicy":"IfNotPresent","repository":"quay.io/cilium/cilium-envoy","tag":"v1.35.9-1770979049-232ed4a26881e4ab4f766f251f258ed424fff663","useDigest":true}` | Envoy container image. |
+| envoy.image | object | `{"digest":"sha256:ba0ab8adac082d50d525fd2c5ba096c8facea3a471561b7c61c7a5b9c2e0de0d","override":null,"pullPolicy":"IfNotPresent","repository":"quay.io/cilium/cilium-envoy","tag":"v1.36.6-1776000132-2437d2edeaf4d9b56ef279bd0d71127440c067aa","useDigest":true}` | Envoy container image. |
 | envoy.initContainers | list | `[]` | Init containers added to the cilium Envoy DaemonSet. |
 | envoy.initialFetchTimeoutSeconds | int | `30` | Time in seconds after which the initial fetch on an xDS stream is considered timed out |
 | envoy.livenessProbe.enabled | bool | `true` | Enable liveness probe for cilium-envoy |
@@ -591,7 +615,7 @@ contributors across the globe, there is almost always someone available to help.
 | hubble.relay.extraVolumes | list | `[]` | Additional hubble-relay volumes. |
 | hubble.relay.gops.enabled | bool | `true` | Enable gops for hubble-relay |
 | hubble.relay.gops.port | int | `9893` | Configure gops listen port for hubble-relay |
-| hubble.relay.image | object | `{"digest":"sha256:d8c4e13bc36a56179292bb52bc6255379cb94cb873700d316ea3139b1bdb8165","override":null,"pullPolicy":"IfNotPresent","repository":"quay.io/cilium/hubble-relay","tag":"v1.19.1","useDigest":true}` | Hubble-relay container image. |
+| hubble.relay.image | object | `{"digest":"sha256:5ee21d57b6ef2aa6db67e603a735fdceb162454b352b7335b651456e308f681b","override":null,"pullPolicy":"IfNotPresent","repository":"quay.io/cilium/hubble-relay","tag":"v1.19.3","useDigest":true}` | Hubble-relay container image. |
 | hubble.relay.listenHost | string | `""` | Host to listen to. Specify an empty string to bind to all the interfaces. |
 | hubble.relay.listenPort | string | `"4245"` | Port to listen to. |
 | hubble.relay.logOptions | object | `{"format":null,"level":null}` | Logging configuration for hubble-relay. |
@@ -709,7 +733,7 @@ contributors across the globe, there is almost always someone available to help.
 | identityAllocationMode | string | `"crd"` | Method to use for identity allocation (`crd`, `kvstore` or `doublewrite-readkvstore` / `doublewrite-readcrd` for migrating between identity backends). |
 | identityChangeGracePeriod | string | `"5s"` | Time to wait before using new identity on endpoint identity change. |
 | identityManagementMode | string | `"agent"` | Control whether CiliumIdentities are created by the agent ("agent"), the operator ("operator") or both ("both"). "Both" should be used only to migrate between "agent" and "operator". Operator-managed identities is a beta feature. |
-| image | object | `{"digest":"sha256:41f1f74a0000de8656f1de4088ea00c8f2d49d6edea579034c73c5fd5fe01792","override":null,"pullPolicy":"IfNotPresent","repository":"quay.io/cilium/cilium","tag":"v1.19.1","useDigest":true}` | Agent container image. |
+| image | object | `{"digest":"sha256:2e61680593cddca8b6c055f6d4c849d87a26a1c91c7e3b8b56c7fb76ab7b7b10","override":null,"pullPolicy":"IfNotPresent","repository":"quay.io/cilium/cilium","tag":"v1.19.3","useDigest":true}` | Agent container image. |
 | imagePullSecrets | list | `[]` | Configure image pull secrets for pulling container images |
 | ingressController.default | bool | `false` | Set cilium ingress controller to be the default ingress controller This will let cilium ingress controller route entries without ingress class set |
 | ingressController.defaultSecretName | string | `nil` | Default secret name for ingresses without .spec.tls[].secretName set. |
@@ -797,12 +821,13 @@ contributors across the globe, there is almost always someone available to help.
 | livenessProbe.failureThreshold | int | `10` | failure threshold of liveness probe |
 | livenessProbe.periodSeconds | int | `30` | interval between checks of the liveness probe |
 | livenessProbe.requireK8sConnectivity | bool | `false` | whether to require k8s connectivity as part of the check. |
-| loadBalancer | object | `{"acceleration":"disabled","l7":{"algorithm":"round_robin","backend":"disabled","ports":[]}}` | Configure service load balancing |
+| loadBalancer | object | `{"acceleration":"disabled","l7":{"algorithm":"round_robin","backend":"disabled","ports":[]},"serviceTopology":false}` | Configure service load balancing |
 | loadBalancer.acceleration | string | `"disabled"` | acceleration is the option to accelerate service handling via XDP Applicable values can be: disabled (do not use XDP), native (XDP BPF program is run directly out of the networking driver's early receive path), or best-effort (use native mode XDP acceleration on devices that support it). |
 | loadBalancer.l7 | object | `{"algorithm":"round_robin","backend":"disabled","ports":[]}` | L7 LoadBalancer |
 | loadBalancer.l7.algorithm | string | `"round_robin"` | Default LB algorithm The default LB algorithm to be used for services, which can be overridden by the service annotation (e.g. service.cilium.io/lb-l7-algorithm) Applicable values: round_robin, least_request, random |
 | loadBalancer.l7.backend | string | `"disabled"` | Enable L7 service load balancing via envoy proxy. The request to a k8s service, which has specific annotation e.g. service.cilium.io/lb-l7, will be forwarded to the local backend proxy to be load balanced to the service endpoints. Please refer to docs for supported annotations for more configuration.  Applicable values:   - envoy: Enable L7 load balancing via envoy proxy. This will automatically set enable-envoy-config as well.   - disabled: Disable L7 load balancing by way of service annotation. |
 | loadBalancer.l7.ports | list | `[]` | List of ports from service to be automatically redirected to above backend. Any service exposing one of these ports will be automatically redirected. Fine-grained control can be achieved by using the service annotation. |
+| loadBalancer.serviceTopology | bool | `false` | serviceTopology enables K8s Topology Aware Hints -based service endpoints filtering |
 | localRedirectPolicies.addressMatcherCIDRs | string | `nil` | Limit the allowed addresses in Address Matcher rule of Local Redirect Policies to the given CIDRs. @schema@ type: [null, array] @schema@ |
 | localRedirectPolicies.enabled | bool | `false` | Enable local redirect policies. |
 | localRedirectPolicy | bool | `false` | Enable Local Redirect Policy (deprecated, please use 'localRedirectPolicies.enabled' instead) |
@@ -860,7 +885,7 @@ contributors across the globe, there is almost always someone available to help.
 | operator.hostNetwork | bool | `true` | HostNetwork setting |
 | operator.identityGCInterval | string | `"15m0s"` | Interval for identity garbage collection. |
 | operator.identityHeartbeatTimeout | string | `"30m0s"` | Timeout for identity heartbeats. |
-| operator.image | object | `{"alibabacloudDigest":"sha256:837b12f4239e88ea5b4b5708ab982c319a94ee05edaecaafe5fd0e5b1962f554","awsDigest":"sha256:18913d05a6c4d205f0b7126c4723bb9ccbd4dc24403da46ed0f9f4bf2a142804","azureDigest":"sha256:82bce78603056e709d4c4e9f9ebb25c222c36d8a07f8c05381c2372d9078eca8","genericDigest":"sha256:e7278d763e448bf6c184b0682cf98cdca078d58a27e1b2f3c906792670aa211a","override":null,"pullPolicy":"IfNotPresent","repository":"quay.io/cilium/operator","suffix":"","tag":"v1.19.1","useDigest":true}` | cilium-operator image. |
+| operator.image | object | `{"alibabacloudDigest":"sha256:176321a65123373ff8c7823b25183102cbad98375e8d6c80b96d68b6e8491103","awsDigest":"sha256:a53dcbfb77282bf2ddd3abbe60f6d49762e7c1389a36cb35b71d504644a56640","azureDigest":"sha256:699c1571a3df1a98882ee13610d47cffb7b34ee7e8d276096db798a5f6c7e4cb","genericDigest":"sha256:205b09b0ed6accbf9fe688d312a9f0fcfc6a316fc081c23fbffb472af5dd62cd","override":null,"pullPolicy":"IfNotPresent","repository":"quay.io/cilium/operator","suffix":"","tag":"v1.19.3","useDigest":true}` | cilium-operator image. |
 | operator.nodeGCInterval | string | `"5m0s"` | Interval for cilium node garbage collection. |
 | operator.nodeSelector | object | `{"kubernetes.io/os":"linux"}` | Node labels for cilium-operator pod assignment ref: https://kubernetes.io/docs/concepts/scheduling-eviction/assign-pod-node/#nodeselector |
 | operator.podAnnotations | object | `{}` | Annotations to be added to cilium-operator pods |
@@ -918,11 +943,11 @@ contributors across the globe, there is almost always someone available to help.
 | preflight.affinity | object | `{"podAffinity":{"requiredDuringSchedulingIgnoredDuringExecution":[{"labelSelector":{"matchLabels":{"k8s-app":"cilium"}},"topologyKey":"kubernetes.io/hostname"}]}}` | Affinity for cilium-preflight |
 | preflight.annotations | object | `{}` | Annotations to be added to all top-level preflight objects (resources under templates/cilium-preflight) |
 | preflight.enabled | bool | `false` | Enable Cilium pre-flight resources (required for upgrade) |
-| preflight.envoy.image | object | `{"digest":"sha256:8188114a2768b5f49d6ce58e168b20d765e0fbc64eee0d83241aa2b150ccd788","override":null,"pullPolicy":"IfNotPresent","repository":"quay.io/cilium/cilium-envoy","tag":"v1.35.9-1770979049-232ed4a26881e4ab4f766f251f258ed424fff663","useDigest":true}` | Envoy pre-flight image. |
+| preflight.envoy.image | object | `{"digest":"sha256:ba0ab8adac082d50d525fd2c5ba096c8facea3a471561b7c61c7a5b9c2e0de0d","override":null,"pullPolicy":"IfNotPresent","repository":"quay.io/cilium/cilium-envoy","tag":"v1.36.6-1776000132-2437d2edeaf4d9b56ef279bd0d71127440c067aa","useDigest":true}` | Envoy pre-flight image. |
 | preflight.extraEnv | list | `[]` | Additional preflight environment variables. |
 | preflight.extraVolumeMounts | list | `[]` | Additional preflight volumeMounts. |
 | preflight.extraVolumes | list | `[]` | Additional preflight volumes. |
-| preflight.image | object | `{"digest":"sha256:41f1f74a0000de8656f1de4088ea00c8f2d49d6edea579034c73c5fd5fe01792","override":null,"pullPolicy":"IfNotPresent","repository":"quay.io/cilium/cilium","tag":"v1.19.1","useDigest":true}` | Cilium pre-flight image. |
+| preflight.image | object | `{"digest":"sha256:2e61680593cddca8b6c055f6d4c849d87a26a1c91c7e3b8b56c7fb76ab7b7b10","override":null,"pullPolicy":"IfNotPresent","repository":"quay.io/cilium/cilium","tag":"v1.19.3","useDigest":true}` | Cilium pre-flight image. |
 | preflight.nodeSelector | object | `{"kubernetes.io/os":"linux"}` | Node labels for preflight pod assignment ref: https://kubernetes.io/docs/concepts/scheduling-eviction/assign-pod-node/#nodeselector |
 | preflight.podAnnotations | object | `{}` | Annotations to be added to preflight pods |
 | preflight.podDisruptionBudget.enabled | bool | `false` | enable PodDisruptionBudget ref: https://kubernetes.io/docs/concepts/workloads/pods/disruptions/ |
@@ -979,6 +1004,7 @@ contributors across the globe, there is almost always someone available to help.
 | serviceAccounts.corednsMCSAPI | object | `{"annotations":{},"automount":true,"create":true,"name":"cilium-coredns-mcsapi-autoconfig"}` | CorednsMCSAPI is used if clustermesh.mcsapi.corednsAutoConfigure.enabled=true |
 | serviceAccounts.hubblecertgen | object | `{"annotations":{},"automount":true,"create":true,"name":"hubble-generate-certs"}` | Hubblecertgen is used if hubble.tls.auto.method=cronJob |
 | serviceAccounts.nodeinit.enabled | bool | `false` | Enabled is temporary until https://github.com/cilium/cilium-cli/issues/1396 is implemented. Cilium CLI doesn't create the SAs for node-init, thus the workaround. Helm is not affected by this issue. Name and automount can be configured, if enabled is set to true. Otherwise, they are ignored. Enabled can be removed once the issue is fixed. Cilium-nodeinit DS must also be fixed. |
+| serviceAccounts.ztunnel | object | `{"annotations":{},"automount":false,"create":true,"name":"ztunnel-cilium"}` | Ztunnel is used if encryption.type=ztunnel |
 | serviceNoBackendResponse | string | `"reject"` | Configure what the response should be to traffic for a service without backends. Possible values:  - reject (default)  - drop |
 | sleepAfterInit | bool | `false` | Do not run Cilium agent when running with clean mode. Useful to completely uninstall Cilium as it will stop Cilium from starting and create artifacts in the node. |
 | socketLB | object | `{"enabled":false}` | Configure socket LB |
diff --git a/packages/system/cilium/charts/cilium/files/cilium-envoy/configmap/bootstrap-config.yaml b/packages/system/cilium/charts/cilium/files/cilium-envoy/configmap/bootstrap-config.yaml
index ea1d3bda..90ebf4ac 100644
--- a/packages/system/cilium/charts/cilium/files/cilium-envoy/configmap/bootstrap-config.yaml
+++ b/packages/system/cilium/charts/cilium/files/cilium-envoy/configmap/bootstrap-config.yaml
@@ -167,6 +167,8 @@ staticResources:
     circuitBreakers:
       thresholds:
       - maxRetries: {{ .Values.envoy.maxConcurrentRetries }}
+        maxConnections: {{ .Values.envoy.clusterMaxConnections }}
+        maxRequests: {{ .Values.envoy.clusterMaxRequests }}
     lbPolicy: "CLUSTER_PROVIDED"
     typedExtensionProtocolOptions:
       envoy.extensions.upstreams.http.v3.HttpProtocolOptions:
@@ -183,6 +185,8 @@ staticResources:
     circuitBreakers:
       thresholds:
       - maxRetries: {{ .Values.envoy.maxConcurrentRetries }}
+        maxConnections: {{ .Values.envoy.clusterMaxConnections }}
+        maxRequests: {{ .Values.envoy.clusterMaxRequests }}
     lbPolicy: "CLUSTER_PROVIDED"
     typedExtensionProtocolOptions:
       envoy.extensions.upstreams.http.v3.HttpProtocolOptions:
@@ -204,6 +208,8 @@ staticResources:
     circuitBreakers:
       thresholds:
       - maxRetries: {{ .Values.envoy.maxConcurrentRetries }}
+        maxConnections: {{ .Values.envoy.clusterMaxConnections }}
+        maxRequests: {{ .Values.envoy.clusterMaxRequests }}
     lbPolicy: "CLUSTER_PROVIDED"
     typedExtensionProtocolOptions:
       envoy.extensions.upstreams.http.v3.HttpProtocolOptions:
@@ -220,6 +226,8 @@ staticResources:
     circuitBreakers:
       thresholds:
       - maxRetries: {{ .Values.envoy.maxConcurrentRetries }}
+        maxConnections: {{ .Values.envoy.clusterMaxConnections }}
+        maxRequests: {{ .Values.envoy.clusterMaxRequests }}
     lbPolicy: "CLUSTER_PROVIDED"
     typedExtensionProtocolOptions:
       envoy.extensions.upstreams.http.v3.HttpProtocolOptions:
@@ -304,3 +312,4 @@ admin:
   address:
     pipe:
       path: "/var/run/cilium/envoy/sockets/admin.sock"
+      mode: 0660
diff --git a/packages/system/cilium/charts/cilium/templates/cilium-agent/daemonset.yaml b/packages/system/cilium/charts/cilium/templates/cilium-agent/daemonset.yaml
index f9a4cec7..fff1b384 100644
--- a/packages/system/cilium/charts/cilium/templates/cilium-agent/daemonset.yaml
+++ b/packages/system/cilium/charts/cilium/templates/cilium-agent/daemonset.yaml
@@ -57,19 +57,6 @@ spec:
         # ensure pods roll when configmap updates
         cilium.io/cilium-configmap-checksum: {{ include (print $.Template.BasePath "/cilium-configmap.yaml") . | sha256sum | quote }}
         {{- end }}
-        {{- if not .Values.securityContext.privileged }}
-        {{- if semverCompare "<1.30.0" (printf "%d.%d.0" (semver .Capabilities.KubeVersion.Version).Major (semver .Capabilities.KubeVersion.Version).Minor) }}
-        # Set app AppArmor's profile to "unconfined". The value of this annotation
-        # can be modified as long users know which profiles they have available
-        # in AppArmor.
-        container.apparmor.security.beta.kubernetes.io/cilium-agent: "unconfined"
-        container.apparmor.security.beta.kubernetes.io/clean-cilium-state: "unconfined"
-        {{- if .Values.cgroup.autoMount.enabled }}
-        container.apparmor.security.beta.kubernetes.io/mount-cgroup: "unconfined"
-        container.apparmor.security.beta.kubernetes.io/apply-sysctl-overwrites: "unconfined"
-        {{- end }}
-        {{- end }}
-        {{- end }}
         {{- with .Values.podAnnotations }}
         {{- toYaml . | nindent 8 }}
         {{- end }}
@@ -559,7 +546,7 @@ spec:
           {{- toYaml .Values.initResources | trim | nindent 10 }}
         {{- end }}
         command:
-        - sh
+        - bash
         - -ec
         # The statically linked Go program binary is invoked to avoid any
         # dependency on utilities like sh and mount that can be missing on certain
@@ -605,7 +592,7 @@ spec:
         - name: BIN_PATH
           value: {{ .Values.cni.binPath }}
         command:
-        - sh
+        - bash
         - -ec
         # The statically linked Go program binary is invoked to avoid any
         # dependency on utilities like sh that can be missing on certain
@@ -673,7 +660,7 @@ spec:
           {{- toYaml . | trim | nindent 10 }}
         {{- end }}
         command:
-        - sh
+        - bash
         - -c
         - |
           until test -s {{ (print "/tmp/cilium-bootstrap.d/" (.Values.nodeinit.bootstrapFile | base)) | quote }}; do
diff --git a/packages/system/cilium/charts/cilium/templates/cilium-configmap.yaml b/packages/system/cilium/charts/cilium/templates/cilium-configmap.yaml
index a3a38c09..5d76944d 100644
--- a/packages/system/cilium/charts/cilium/templates/cilium-configmap.yaml
+++ b/packages/system/cilium/charts/cilium/templates/cilium-configmap.yaml
@@ -464,6 +464,9 @@ data:
 {{- if has (kindOf .Values.bpf.policyMapPressureMetricsThreshold) (list "int64" "float64") }}
   bpf-policy-map-pressure-metrics-threshold: {{ .Values.bpf.policyMapPressureMetricsThreshold | quote }}
 {{- end }}
+{{- if .Values.endpointPolicyUpdateTimeoutDuration }}
+  endpoint-policy-update-timeout: {{ .Values.endpointPolicyUpdateTimeoutDuration | quote }}
+{{- end }}
 {{- if hasKey .Values.bpf "policyStatsMapMax" }}
   # bpf-policy-stats-map-max specifies the maximum number of entries in global
   # policy stats map
@@ -706,7 +709,6 @@ data:
   enable-ipv4-big-tcp: {{ .Values.enableIPv4BIGTCP | quote }}
   enable-ipv6-big-tcp: {{ .Values.enableIPv6BIGTCP | quote }}
   enable-ipv6-masquerade: {{ .Values.enableIPv6Masquerade | quote }}
-  enable-tunnel-big-tcp: {{ .Values.enableTunnelBIGTCP | quote }}
 
 {{- if hasKey .Values.bpf "enableTCX" }}
   enable-tcx: {{ .Values.bpf.enableTCX | quote }}
@@ -906,9 +908,9 @@ data:
 {{- end }}
 {{- if hasKey .Values.loadBalancer "serviceTopology" }}
   enable-service-topology: {{ .Values.loadBalancer.serviceTopology | quote }}
-# {{- end }}
-
 {{- end }}
+{{- end }}
+
 {{- if hasKey .Values.maglev "tableSize" }}
   bpf-lb-maglev-table-size: {{ .Values.maglev.tableSize | quote}}
 {{- end }}
@@ -1380,11 +1382,7 @@ data:
 
 {{- if .Values.operator.unmanagedPodWatcher.restart }}
   {{- $interval := .Values.operator.unmanagedPodWatcher.intervalSeconds }}
-  {{- if kindIs "float64" $interval }}
   unmanaged-pod-watcher-interval: {{ printf "%ds" (int $interval) | quote }}
-  {{- else }}
-  unmanaged-pod-watcher-interval: {{ $interval | quote }}
-  {{- end }}
 {{- else }}
   unmanaged-pod-watcher-interval: "0"
 {{- end }}
@@ -1517,6 +1515,14 @@ data:
   connectivity-probe-frequency-ratio: {{ .Values.connectivityProbeFrequencyRatio | quote }}
 {{- end }}
 
+{{- if hasKey .Values "configDriftDetection" }}
+  enable-dynamic-config: {{ .Values.configDriftDetection.enabled | quote }}
+  enable-drift-checker: {{ .Values.configDriftDetection.driftChecker | quote }}
+  {{- if .Values.configDriftDetection.ignoredKeys }}
+  ignore-flags-drift-checker: {{ join "," .Values.configDriftDetection.ignoredKeys | quote }}
+  {{- end }}
+{{- end }}
+
 # Extra config allows adding arbitrary properties to the cilium config.
 # By putting it at the end of the ConfigMap, it's also possible to override existing properties.
 {{- if .Values.extraConfig }}
diff --git a/packages/system/cilium/charts/cilium/templates/cilium-operator/clusterrole.yaml b/packages/system/cilium/charts/cilium/templates/cilium-operator/clusterrole.yaml
index 96f72c7f..c147e2b5 100644
--- a/packages/system/cilium/charts/cilium/templates/cilium-operator/clusterrole.yaml
+++ b/packages/system/cilium/charts/cilium/templates/cilium-operator/clusterrole.yaml
@@ -407,6 +407,15 @@ rules:
   verbs:
   - update
   - patch
+- apiGroups:
+  - multicluster.x-k8s.io
+  resources:
+  # The controller needs to be able to set serviceimport finalizers to be able to create a derived Service
+  # resource that is owned by the ServiceImport and sets blockOwnerDeletion=true in its ownerRef.
+  # This is required when the admission plugin OwnerReferencesPermissionEnforcement is activated.
+  - serviceimports/finalizers
+  verbs:
+  - update
 - apiGroups:
   - multicluster.x-k8s.io
   resources:
diff --git a/packages/system/cilium/charts/cilium/templates/ztunnel/daemonset.yaml b/packages/system/cilium/charts/cilium/templates/ztunnel/daemonset.yaml
new file mode 100644
index 00000000..f9363068
--- /dev/null
+++ b/packages/system/cilium/charts/cilium/templates/ztunnel/daemonset.yaml
@@ -0,0 +1,165 @@
+{{- if and .Values.encryption.enabled (eq .Values.encryption.type "ztunnel") }}
+---
+kind: DaemonSet
+apiVersion: apps/v1
+metadata:
+  name: ztunnel-cilium
+  namespace: {{ include "cilium.namespace" . }}
+  {{- with .Values.encryption.ztunnel.annotations }}
+  annotations:
+    {{- toYaml . | nindent 4 }}
+  {{- end }}
+  labels:
+    app: ztunnel-cilium
+    app.kubernetes.io/part-of: cilium
+    app.kubernetes.io/name: ztunnel-cilium
+    {{- with .Values.commonLabels }}
+    {{- toYaml . | nindent 4 }}
+    {{- end }}
+spec:
+  revisionHistoryLimit: 10
+  selector:
+    matchLabels:
+      app: ztunnel-cilium
+  {{- with .Values.encryption.ztunnel.updateStrategy }}
+  updateStrategy:
+    {{- toYaml . | trim | nindent 4 }}
+  {{- end }}
+  template:
+    metadata:
+      annotations:
+        sidecar.istio.io/inject: "false"
+        {{- with .Values.encryption.ztunnel.podAnnotations }}
+        {{- toYaml . | nindent 8 }}
+        {{- end }}
+      labels:
+        app: ztunnel-cilium
+        app.kubernetes.io/part-of: cilium
+        app.kubernetes.io/name: ztunnel-cilium
+        {{- with .Values.commonLabels }}
+        {{- toYaml . | nindent 8 }}
+        {{- end }}
+        {{- with .Values.encryption.ztunnel.podLabels }}
+        {{- toYaml . | nindent 8 }}
+        {{- end }}
+    spec:
+      {{- with .Values.imagePullSecrets }}
+      imagePullSecrets:
+        {{- toYaml . | nindent 8 }}
+      {{- end }}
+      hostNetwork: true
+      dnsPolicy: ClusterFirst
+      containers:
+        - name: istio-proxy
+          image: {{ include "cilium.image" .Values.encryption.ztunnel.image | quote }}
+          imagePullPolicy: {{ .Values.encryption.ztunnel.image.pullPolicy }}
+          args:
+            - proxy
+            - ztunnel
+          env:
+            - name: XDS_ADDRESS
+              value: "https://localhost:15012"
+            - name: XDS_ROOT_CA
+              value: "/etc/ztunnel/bootstrap-root.crt"
+            - name: CA_ROOT_CA
+              value: "/etc/ztunnel/bootstrap-root.crt"
+            - name: CA_ADDRESS
+              value: {{ .Values.encryption.ztunnel.caAddress | quote }}
+            - name: ISTIO_META_DNS_CAPTURE
+              value: "false"
+            - name: INPOD_UDS
+              value: "/var/run/cilium/ztunnel.sock"
+            - name: NODE_NAME
+              valueFrom:
+                fieldRef:
+                  apiVersion: v1
+                  fieldPath: status.hostIP
+            - name: INSTANCE_IP
+              valueFrom:
+                fieldRef:
+                  apiVersion: v1
+                  fieldPath: status.podIP
+            - name: ISTIO_META_ENABLE_HBONE
+              value: "true"
+            {{- with .Values.encryption.ztunnel.extraEnv }}
+            {{- toYaml . | trim | nindent 12 }}
+            {{- end }}
+          readinessProbe:
+            httpGet:
+              path: /healthz/ready
+              port: {{ .Values.encryption.ztunnel.healthPort }}
+              host: "127.0.0.1"
+              scheme: HTTP
+            initialDelaySeconds: {{ .Values.encryption.ztunnel.readinessProbe.initialDelaySeconds }}
+            periodSeconds: {{ .Values.encryption.ztunnel.readinessProbe.periodSeconds }}
+            failureThreshold: {{ .Values.encryption.ztunnel.readinessProbe.failureThreshold }}
+            successThreshold: 1
+            timeoutSeconds: 1
+          {{- with .Values.encryption.ztunnel.resources }}
+          resources:
+            {{- toYaml . | trim | nindent 12 }}
+          {{- end }}
+          securityContext:
+            allowPrivilegeEscalation: true
+            capabilities:
+              add:
+                - NET_ADMIN
+                - SYS_ADMIN
+                - NET_RAW
+              drop:
+                - ALL
+            privileged: false
+            readOnlyRootFilesystem: true
+            runAsGroup: 1337
+            runAsNonRoot: false
+            runAsUser: 0
+          terminationMessagePath: /dev/termination-log
+          terminationMessagePolicy: File
+          volumeMounts:
+            - mountPath: /var/run/cilium
+              name: cilium-dir
+              readOnly: false
+            - mountPath: /etc/ztunnel
+              name: cilium-ztunnel-secrets
+              readOnly: true
+            {{- with .Values.encryption.ztunnel.extraVolumeMounts }}
+            {{- toYaml . | nindent 12 }}
+            {{- end }}
+      {{- with .Values.encryption.ztunnel.affinity }}
+      affinity:
+        {{- toYaml . | nindent 8 }}
+      {{- end }}
+      {{- with .Values.encryption.ztunnel.nodeSelector }}
+      nodeSelector:
+        {{- toYaml . | trim | nindent 8 }}
+      {{- end }}
+      {{- with .Values.encryption.ztunnel.tolerations }}
+      tolerations:
+        {{- toYaml . | trim | nindent 8 }}
+      {{- end }}
+      priorityClassName: {{ include "cilium.priorityClass" (list $ .Values.encryption.ztunnel.priorityClassName "system-node-critical") }}
+      restartPolicy: Always
+      terminationGracePeriodSeconds: {{ .Values.encryption.ztunnel.terminationGracePeriodSeconds }}
+      {{- if .Values.serviceAccounts.ztunnel.create }}
+      serviceAccountName: {{ .Values.serviceAccounts.ztunnel.name | quote }}
+      automountServiceAccountToken: {{ .Values.serviceAccounts.ztunnel.automount }}
+      {{- else }}
+      automountServiceAccountToken: false
+      {{- end }}
+      volumes:
+        - name: cilium-dir
+          hostPath:
+            path: /var/run/cilium
+            type: DirectoryOrCreate
+        - name: cilium-ztunnel-secrets
+          secret:
+            secretName: cilium-ztunnel-secrets
+            defaultMode: 420
+            items:
+              - key: bootstrap-root.crt
+                path: bootstrap-root.crt
+                mode: 420
+        {{- with .Values.encryption.ztunnel.extraVolumes }}
+        {{- toYaml . | nindent 8 }}
+        {{- end }}
+{{- end }}
diff --git a/packages/system/cilium/charts/cilium/templates/ztunnel/secret.yaml b/packages/system/cilium/charts/cilium/templates/ztunnel/secret.yaml
new file mode 100644
index 00000000..520857dc
--- /dev/null
+++ b/packages/system/cilium/charts/cilium/templates/ztunnel/secret.yaml
@@ -0,0 +1,23 @@
+{{- if and .Values.encryption.enabled (eq .Values.encryption.type "ztunnel") }}
+{{- if .Values.encryption.ztunnel.secrets.bootstrapRootCert }}
+---
+apiVersion: v1
+kind: Secret
+metadata:
+  name: cilium-ztunnel-secrets
+  namespace: {{ include "cilium.namespace" . }}
+  {{- with .Values.encryption.ztunnel.annotations }}
+  annotations:
+    {{- toYaml . | nindent 4 }}
+  {{- end }}
+  labels:
+    app.kubernetes.io/part-of: cilium
+    app.kubernetes.io/name: ztunnel-cilium
+    {{- with .Values.commonLabels }}
+    {{- toYaml . | nindent 4 }}
+    {{- end }}
+type: Opaque
+data:
+  bootstrap-root.crt: {{ .Values.encryption.ztunnel.secrets.bootstrapRootCert | b64enc }}
+{{- end }}
+{{- end }}
diff --git a/packages/system/cilium/charts/cilium/templates/ztunnel/serviceaccount.yaml b/packages/system/cilium/charts/cilium/templates/ztunnel/serviceaccount.yaml
new file mode 100644
index 00000000..4ef9e2bc
--- /dev/null
+++ b/packages/system/cilium/charts/cilium/templates/ztunnel/serviceaccount.yaml
@@ -0,0 +1,21 @@
+{{- if and .Values.encryption.enabled (eq .Values.encryption.type "ztunnel") .Values.serviceAccounts.ztunnel.create }}
+---
+apiVersion: v1
+kind: ServiceAccount
+metadata:
+  name: {{ .Values.serviceAccounts.ztunnel.name | quote }}
+  namespace: {{ include "cilium.namespace" . }}
+  {{- with .Values.commonLabels }}
+  labels:
+    {{- toYaml . | nindent 4 }}
+  {{- end }}
+  {{- if or .Values.serviceAccounts.ztunnel.annotations .Values.encryption.ztunnel.annotations }}
+  annotations:
+    {{- with .Values.encryption.ztunnel.annotations }}
+      {{- toYaml . | nindent 4 }}
+    {{- end }}
+    {{- with .Values.serviceAccounts.ztunnel.annotations }}
+      {{- toYaml . | nindent 4 }}
+    {{- end }}
+  {{- end }}
+{{- end }}
diff --git a/packages/system/cilium/charts/cilium/values.schema.json b/packages/system/cilium/charts/cilium/values.schema.json
index d18566d7..65c84af5 100644
--- a/packages/system/cilium/charts/cilium/values.schema.json
+++ b/packages/system/cilium/charts/cilium/values.schema.json
@@ -1714,6 +1714,21 @@
         "object"
       ]
     },
+    "configDriftDetection": {
+      "properties": {
+        "driftChecker": {
+          "type": "boolean"
+        },
+        "enabled": {
+          "type": "boolean"
+        },
+        "ignoredKeys": {
+          "items": {},
+          "type": "array"
+        }
+      },
+      "type": "object"
+    },
     "connectivityProbeFrequencyRatio": {
       "type": [
         "null",
@@ -1898,9 +1913,6 @@
     "enableNonDefaultDenyPolicies": {
       "type": "boolean"
     },
-    "enableTunnelBIGTCP": {
-      "type": "boolean"
-    },
     "enableXTSocketFallback": {
       "type": "boolean"
     },
@@ -1984,6 +1996,184 @@
             }
           },
           "type": "object"
+        },
+        "ztunnel": {
+          "properties": {
+            "affinity": {
+              "type": "object"
+            },
+            "annotations": {
+              "type": "object"
+            },
+            "caAddress": {
+              "type": "string"
+            },
+            "extraEnv": {
+              "items": {},
+              "type": "array"
+            },
+            "extraVolumeMounts": {
+              "items": {},
+              "type": "array"
+            },
+            "extraVolumes": {
+              "items": {},
+              "type": "array"
+            },
+            "healthPort": {
+              "type": "integer"
+            },
+            "image": {
+              "properties": {
+                "digest": {
+                  "type": [
+                    "null",
+                    "string"
+                  ]
+                },
+                "override": {
+                  "type": [
+                    "null",
+                    "string"
+                  ]
+                },
+                "pullPolicy": {
+                  "type": "string"
+                },
+                "repository": {
+                  "type": "string"
+                },
+                "tag": {
+                  "type": "string"
+                },
+                "useDigest": {
+                  "type": "boolean"
+                }
+              },
+              "type": "object"
+            },
+            "nodeSelector": {
+              "properties": {
+                "kubernetes.io/os": {
+                  "type": "string"
+                }
+              },
+              "type": "object"
+            },
+            "podAnnotations": {
+              "type": "object"
+            },
+            "podLabels": {
+              "type": "object"
+            },
+            "priorityClassName": {
+              "type": [
+                "null",
+                "string"
+              ]
+            },
+            "readinessProbe": {
+              "properties": {
+                "failureThreshold": {
+                  "type": "integer"
+                },
+                "initialDelaySeconds": {
+                  "type": "integer"
+                },
+                "periodSeconds": {
+                  "type": "integer"
+                }
+              },
+              "type": "object"
+            },
+            "resources": {
+              "properties": {
+                "requests": {
+                  "properties": {
+                    "cpu": {
+                      "type": "string"
+                    },
+                    "memory": {
+                      "type": "string"
+                    }
+                  },
+                  "type": "object"
+                }
+              },
+              "type": "object"
+            },
+            "secrets": {
+              "properties": {
+                "bootstrapRootCert": {
+                  "type": [
+                    "null",
+                    "string"
+                  ]
+                }
+              },
+              "type": "object"
+            },
+            "terminationGracePeriodSeconds": {
+              "type": "integer"
+            },
+            "tolerations": {
+              "items": {
+                "anyOf": [
+                  {
+                    "properties": {
+                      "effect": {
+                        "type": "string"
+                      },
+                      "operator": {
+                        "type": "string"
+                      }
+                    }
+                  },
+                  {
+                    "properties": {
+                      "key": {
+                        "type": "string"
+                      },
+                      "operator": {
+                        "type": "string"
+                      }
+                    }
+                  },
+                  {
+                    "properties": {
+                      "effect": {
+                        "type": "string"
+                      },
+                      "operator": {
+                        "type": "string"
+                      }
+                    }
+                  }
+                ]
+              },
+              "type": "array"
+            },
+            "updateStrategy": {
+              "properties": {
+                "rollingUpdate": {
+                  "properties": {
+                    "maxSurge": {
+                      "type": "integer"
+                    },
+                    "maxUnavailable": {
+                      "type": "integer"
+                    }
+                  },
+                  "type": "object"
+                },
+                "type": {
+                  "type": "string"
+                }
+              },
+              "type": "object"
+            }
+          },
+          "type": "object"
         }
       },
       "type": "object"
@@ -1999,6 +2189,9 @@
     "endpointLockdownOnMapOverflow": {
       "type": "boolean"
     },
+    "endpointPolicyUpdateTimeoutDuration": {
+      "type": "null"
+    },
     "endpointRoutes": {
       "properties": {
         "enabled": {
@@ -4484,6 +4677,9 @@
             }
           },
           "type": "object"
+        },
+        "serviceTopology": {
+          "type": "boolean"
         }
       },
       "type": "object"
@@ -6029,6 +6225,23 @@
             }
           },
           "type": "object"
+        },
+        "ztunnel": {
+          "properties": {
+            "annotations": {
+              "type": "object"
+            },
+            "automount": {
+              "type": "boolean"
+            },
+            "create": {
+              "type": "boolean"
+            },
+            "name": {
+              "type": "string"
+            }
+          },
+          "type": "object"
         }
       },
       "type": "object"
diff --git a/packages/system/cilium/charts/cilium/values.yaml b/packages/system/cilium/charts/cilium/values.yaml
index b9b30830..66f76982 100644
--- a/packages/system/cilium/charts/cilium/values.yaml
+++ b/packages/system/cilium/charts/cilium/values.yaml
@@ -210,6 +210,12 @@ serviceAccounts:
     name: cilium-coredns-mcsapi-autoconfig
     automount: true
     annotations: {}
+  # -- Ztunnel is used if encryption.type=ztunnel
+  ztunnel:
+    create: true
+    name: ztunnel-cilium
+    automount: false
+    annotations: {}
 # -- Configure termination grace period for cilium-agent DaemonSet.
 terminationGracePeriodSeconds: 1
 # -- Install the cilium agent resources.
@@ -218,6 +224,22 @@ agent: true
 name: cilium
 # -- Roll out cilium agent pods automatically when configmap is updated.
 rollOutCiliumPods: false
+# -- Configuration for the ConfigMap drift detection feature.
+# When enabled, the agent continuously watches the cilium-config ConfigMap
+# and exposes a cilium_drift_checker_config_delta Prometheus metric reporting
+# the number of keys that differ between the ConfigMap and the agent's active
+# settings. A non-zero value indicates that the agent has not yet applied all
+# current ConfigMap changes and needs to be restarted.
+configDriftDetection:
+  # -- Enable watching of the cilium-config ConfigMap and reflecting its
+  # contents into the agent's internal DynamicConfig table.
+  enabled: true
+  # -- Enable the drift checker which compares the DynamicConfig table against
+  # the agent's active settings and publishes the
+  # cilium_drift_checker_config_delta metric.
+  driftChecker: true
+  # -- List of config-map keys to ignore when computing the drift delta.
+  ignoredKeys: []
 # -- Agent container image.
 image:
   # @schema
@@ -225,10 +247,10 @@ image:
   # @schema
   override: ~
   repository: "quay.io/cilium/cilium"
-  tag: "v1.19.1"
+  tag: "v1.19.3"
   pullPolicy: "IfNotPresent"
   # cilium-digest
-  digest: sha256:41f1f74a0000de8656f1de4088ea00c8f2d49d6edea579034c73c5fd5fe01792
+  digest: sha256:2e61680593cddca8b6c055f6d4c849d87a26a1c91c7e3b8b56c7fb76ab7b7b10
   useDigest: true
 # -- Scheduling configurations for cilium pods
 scheduling:
@@ -1133,9 +1155,89 @@ encryption:
   wireguard:
     # -- Controls WireGuard PersistentKeepalive option. Set 0s to disable.
     persistentKeepalive: 0s
+  # -- ztunnel encryption configuration.
+  # ztunnel is Istio's purpose-built, per-node proxy for handling L4 traffic in ambient mesh mode.
+  # These settings only apply when encryption.type is set to "ztunnel".
+  ztunnel:
+    # -- ztunnel container image.
+    image:
+      # @schema
+      # type: [null, string]
+      # @schema
+      override: ~
+      repository: "docker.io/istio/ztunnel"
+      tag: "1.28.0-distroless"
+      pullPolicy: "IfNotPresent"
+      # @schema
+      # type: [null, string]
+      # @schema
+      digest: ~
+      useDigest: false
+    # -- CA server address for certificate requests.
+    caAddress: "https://localhost:15012"
+    # -- TCP port for the health API.
+    healthPort: 15021
+    # -- ztunnel resource limits & requests.
+    resources:
+      requests:
+        cpu: 200m
+        memory: 512Mi
+    # -- ztunnel update strategy.
+    updateStrategy:
+      type: RollingUpdate
+      rollingUpdate:
+        maxSurge: 1
+        maxUnavailable: 0
+    # -- Configure termination grace period for ztunnel DaemonSet.
+    terminationGracePeriodSeconds: 30
+    # -- Readiness probe configuration.
+    readinessProbe:
+      initialDelaySeconds: 0
+      periodSeconds: 10
+      failureThreshold: 3
+    # -- Node selector for ztunnel pods.
+    nodeSelector:
+      kubernetes.io/os: linux
+    # -- Node tolerations for ztunnel scheduling.
+    tolerations:
+      - effect: NoSchedule
+        operator: Exists
+      - key: CriticalAddonsOnly
+        operator: Exists
+      - effect: NoExecute
+        operator: Exists
+    # -- Affinity for ztunnel pods.
+    affinity: {}
+    # @schema
+    # type: [null, string]
+    # @schema
+    # -- The priority class to use for ztunnel pods.
+    priorityClassName: ~
+    # -- Annotations to be added to all ztunnel resources.
+    annotations: {}
+    # -- Annotations to be added to ztunnel pods.
+    podAnnotations: {}
+    # -- Labels to be added to ztunnel pods.
+    podLabels: {}
+    # -- Additional ztunnel container environment variables.
+    extraEnv: []
+    # -- Additional ztunnel volumes.
+    extraVolumes: []
+    # -- Additional ztunnel volumeMounts.
+    extraVolumeMounts: []
+    # -- ztunnel secrets configuration.
+    secrets:
+      # -- Base64-encoded bootstrap root certificate content.
+      # If not provided, the secret must be created manually before deploying.
+      # @schema
+      # type: [null, string]
+      # @schema
+      bootstrapRootCert: ~
 endpointHealthChecking:
   # -- Enable connectivity health checking between virtual endpoints.
   enabled: true
+# -- Max duration to wait for envoy to respond to configuration changes. Default "10s".
+endpointPolicyUpdateTimeoutDuration: null
 endpointRoutes:
   # @schema
   # type: [boolean, string]
@@ -1250,8 +1352,8 @@ certgen:
     # @schema
     override: ~
     repository: "quay.io/cilium/certgen"
-    tag: "v0.3.2"
-    digest: "sha256:19921f48ee7e2295ea4dca955878a6cd8d70e6d4219d08f688e866ece9d95d4d"
+    tag: "v0.4.1"
+    digest: "sha256:f0c656830e856d26b24b0e144df1f8b327d3b46748d76a630514111fc365b697"
     useDigest: true
     pullPolicy: "IfNotPresent"
   # @schema
@@ -1599,9 +1701,9 @@ hubble:
       # @schema
       override: ~
       repository: "quay.io/cilium/hubble-relay"
-      tag: "v1.19.1"
+      tag: "v1.19.3"
       # hubble-relay-digest
-      digest: sha256:d8c4e13bc36a56179292bb52bc6255379cb94cb873700d316ea3139b1bdb8165
+      digest: sha256:5ee21d57b6ef2aa6db67e603a735fdceb162454b352b7335b651456e308f681b
       useDigest: true
       pullPolicy: "IfNotPresent"
     # -- Specifies the resources for the hubble-relay pods
@@ -2313,8 +2415,6 @@ enableMasqueradeRouteSource: false
 enableIPv4BIGTCP: false
 # -- Enables IPv6 BIG TCP support which increases maximum IPv6 GSO/GRO limits for nodes and pods
 enableIPv6BIGTCP: false
-# -- Enable BIG TCP in tunneling mode and increase maximum GRO/GSO limits for VXLAN/GENEVE tunnels
-enableTunnelBIGTCP: false
 nat:
   # -- Number of the top-k SNAT map connections to track in Cilium statedb.
   mapStatsEntries: 32
@@ -2392,8 +2492,7 @@ loadBalancer:
 
   # -- serviceTopology enables K8s Topology Aware Hints -based service
   # endpoints filtering
-  # serviceTopology: false
-
+  serviceTopology: false
   # -- L7 LoadBalancer
   l7:
     # -- Enable L7 service load balancing via envoy proxy.
@@ -2626,9 +2725,9 @@ envoy:
     # @schema
     override: ~
     repository: "quay.io/cilium/cilium-envoy"
-    tag: "v1.35.9-1770979049-232ed4a26881e4ab4f766f251f258ed424fff663"
+    tag: "v1.36.6-1776000132-2437d2edeaf4d9b56ef279bd0d71127440c067aa"
     pullPolicy: "IfNotPresent"
-    digest: "sha256:8188114a2768b5f49d6ce58e168b20d765e0fbc64eee0d83241aa2b150ccd788"
+    digest: "sha256:ba0ab8adac082d50d525fd2c5ba096c8facea3a471561b7c61c7a5b9c2e0de0d"
     useDigest: true
   # -- Init containers added to the cilium Envoy DaemonSet.
   initContainers: []
@@ -3011,15 +3110,15 @@ operator:
     # @schema
     override: ~
     repository: "quay.io/cilium/operator"
-    tag: "v1.19.1"
+    tag: "v1.19.3"
     # operator-generic-digest
-    genericDigest: sha256:e7278d763e448bf6c184b0682cf98cdca078d58a27e1b2f3c906792670aa211a
+    genericDigest: sha256:205b09b0ed6accbf9fe688d312a9f0fcfc6a316fc081c23fbffb472af5dd62cd
     # operator-azure-digest
-    azureDigest: sha256:82bce78603056e709d4c4e9f9ebb25c222c36d8a07f8c05381c2372d9078eca8
+    azureDigest: sha256:699c1571a3df1a98882ee13610d47cffb7b34ee7e8d276096db798a5f6c7e4cb
     # operator-aws-digest
-    awsDigest: sha256:18913d05a6c4d205f0b7126c4723bb9ccbd4dc24403da46ed0f9f4bf2a142804
+    awsDigest: sha256:a53dcbfb77282bf2ddd3abbe60f6d49762e7c1389a36cb35b71d504644a56640
     # operator-alibabacloud-digest
-    alibabacloudDigest: sha256:837b12f4239e88ea5b4b5708ab982c319a94ee05edaecaafe5fd0e5b1962f554
+    alibabacloudDigest: sha256:176321a65123373ff8c7823b25183102cbad98375e8d6c80b96d68b6e8491103
     useDigest: true
     pullPolicy: "IfNotPresent"
     suffix: ""
@@ -3344,9 +3443,9 @@ preflight:
     # @schema
     override: ~
     repository: "quay.io/cilium/cilium"
-    tag: "v1.19.1"
+    tag: "v1.19.3"
     # cilium-digest
-    digest: sha256:41f1f74a0000de8656f1de4088ea00c8f2d49d6edea579034c73c5fd5fe01792
+    digest: sha256:2e61680593cddca8b6c055f6d4c849d87a26a1c91c7e3b8b56c7fb76ab7b7b10
     useDigest: true
     pullPolicy: "IfNotPresent"
   envoy:
@@ -3357,9 +3456,9 @@ preflight:
       # @schema
       override: ~
       repository: "quay.io/cilium/cilium-envoy"
-      tag: "v1.35.9-1770979049-232ed4a26881e4ab4f766f251f258ed424fff663"
+      tag: "v1.36.6-1776000132-2437d2edeaf4d9b56ef279bd0d71127440c067aa"
       pullPolicy: "IfNotPresent"
-      digest: "sha256:8188114a2768b5f49d6ce58e168b20d765e0fbc64eee0d83241aa2b150ccd788"
+      digest: "sha256:ba0ab8adac082d50d525fd2c5ba096c8facea3a471561b7c61c7a5b9c2e0de0d"
       useDigest: true
   # -- The priority class to use for the preflight pod.
   priorityClassName: ""
@@ -3603,9 +3702,9 @@ clustermesh:
       # @schema
       override: ~
       repository: "quay.io/cilium/clustermesh-apiserver"
-      tag: "v1.19.1"
+      tag: "v1.19.3"
       # clustermesh-apiserver-digest
-      digest: sha256:56d6c3dc13b50126b80ecb571707a0ea97f6db694182b9d61efd386d04e5bb28
+      digest: sha256:a8136a7615d6c6041d3aa6f2674d17beaec238170d669507ccc05328a778e2b7
       useDigest: true
       pullPolicy: "IfNotPresent"
     # -- TCP port for the clustermesh-apiserver health API.
@@ -4140,7 +4239,7 @@ authentication:
           override: ~
           repository: "docker.io/library/busybox"
           tag: "1.37.0"
-          digest: "sha256:b3255e7dfbcd10cb367af0d409747d511aeb66dfac98cf30e97e87e4207dd76f"
+          digest: "sha256:1487d0af5f52b4ba31c7e465126ee2123fe3f2305d638e7827681e7cf6c83d5e"
           useDigest: true
           pullPolicy: "IfNotPresent"
         # SPIRE agent configuration
diff --git a/packages/system/cilium/charts/cilium/values.yaml.tmpl b/packages/system/cilium/charts/cilium/values.yaml.tmpl
index c21f1c26..8039b40c 100644
--- a/packages/system/cilium/charts/cilium/values.yaml.tmpl
+++ b/packages/system/cilium/charts/cilium/values.yaml.tmpl
@@ -213,6 +213,12 @@ serviceAccounts:
     name: cilium-coredns-mcsapi-autoconfig
     automount: true
     annotations: {}
+  # -- Ztunnel is used if encryption.type=ztunnel
+  ztunnel:
+    create: true
+    name: ztunnel-cilium
+    automount: false
+    annotations: {}
 # -- Configure termination grace period for cilium-agent DaemonSet.
 terminationGracePeriodSeconds: 1
 # -- Install the cilium agent resources.
@@ -221,6 +227,22 @@ agent: true
 name: cilium
 # -- Roll out cilium agent pods automatically when configmap is updated.
 rollOutCiliumPods: false
+# -- Configuration for the ConfigMap drift detection feature.
+# When enabled, the agent continuously watches the cilium-config ConfigMap
+# and exposes a cilium_drift_checker_config_delta Prometheus metric reporting
+# the number of keys that differ between the ConfigMap and the agent's active
+# settings. A non-zero value indicates that the agent has not yet applied all
+# current ConfigMap changes and needs to be restarted.
+configDriftDetection:
+  # -- Enable watching of the cilium-config ConfigMap and reflecting its
+  # contents into the agent's internal DynamicConfig table.
+  enabled: true
+  # -- Enable the drift checker which compares the DynamicConfig table against
+  # the agent's active settings and publishes the
+  # cilium_drift_checker_config_delta metric.
+  driftChecker: true
+  # -- List of config-map keys to ignore when computing the drift delta.
+  ignoredKeys: []
 # -- Agent container image.
 image:
   # @schema
@@ -1148,9 +1170,89 @@ encryption:
   wireguard:
     # -- Controls WireGuard PersistentKeepalive option. Set 0s to disable.
     persistentKeepalive: 0s
+  # -- ztunnel encryption configuration.
+  # ztunnel is Istio's purpose-built, per-node proxy for handling L4 traffic in ambient mesh mode.
+  # These settings only apply when encryption.type is set to "ztunnel".
+  ztunnel:
+    # -- ztunnel container image.
+    image:
+      # @schema
+      # type: [null, string]
+      # @schema
+      override: ~
+      repository: "docker.io/istio/ztunnel"
+      tag: "1.28.0-distroless"
+      pullPolicy: "IfNotPresent"
+      # @schema
+      # type: [null, string]
+      # @schema
+      digest: ~
+      useDigest: false
+    # -- CA server address for certificate requests.
+    caAddress: "https://localhost:15012"
+    # -- TCP port for the health API.
+    healthPort: 15021
+    # -- ztunnel resource limits & requests.
+    resources:
+      requests:
+        cpu: 200m
+        memory: 512Mi
+    # -- ztunnel update strategy.
+    updateStrategy:
+      type: RollingUpdate
+      rollingUpdate:
+        maxSurge: 1
+        maxUnavailable: 0
+    # -- Configure termination grace period for ztunnel DaemonSet.
+    terminationGracePeriodSeconds: 30
+    # -- Readiness probe configuration.
+    readinessProbe:
+      initialDelaySeconds: 0
+      periodSeconds: 10
+      failureThreshold: 3
+    # -- Node selector for ztunnel pods.
+    nodeSelector:
+      kubernetes.io/os: linux
+    # -- Node tolerations for ztunnel scheduling.
+    tolerations:
+      - effect: NoSchedule
+        operator: Exists
+      - key: CriticalAddonsOnly
+        operator: Exists
+      - effect: NoExecute
+        operator: Exists
+    # -- Affinity for ztunnel pods.
+    affinity: {}
+    # @schema
+    # type: [null, string]
+    # @schema
+    # -- The priority class to use for ztunnel pods.
+    priorityClassName: ~
+    # -- Annotations to be added to all ztunnel resources.
+    annotations: {}
+    # -- Annotations to be added to ztunnel pods.
+    podAnnotations: {}
+    # -- Labels to be added to ztunnel pods.
+    podLabels: {}
+    # -- Additional ztunnel container environment variables.
+    extraEnv: []
+    # -- Additional ztunnel volumes.
+    extraVolumes: []
+    # -- Additional ztunnel volumeMounts.
+    extraVolumeMounts: []
+    # -- ztunnel secrets configuration.
+    secrets:
+      # -- Base64-encoded bootstrap root certificate content.
+      # If not provided, the secret must be created manually before deploying.
+      # @schema
+      # type: [null, string]
+      # @schema
+      bootstrapRootCert: ~
 endpointHealthChecking:
   # -- Enable connectivity health checking between virtual endpoints.
   enabled: true
+# -- Max duration to wait for envoy to respond to configuration changes. Default "10s".
+endpointPolicyUpdateTimeoutDuration: null
 endpointRoutes:
   # @schema
   # type: [boolean, string]
@@ -2339,8 +2441,6 @@ enableMasqueradeRouteSource: false
 enableIPv4BIGTCP: false
 # -- Enables IPv6 BIG TCP support which increases maximum IPv6 GSO/GRO limits for nodes and pods
 enableIPv6BIGTCP: false
-# -- Enable BIG TCP in tunneling mode and increase maximum GRO/GSO limits for VXLAN/GENEVE tunnels
-enableTunnelBIGTCP: false
 
 nat:
   # -- Number of the top-k SNAT map connections to track in Cilium statedb.
@@ -2420,7 +2520,7 @@ loadBalancer:
 
   # -- serviceTopology enables K8s Topology Aware Hints -based service
   # endpoints filtering
-  # serviceTopology: false
+  serviceTopology: false
 
   # -- L7 LoadBalancer
   l7:
diff --git a/packages/system/cilium/images/cilium/Dockerfile b/packages/system/cilium/images/cilium/Dockerfile
index b62f9b46..32fc6fb8 100644
--- a/packages/system/cilium/images/cilium/Dockerfile
+++ b/packages/system/cilium/images/cilium/Dockerfile
@@ -1,2 +1,2 @@
-ARG VERSION=v1.19.1
+ARG VERSION=v1.19.3
 FROM quay.io/cilium/cilium:${VERSION}
diff --git a/packages/system/cilium/values-apparmor.yaml b/packages/system/cilium/values-apparmor.yaml
new file mode 100644
index 00000000..0712836c
--- /dev/null
+++ b/packages/system/cilium/values-apparmor.yaml
@@ -0,0 +1,20 @@
+cilium:
+  # Opt out of the cri-containerd.apparmor.d profile for containers that invoke
+  # nsenter to join the host cgroup/mount namespace. The kernel reports the
+  # denial as a blocked "ptrace" operation. Required on Ubuntu 22.04+ and
+  # other distros that load this AppArmor profile by default, where the
+  # denial otherwise puts cilium-agent into Init:CrashLoopBackOff.
+  # The deprecated annotations are used because k8s >= 1.30 pod-level
+  # appArmorProfile: Unconfined is not honoured by containerd CRI today
+  # (kubernetes/kubernetes#125069).
+  # Only applied on non-Talos cilium variants where cgroup.autoMount.enabled is
+  # true — on Talos mount-cgroup is not rendered and the kube-apiserver would
+  # reject an annotation referencing a non-existent container.
+  # mount-bpf-fs is intentionally excluded: it renders with its own
+  # securityContext.privileged=true, and the kernel does not apply AppArmor
+  # profiles to privileged containers.
+  podAnnotations:
+    container.apparmor.security.beta.kubernetes.io/cilium-agent: unconfined
+    container.apparmor.security.beta.kubernetes.io/clean-cilium-state: unconfined
+    container.apparmor.security.beta.kubernetes.io/mount-cgroup: unconfined
+    container.apparmor.security.beta.kubernetes.io/apply-sysctl-overwrites: unconfined
diff --git a/packages/system/cilium/values.yaml b/packages/system/cilium/values.yaml
index cfb2cd6b..9aba7cb9 100644
--- a/packages/system/cilium/values.yaml
+++ b/packages/system/cilium/values.yaml
@@ -15,8 +15,8 @@ cilium:
     mode: "kubernetes"
   image:
     repository: ghcr.io/cozystack/cozystack/cilium
-    tag: 1.19.1
-    digest: "sha256:ab3acf270821df4614a8456348a4e0d3098aed72a4b2016a0edfa30d91428c3d"
+    tag: 1.19.3
+    digest: "sha256:700f06f4803a838a8e830be5ace4650e3ad82bdefabfb2f4d110368d307a5efb"
   envoy:
     enabled: false
   rollOutCiliumPods: true
diff --git a/packages/system/cozy-proxy/charts/cozy-proxy/Chart.yaml b/packages/system/cozy-proxy/charts/cozy-proxy/Chart.yaml
index 13352680..58f166f6 100644
--- a/packages/system/cozy-proxy/charts/cozy-proxy/Chart.yaml
+++ b/packages/system/cozy-proxy/charts/cozy-proxy/Chart.yaml
@@ -2,5 +2,5 @@ apiVersion: v2
 name: cozy-proxy
 description: A simple kube-proxy addon for 1:1 NAT services in Kubernetes using an NFT backend
 type: application
-version: 0.2.0
-appVersion: 0.2.0
+version: 0.3.0
+appVersion: 0.3.0
diff --git a/packages/system/cozy-proxy/charts/cozy-proxy/values.yaml b/packages/system/cozy-proxy/charts/cozy-proxy/values.yaml
index 8cde5bed..e143e926 100644
--- a/packages/system/cozy-proxy/charts/cozy-proxy/values.yaml
+++ b/packages/system/cozy-proxy/charts/cozy-proxy/values.yaml
@@ -1,6 +1,6 @@
 image:
   repository: ghcr.io/cozystack/cozystack/cozy-proxy
-  tag: v0.2.0
+  tag: v0.3.0
   pullPolicy: IfNotPresent
 
 daemonset:
diff --git a/packages/system/cozystack-api/values.yaml b/packages/system/cozystack-api/values.yaml
index ca4cf5e2..dddabf54 100644
--- a/packages/system/cozystack-api/values.yaml
+++ b/packages/system/cozystack-api/values.yaml
@@ -1,3 +1,3 @@
 cozystackAPI:
-  image: ghcr.io/cozystack/cozystack/cozystack-api:v1.2.1@sha256:afcc72ec5b8ff6eea7fbf927d518f277a7bfead44948f45ed17eda90d47253d5
+  image: ghcr.io/cozystack/cozystack/cozystack-api:v1.3.0@sha256:5fa8648821cf1e9e08cf7c2899c4b1c4226bb74c6773327141456c7d3f2b9b7e
   replicas: 2
diff --git a/packages/system/cozystack-controller/values.yaml b/packages/system/cozystack-controller/values.yaml
index 030a4564..4d0d0fde 100644
--- a/packages/system/cozystack-controller/values.yaml
+++ b/packages/system/cozystack-controller/values.yaml
@@ -1,4 +1,4 @@
 cozystackController:
-  image: ghcr.io/cozystack/cozystack/cozystack-controller:v1.2.1@sha256:0b1d49b29e8860f13f7afc8d7b6265f36c9d24d6ad7c40f5164535dd8e933c73
+  image: ghcr.io/cozystack/cozystack/cozystack-controller:v1.3.0@sha256:d03d19b78c4c98f970ac549a68b01ef6bd1ad755f5e0dcb9e08503511cdf2fdc
   debug: false
   disableTelemetry: false
diff --git a/packages/system/cozystack-scheduler/Chart.yaml b/packages/system/cozystack-scheduler/Chart.yaml
index c869abb4..f2dd13ee 100644
--- a/packages/system/cozystack-scheduler/Chart.yaml
+++ b/packages/system/cozystack-scheduler/Chart.yaml
@@ -1,3 +1,3 @@
 apiVersion: v2
 name: cozy-cozystack-scheduler
-version: 0.2.0
+version: 0.3.0
diff --git a/packages/system/cozystack-scheduler/Makefile b/packages/system/cozystack-scheduler/Makefile
index fd4bdf30..d075ec63 100644
--- a/packages/system/cozystack-scheduler/Makefile
+++ b/packages/system/cozystack-scheduler/Makefile
@@ -3,6 +3,9 @@ export NAMESPACE=kube-system
 
 include ../../../hack/package.mk
 
+test:
+	helm unittest .
+
 update:
 	rm -rf charts
 	tag=$$(git ls-remote --tags --sort="v:refname" https://github.com/cozystack/cozystack-scheduler | awk -F'[/^]' 'END{print $$3}') && \
diff --git a/packages/system/cozystack-scheduler/charts/cozystack-scheduler/Chart.yaml b/packages/system/cozystack-scheduler/charts/cozystack-scheduler/Chart.yaml
index c869abb4..f2dd13ee 100644
--- a/packages/system/cozystack-scheduler/charts/cozystack-scheduler/Chart.yaml
+++ b/packages/system/cozystack-scheduler/charts/cozystack-scheduler/Chart.yaml
@@ -1,3 +1,3 @@
 apiVersion: v2
 name: cozy-cozystack-scheduler
-version: 0.2.0
+version: 0.3.0
diff --git a/packages/system/cozystack-scheduler/charts/cozystack-scheduler/templates/configmap.yaml b/packages/system/cozystack-scheduler/charts/cozystack-scheduler/templates/configmap.yaml
index 1d78b225..24e0bc6f 100644
--- a/packages/system/cozystack-scheduler/charts/cozystack-scheduler/templates/configmap.yaml
+++ b/packages/system/cozystack-scheduler/charts/cozystack-scheduler/templates/configmap.yaml
@@ -52,3 +52,7 @@ data:
               - name: CozystackInterPodAffinity
               - name: CozystackNodeAffinity
               - name: CozystackPodTopologySpread
+    {{- with .Values.extenders }}
+    extenders:
+      {{- toYaml . | nindent 6 }}
+    {{- end }}
diff --git a/packages/system/cozystack-scheduler/charts/cozystack-scheduler/values.yaml b/packages/system/cozystack-scheduler/charts/cozystack-scheduler/values.yaml
index 5ab78dbe..55b6faff 100644
--- a/packages/system/cozystack-scheduler/charts/cozystack-scheduler/values.yaml
+++ b/packages/system/cozystack-scheduler/charts/cozystack-scheduler/values.yaml
@@ -1,4 +1,4 @@
-image: ghcr.io/cozystack/cozystack/cozystack-scheduler:v0.2.0@sha256:89c285c5c5fe3ed8d7d597acf32fc9f045394e0f8efafd1878d6080cf112f6c2
+image: ghcr.io/cozystack/cozystack/cozystack-scheduler:v0.3.0@sha256:89c285c5c5fe3ed8d7d597acf32fc9f045394e0f8efafd1878d6080cf112f6c2
 replicas: 1
 # defaultLabelSelectorKeys overrides the pod label keys used to auto-populate
 # nil LabelSelectors in SchedulingClass affinity and topology spread terms.
@@ -7,3 +7,17 @@ replicas: 1
 #   - apps.cozystack.io/application.kind
 #   - apps.cozystack.io/application.name
 # defaultLabelSelectorKeys: []
+
+# extenders is a list of scheduler extenders to call during the scheduling cycle.
+# Each entry is passed directly into KubeSchedulerConfiguration.extenders[].
+# Example:
+#   extenders:
+#     - urlPrefix: http://linstor-scheduler-extender.cozy-linstor.svc:8099
+#       filterVerb: filter
+#       prioritizeVerb: prioritize
+#       weight: 5
+#       enableHTTPS: false
+#       httpTimeout: 10s
+#       nodeCacheCapable: false
+#       ignorable: true
+extenders: []
diff --git a/packages/system/cozystack-scheduler/tests/configmap_test.yaml b/packages/system/cozystack-scheduler/tests/configmap_test.yaml
new file mode 100644
index 00000000..ce8f372d
--- /dev/null
+++ b/packages/system/cozystack-scheduler/tests/configmap_test.yaml
@@ -0,0 +1,26 @@
+suite: scheduler configmap tests
+
+templates:
+  - charts/cozy-cozystack-scheduler/templates/configmap.yaml
+
+tests:
+  - it: renders extenders when configured
+    values:
+      - ../values-linstor.yaml
+    asserts:
+      - isKind:
+          of: ConfigMap
+      - matchRegex:
+          path: data["scheduler-config.yaml"]
+          pattern: "urlPrefix: http://linstor-scheduler-extender"
+      - matchRegex:
+          path: data["scheduler-config.yaml"]
+          pattern: "ignorable: true"
+
+  - it: omits extenders by default
+    asserts:
+      - isKind:
+          of: ConfigMap
+      - notMatchRegex:
+          path: data["scheduler-config.yaml"]
+          pattern: "extenders:"
diff --git a/packages/system/cozystack-scheduler/values-linstor.yaml b/packages/system/cozystack-scheduler/values-linstor.yaml
new file mode 100644
index 00000000..ab16ca57
--- /dev/null
+++ b/packages/system/cozystack-scheduler/values-linstor.yaml
@@ -0,0 +1,10 @@
+cozy-cozystack-scheduler:
+  extenders:
+    - urlPrefix: http://linstor-scheduler-extender.cozy-linstor.svc:8099
+      filterVerb: filter
+      prioritizeVerb: prioritize
+      weight: 5
+      enableHTTPS: false
+      httpTimeout: 10s
+      nodeCacheCapable: false
+      ignorable: true
diff --git a/packages/system/dashboard/templates/configmap.yaml b/packages/system/dashboard/templates/configmap.yaml
index d260c5d9..a2b7684c 100644
--- a/packages/system/dashboard/templates/configmap.yaml
+++ b/packages/system/dashboard/templates/configmap.yaml
@@ -1,6 +1,6 @@
 {{- $brandingConfig := .Values._cluster.branding | default dict }}
 
-{{- $tenantText := "v1.2.1" }}
+{{- $tenantText := "v1.3.0" }}
 {{- $footerText := "Cozystack" }}
 {{- $titleText := "Cozystack Dashboard" }}
 {{- $logoText := "" }}
diff --git a/packages/system/dashboard/templates/ingress.yaml b/packages/system/dashboard/templates/ingress.yaml
index a10797fb..6cf01490 100644
--- a/packages/system/dashboard/templates/ingress.yaml
+++ b/packages/system/dashboard/templates/ingress.yaml
@@ -11,7 +11,7 @@ metadata:
   annotations:
     cert-manager.io/cluster-issuer: {{ $clusterIssuer }}
     {{- if eq $solver "http01" }}
-    acme.cert-manager.io/http01-ingress-class: {{ $exposeIngress }}
+    acme.cert-manager.io/http01-ingress-ingressclassname: {{ $exposeIngress }}
     {{- end }}
     nginx.ingress.kubernetes.io/rewrite-target: /
     nginx.ingress.kubernetes.io/client-max-body-size: 100m
diff --git a/packages/system/dashboard/values.yaml b/packages/system/dashboard/values.yaml
index 1aa69b09..a5188825 100644
--- a/packages/system/dashboard/values.yaml
+++ b/packages/system/dashboard/values.yaml
@@ -1,6 +1,6 @@
 openapiUI:
-  image: ghcr.io/cozystack/cozystack/openapi-ui:v1.2.1@sha256:a77c502c4b67fa5f0698f66a82eb2ad30e138f9bd2700aa40a63f73cb86a559d
+  image: ghcr.io/cozystack/cozystack/openapi-ui:v1.3.0@sha256:0fa79c373a62840a617ff1ca1b0e31931c13a6cf7b0bb0ff0dc191f047a465a3
 openapiUIK8sBff:
-  image: ghcr.io/cozystack/cozystack/openapi-ui-k8s-bff:v1.2.1@sha256:1b3ea6d4c7dbbe6a8def3b2807fffdfab2ac4afc39d7a846e57dd491fa168f92
+  image: ghcr.io/cozystack/cozystack/openapi-ui-k8s-bff:v1.3.0@sha256:1b3ea6d4c7dbbe6a8def3b2807fffdfab2ac4afc39d7a846e57dd491fa168f92
 tokenProxy:
-  image: ghcr.io/cozystack/cozystack/token-proxy:v1.2.1@sha256:2e280991e07853ea48f97b0a42946afffa10d03d6a83d41099ed83e6ffc94fdc
+  image: ghcr.io/cozystack/cozystack/token-proxy:v1.3.0@sha256:2e280991e07853ea48f97b0a42946afffa10d03d6a83d41099ed83e6ffc94fdc
diff --git a/packages/system/gpu-operator/examples/README.md b/packages/system/gpu-operator/examples/README.md
new file mode 100644
index 00000000..b6f170c8
--- /dev/null
+++ b/packages/system/gpu-operator/examples/README.md
@@ -0,0 +1,121 @@
+# GPU operator — native pod workload on Talos (reference)
+
+The files in this directory are **not** templates. They are reference
+artifacts that document one working configuration for running GPU
+workloads directly in pods on a Talos-based Cozystack cluster, together
+with the DCGM metrics needed by the `gpu/gpu-performance` Grafana
+dashboard.
+
+The out-of-the-box `values-talos.yaml` for this package targets the
+sandbox (VFIO passthrough to KubeVirt VMs) scenario. The files here
+illustrate an alternative — running CUDA workloads in regular pods with
+the NVIDIA device plugin — and the workarounds it currently requires on
+Talos.
+
+## Files
+
+- [`values-native-talos.yaml`](./values-native-talos.yaml) — Cozystack
+  `Package` values that disable sandbox workloads, enable the device
+  plugin, point `hostPaths.driverInstallDir` at the staging location
+  used by the compat DaemonSet, and wire DCGM to the custom metrics
+  ConfigMap.
+- [`dcgm-custom-metrics.yaml`](./dcgm-custom-metrics.yaml) — `ConfigMap`
+  with a DCGM metrics CSV that adds profiling, ECC, throttling and
+  energy counters on top of the upstream defaults. The CSV is a
+  superset needed for full coverage of the `gpu/gpu-performance`
+  dashboard. Which parts are actually required depends on which
+  dashboards you ship — see the table below.
+- [`nvidia-driver-compat.yaml`](./nvidia-driver-compat.yaml) — DaemonSet
+  that stages `libnvidia-ml.so.1` and `nvidia-smi` from the Talos glibc
+  tree into a path where the NVIDIA GPU Operator validator expects
+  them. See the "Why the compat DaemonSet exists" section below.
+
+## Why these are reference, not templates
+
+Shipping these as first-class templates would silently impose
+assumptions that do not hold for every user:
+
+- Whether the NVIDIA Talos system extension is installed on the nodes.
+- Whether GPUs are exposed directly to pods or passed through to VMs.
+- The exact path the installed driver ends up at (depends on the
+  extension version and Talos release).
+
+The sandbox-oriented `values-talos.yaml` remains the default. Operators
+who want native pod GPU workloads can start from this directory and
+adapt as needed.
+
+## Why the compat DaemonSet exists
+
+The NVIDIA GPU Operator validator checks for `libnvidia-ml.so.1` and
+`bin/nvidia-smi` in the path given by `hostPaths.driverInstallDir`.
+Talos installs them under `/usr/local/glibc/usr/lib/` and
+`/usr/local/bin/`, which the validator does not look at. Until upstream
+addresses [NVIDIA/gpu-operator#1687][1], the DaemonSet copies those
+files into a directory the validator does inspect and creates the
+`.driver-ctr-ready` flag file so the validator proceeds.
+
+[1]: https://github.com/NVIDIA/gpu-operator/issues/1687
+
+The compat DaemonSet runs privileged and bind-mounts host paths, so
+the target namespace must allow privileged pods. On clusters that
+enforce the Kubernetes Pod Security Standards at `baseline` or
+`restricted`, label the namespace with
+`pod-security.kubernetes.io/enforce: privileged` (and the matching
+`audit`/`warn` labels if the admission webhook is configured to
+surface violations) before applying the manifest.
+
+## Dashboards and what DCGM metrics they need
+
+Five GPU dashboards live under `gpu/*` in
+`packages/system/monitoring/dashboards-infra.list`. All of them share
+`packages/system/monitoring-agents/alerts/gpu-recording.rules.yaml` as
+their source of aggregated series. The recording rules are safe to
+ship on any cluster — they evaluate to empty series when DCGM is not
+scraped, or when optional counters are missing.
+
+What each dashboard needs on top of the upstream DCGM Exporter
+[`default-counters.csv`][default-csv]:
+
+| Dashboard         | Scope                              | Needs beyond defaults                                                   |
+| ----------------- | ---------------------------------- | ----------------------------------------------------------------------- |
+| `gpu-performance` | Per-node, per-GPU deep dive        | `DCGM_FI_DEV_POWER_VIOLATION`, `DCGM_FI_DEV_THERMAL_VIOLATION`          |
+| `gpu-efficiency`  | Per-workload util vs tensor active | `DCGM_FI_DEV_POWER_VIOLATION`, `DCGM_FI_DEV_THERMAL_VIOLATION` (via `gpu:*_throttle_fraction:rate5m` recording rules) |
+| `gpu-fleet`       | Cluster-wide admin inventory       | `DCGM_FI_DEV_POWER_MGMT_LIMIT` (for the TDP vs draw panel)              |
+| `gpu-quotas`      | Kube-quota vs live usage           | `kube_pod_container_resource_requests`, `kube_pod_status_phase`, `kube_node_status_allocatable` (via `namespace:gpu_count:allocated` / `cluster:gpu_count:*` recording rules) |
+| `gpu-tenants`     | Per-namespace tenant view          | nothing (works on default counters)                                     |
+
+`DCGM_FI_PROF_PIPE_TENSOR_ACTIVE` and `DCGM_FI_PROF_GR_ENGINE_ACTIVE`
+are already in the upstream default set for the pinned DCGM Exporter
+version, so the tensor-saturation and engine-active panels work without
+any CSV override. The three counters listed in the table — throttling
+violations and the power management limit — are the only extras the
+tracked dashboards need. The recording rules in
+`gpu-recording.rules.yaml` consume utilization, FB used, power,
+temperature and the tensor-active profiling counter from the default
+set, plus `DCGM_FI_DEV_POWER_VIOLATION` and
+`DCGM_FI_DEV_THERMAL_VIOLATION` — used by the
+`gpu.recording.efficiency.1m` group to derive the
+`gpu:power_throttle_fraction:rate5m` and
+`gpu:thermal_throttle_fraction:rate5m` series consumed by the
+throttling panels on the efficiency and fleet dashboards.
+
+The `gpu.recording.throttle.validation.5m` group additionally ships the
+`GPUThrottleFractionOverOne` alert (severity `warning`) as a regression
+detector: it fires when either throttle-fraction series exceeds 1.0,
+which would indicate that DCGM changed the scale/divisor of the
+underlying violation counters and the recording rules need to be
+re-derived.
+
+## Verification status
+
+The minimum-CSV claims above are verified by
+`hack/check-gpu-recording-rules.bats`, which cross-checks every
+`DCGM_FI_*` reference in the tracked GPU dashboards and recording rules
+against the union of the upstream default set (snapshotted at
+`hack/dcgm-default-counters.csv` for the pinned DCGM Exporter version)
+and the custom CSV in `dcgm-custom-metrics.yaml`. When the DCGM
+Exporter image in `packages/system/gpu-operator/charts/gpu-operator/values.yaml`
+is bumped, refresh the snapshot from the matching tag of the
+[`NVIDIA/dcgm-exporter`][default-csv] repository and rerun the test.
+
+[default-csv]: https://github.com/NVIDIA/dcgm-exporter/blob/main/etc/default-counters.csv
diff --git a/packages/system/gpu-operator/examples/dcgm-custom-metrics.yaml b/packages/system/gpu-operator/examples/dcgm-custom-metrics.yaml
new file mode 100644
index 00000000..ef2b0b88
--- /dev/null
+++ b/packages/system/gpu-operator/examples/dcgm-custom-metrics.yaml
@@ -0,0 +1,88 @@
+# Custom DCGM Exporter metrics CSV. Referenced from
+# examples/values-native-talos.yaml via dcgmExporter.config.name.
+#
+# Extends the upstream default set with profiling counters, ECC, page
+# retirement, row remap, energy and throttling violations — everything
+# the gpu/gpu-performance dashboard and the GPU recording rules in
+# monitoring-agents expect.
+apiVersion: v1
+kind: ConfigMap
+metadata:
+  name: dcgm-custom-metrics
+  namespace: cozy-gpu-operator
+data:
+  dcgm-metrics.csv: |
+    # Format
+    # If line starts with a '#' it is considered a comment
+    # DCGM FIELD, Prometheus metric type, help message
+
+    # Identity
+    DCGM_FI_DRIVER_VERSION, label, Driver version.
+
+    # Clocks
+    DCGM_FI_DEV_SM_CLOCK,  gauge, SM clock frequency (in MHz).
+    DCGM_FI_DEV_MEM_CLOCK, gauge, Memory clock frequency (in MHz).
+
+    # Temperature
+    DCGM_FI_DEV_MEMORY_TEMP, gauge, Memory temperature (in C).
+    DCGM_FI_DEV_GPU_TEMP,    gauge, GPU temperature (in C).
+
+    # Power
+    DCGM_FI_DEV_POWER_USAGE,              gauge, Power draw (in W).
+    DCGM_FI_DEV_POWER_MGMT_LIMIT,         gauge, Current power management limit (in W).
+    DCGM_FI_DEV_TOTAL_ENERGY_CONSUMPTION, counter, Total energy consumption since boot (in mJ).
+
+    # PCIE
+    DCGM_FI_DEV_PCIE_REPLAY_COUNTER, counter, Total number of PCIe retries.
+
+    # Utilization (the sample period varies depending on the product)
+    DCGM_FI_DEV_GPU_UTIL,      gauge, GPU utilization (in %).
+    DCGM_FI_DEV_MEM_COPY_UTIL, gauge, Memory utilization (in %).
+    DCGM_FI_DEV_ENC_UTIL,      gauge, Encoder utilization (in %).
+    DCGM_FI_DEV_DEC_UTIL,      gauge, Decoder utilization (in %).
+
+    # Errors and violations
+    DCGM_FI_DEV_XID_ERRORS, gauge, Value of the last XID error encountered.
+
+    # Memory usage
+    DCGM_FI_DEV_FB_FREE,     gauge, Framebuffer memory free (in MiB).
+    DCGM_FI_DEV_FB_USED,     gauge, Framebuffer memory used (in MiB).
+    DCGM_FI_DEV_FB_RESERVED, gauge, Framebuffer memory reserved (in MiB).
+
+    # ECC (supported on datacenter-class GPUs such as A10)
+    DCGM_FI_DEV_ECC_SBE_VOL_TOTAL, counter, Total number of single-bit volatile ECC errors.
+    DCGM_FI_DEV_ECC_DBE_VOL_TOTAL, counter, Total number of double-bit volatile ECC errors.
+    DCGM_FI_DEV_ECC_SBE_AGG_TOTAL, counter, Total number of single-bit persistent ECC errors.
+    DCGM_FI_DEV_ECC_DBE_AGG_TOTAL, counter, Total number of double-bit persistent ECC errors.
+
+    # Retired pages
+    DCGM_FI_DEV_RETIRED_SBE,     counter, Total number of retired pages due to single-bit errors.
+    DCGM_FI_DEV_RETIRED_DBE,     counter, Total number of retired pages due to double-bit errors.
+    DCGM_FI_DEV_RETIRED_PENDING, counter, Total number of pages pending retirement.
+
+    # Row remapping (not applicable to GDDR6 but left for datacenter GPUs)
+    DCGM_FI_DEV_UNCORRECTABLE_REMAPPED_ROWS, counter, Number of remapped rows for uncorrectable errors.
+    DCGM_FI_DEV_CORRECTABLE_REMAPPED_ROWS,   counter, Number of remapped rows for correctable errors.
+    DCGM_FI_DEV_ROW_REMAP_FAILURE,           gauge,   Whether remapping of rows has failed.
+
+    # Throttle / violation counters (crucial for SLA and tenant monitoring)
+    DCGM_FI_DEV_POWER_VIOLATION,       counter, Throttling duration due to power constraints (us per docs; ns on DCGM 3.x).
+    DCGM_FI_DEV_THERMAL_VIOLATION,     counter, Throttling duration due to thermal constraints (us per docs; ns on DCGM 3.x).
+    DCGM_FI_DEV_SYNC_BOOST_VIOLATION,  counter, Throttling duration due to sync-boost constraints (us per docs; ns on DCGM 3.x).
+    DCGM_FI_DEV_BOARD_LIMIT_VIOLATION, counter, Throttling duration due to board limit constraints (us per docs; ns on DCGM 3.x).
+    DCGM_FI_DEV_LOW_UTIL_VIOLATION,    counter, Throttling duration due to low utilization (us per docs; ns on DCGM 3.x).
+    DCGM_FI_DEV_RELIABILITY_VIOLATION, counter, Throttling duration due to reliability constraints (us per docs; ns on DCGM 3.x).
+
+    # NVLink — DCGM silently drops this metric on GPUs without NVLink, so
+    # enabling it unconditionally is safe and keeps this file reusable
+    # across GPU families.
+    DCGM_FI_DEV_NVLINK_BANDWIDTH_TOTAL, counter, Total number of NVLink bandwidth counters for all lanes.
+
+    # DCP (profiling) metrics — supported from Ampere onwards.
+    DCGM_FI_PROF_GR_ENGINE_ACTIVE,   gauge, Ratio of time the graphics engine is active.
+    DCGM_FI_PROF_SM_ACTIVE,          gauge, The ratio of cycles an SM has at least 1 warp assigned.
+    DCGM_FI_PROF_SM_OCCUPANCY,       gauge, The ratio of number of warps resident on an SM.
+    DCGM_FI_PROF_PIPE_TENSOR_ACTIVE, gauge, Ratio of cycles the tensor (HMMA) pipe is active.
+    DCGM_FI_PROF_DRAM_ACTIVE,        gauge, Ratio of cycles the device memory interface is active sending or receiving data.
+    DCGM_FI_PROF_PCIE_TX_BYTES,      counter, The number of bytes of active PCIe tx (transmit) data including both header and payload.
+    DCGM_FI_PROF_PCIE_RX_BYTES,      counter, The number of bytes of active PCIe rx (read) data including both header and payload.
diff --git a/packages/system/gpu-operator/examples/nvidia-driver-compat.yaml b/packages/system/gpu-operator/examples/nvidia-driver-compat.yaml
new file mode 100644
index 00000000..fc9d5981
--- /dev/null
+++ b/packages/system/gpu-operator/examples/nvidia-driver-compat.yaml
@@ -0,0 +1,109 @@
+# Workaround for https://github.com/NVIDIA/gpu-operator/issues/1687
+#
+# On Talos, the NVIDIA system extension installs driver files under
+# /usr/local/glibc/usr/lib/ and /usr/local/bin/. The GPU Operator
+# validator only looks for libnvidia-ml.so.1 and bin/nvidia-smi under
+# the path configured as hostPaths.driverInstallDir. This DaemonSet
+# stages the required files into /var/nvidia-driver on each node and
+# creates the .driver-ctr-ready flag so the validator proceeds.
+#
+# Paired with examples/values-native-talos.yaml, which sets
+# hostPaths.driverInstallDir to /var/nvidia-driver.
+apiVersion: apps/v1
+kind: DaemonSet
+metadata:
+  name: nvidia-driver-compat
+  namespace: cozy-gpu-operator
+  labels:
+    app: nvidia-driver-compat
+spec:
+  selector:
+    matchLabels:
+      app: nvidia-driver-compat
+  template:
+    metadata:
+      labels:
+        app: nvidia-driver-compat
+    # DaemonSet rather than a one-shot Job: staging is idempotent per-node,
+    # must re-run after every Talos reboot (hostPath contents are wiped on
+    # reboot when the system extension re-installs), and the pause container
+    # keeps the pod visible for operator observability (kubectl get pods).
+    spec:
+      priorityClassName: system-node-critical
+      # Restrict to GPU nodes only. Without this the DaemonSet schedules onto
+      # every node (including control-plane and CPU-only workers) and burns
+      # resources on hosts where the compat tree is meaningless.
+      # The label is published by Node Feature Discovery / GPU Operator's NFD
+      # subchart; if NFD is disabled, replace this with whatever label your
+      # cluster uses to mark GPU hosts.
+      nodeSelector:
+        nvidia.com/gpu.present: "true"
+      # Tolerate-all is intentionally broad: the nodeSelector above already
+      # confines scheduling to GPU nodes, so the "Exists" toleration only
+      # matters when those nodes carry extra taints (dedicated=gpu:NoSchedule,
+      # nvidia.com/gpu:NoSchedule from the GPU Operator's default policy,
+      # etc.). This mirrors the stance of the upstream nvidia-driver-daemonset
+      # and nvidia-device-plugin DaemonSets — blanket tolerations plus a
+      # narrow nodeSelector.
+      tolerations:
+        - operator: Exists
+      initContainers:
+        - name: create-driver-tree
+          image: busybox:1.37
+          command:
+            - sh
+            - -c
+            - |
+              set -e
+              GLIBC_LIB="/host/usr/local/glibc/usr/lib"
+              SRC_BIN="/host/usr/local/bin"
+              DST="/host/var/nvidia-driver"
+
+              mkdir -p "$DST/bin"
+
+              [ -f "$GLIBC_LIB/libnvidia-ml.so.1" ] || {
+                echo "missing $GLIBC_LIB/libnvidia-ml.so.1" >&2
+                exit 1
+              }
+              [ -f "$SRC_BIN/nvidia-smi" ] || {
+                echo "missing $SRC_BIN/nvidia-smi" >&2
+                exit 1
+              }
+
+              cp -f "$GLIBC_LIB/libnvidia-ml.so.1" "$DST/libnvidia-ml.so.1"
+              echo "Copied libnvidia-ml.so.1"
+
+              cp -f "$SRC_BIN/nvidia-smi" "$DST/bin/nvidia-smi"
+              chmod +x "$DST/bin/nvidia-smi"
+              echo "Copied nvidia-smi"
+
+              mkdir -p /host/run/nvidia/validations
+              touch /host/run/nvidia/validations/.driver-ctr-ready
+              echo "Created driver-ctr-ready flag"
+              echo "Done"
+          securityContext:
+            privileged: true
+          resources:
+            requests:
+              cpu: 10m
+              memory: 16Mi
+            limits:
+              cpu: 100m
+              memory: 64Mi
+          volumeMounts:
+            - name: host-root
+              mountPath: /host
+      containers:
+        - name: pause
+          image: registry.k8s.io/pause:3.10
+          resources:
+            requests:
+              cpu: 10m
+              memory: 8Mi
+            limits:
+              cpu: 50m
+              memory: 16Mi
+      volumes:
+        - name: host-root
+          hostPath:
+            path: /
diff --git a/packages/system/gpu-operator/examples/values-native-talos.yaml b/packages/system/gpu-operator/examples/values-native-talos.yaml
new file mode 100644
index 00000000..ef7b2891
--- /dev/null
+++ b/packages/system/gpu-operator/examples/values-native-talos.yaml
@@ -0,0 +1,52 @@
+# Cozystack Package values for running GPU workloads natively in pods
+# on Talos. This is the counterpart to values-talos.yaml, which targets
+# the sandbox (VFIO passthrough) scenario.
+#
+# Prerequisites:
+#   - NVIDIA Talos system extension installed on GPU nodes.
+#   - examples/nvidia-driver-compat.yaml deployed to stage driver files
+#     where the gpu-operator validator looks for them.
+#   - examples/dcgm-custom-metrics.yaml applied so DCGM Exporter exports
+#     the full set of metrics used by the dashboard and recording rules.
+apiVersion: cozystack.io/v1alpha1
+kind: Package
+metadata:
+  name: cozystack.gpu-operator
+spec:
+  variant: default
+  components:
+    gpu-operator:
+      values:
+        gpu-operator:
+          # The compat DaemonSet stages libnvidia-ml.so.1 and nvidia-smi
+          # under /var/nvidia-driver. Point the validator at that path.
+          hostPaths:
+            driverInstallDir: "/var/nvidia-driver"
+          # Disable the sandbox path — workloads run in pods, not VMs.
+          sandboxWorkloads:
+            enabled: false
+          vfioManager:
+            enabled: false
+          # The Talos extension provides the driver and runtime hooks,
+          # so the operator's own toolkit and driver components must be
+          # switched off to avoid conflicting installations.
+          toolkit:
+            enabled: false
+          devicePlugin:
+            enabled: true
+          # Export full set of DCGM metrics using the ConfigMap in
+          # examples/dcgm-custom-metrics.yaml.
+          dcgmExporter:
+            serviceMonitor:
+              enabled: true
+              # Matches the 1m evaluation cadence of GPU recording rules with
+              # 4x oversampling headroom. DCGM_FI_PROF_* counters have
+              # documented overhead concerns below 1s; 15s is safe.
+              interval: "15s"
+              honorLabels: true
+              relabelings:
+                - sourceLabels: [__meta_kubernetes_pod_node_name]
+                  targetLabel: node
+                  action: replace
+            config:
+              name: dcgm-custom-metrics
diff --git a/packages/system/grafana-operator/images/grafana-dashboards.tag b/packages/system/grafana-operator/images/grafana-dashboards.tag
index 08f6bd0b..cde83a70 100644
--- a/packages/system/grafana-operator/images/grafana-dashboards.tag
+++ b/packages/system/grafana-operator/images/grafana-dashboards.tag
@@ -1 +1 @@
-ghcr.io/cozystack/cozystack/grafana-dashboards:v1.2.1@sha256:2c9aa0b48e2bf6167db198f4d15882bfe51700108edf2e9f6d0942940a2c1204
+ghcr.io/cozystack/cozystack/grafana-dashboards:v1.3.0@sha256:2c9aa0b48e2bf6167db198f4d15882bfe51700108edf2e9f6d0942940a2c1204
diff --git a/packages/system/hami/Chart.yaml b/packages/system/hami/Chart.yaml
new file mode 100644
index 00000000..3fcf4c5d
--- /dev/null
+++ b/packages/system/hami/Chart.yaml
@@ -0,0 +1,3 @@
+apiVersion: v2
+name: cozy-hami
+version: 0.0.0 # Placeholder, the actual version will be automatically set during the build process
diff --git a/packages/system/hami/Makefile b/packages/system/hami/Makefile
new file mode 100644
index 00000000..83663a66
--- /dev/null
+++ b/packages/system/hami/Makefile
@@ -0,0 +1,43 @@
+export NAME=hami
+export NAMESPACE=cozy-$(NAME)
+
+include ../../../hack/common-envs.mk
+include ../../../hack/package.mk
+
+# When bumping the HAMi version, run `make update` and then review
+# the resulting diff in `charts/hami/`. The recipe below reproduces the
+# top-level vendoring overrides automatically:
+#
+#   1. Removes the broken hami-dra subchart. Upstream's NVIDIA DRA driver
+#      path requires kubelet DRA support that cozystack does not enable
+#      and has no upstream fix tracked. See commit 3c5521e.
+#   2. Empties Chart.yaml dependencies and drops Chart.lock so Helm does
+#      not try to re-pull hami-dra at build time. See commit 2734dc0.
+#   3. Strips dra/hami-dra/podSecurityPolicy blocks from the upstream
+#      values.yaml since the corresponding code paths are gone. PSP is
+#      removed from Kubernetes 1.25+ and is unused by cozystack.
+#
+# Template-level patches are NOT reproduced automatically:
+#
+#   * Scheduler templates have `{{- if .Values.dra.enabled }}` blocks
+#     that need to be removed because the dra value is gone (commit
+#     2734dc0 stripped them).
+#   * device-plugin/monitorservice.yaml uses `indent` with leading
+#     whitespace; it must be rewritten to `nindent` for the labels block
+#     to render correctly when devicePlugin.service.labels is set
+#     (commit 3685254).
+#
+# After `make update`, run `git diff -- charts/hami/templates/` and
+# replay these template patches against the new upstream version, then
+# verify with `helm unittest`. If upstream restructured the affected
+# files, the patches may need to be redesigned rather than reapplied.
+
+update:
+	rm -rf charts
+	helm repo add hami-charts https://project-hami.github.io/HAMi/
+	helm repo update hami-charts
+	helm pull hami-charts/hami --untar --untardir charts
+	rm -rf charts/hami/charts/hami-dra
+	yq --inplace '.dependencies = []' charts/hami/Chart.yaml
+	rm -f charts/hami/Chart.lock
+	yq --inplace 'del(.dra) | del(.["hami-dra"]) | del(.podSecurityPolicy)' charts/hami/values.yaml
diff --git a/packages/system/hami/README.md b/packages/system/hami/README.md
new file mode 100644
index 00000000..68669ac7
--- /dev/null
+++ b/packages/system/hami/README.md
@@ -0,0 +1,82 @@
+# HAMi — GPU Virtualization Middleware
+
+[HAMi](https://github.com/Project-HAMi/HAMi) (Heterogeneous AI Computing Virtualization Middleware) is a CNCF Sandbox project that enables fractional GPU sharing in Kubernetes. It allows workloads to request specific amounts of GPU memory and compute cores instead of claiming entire GPUs.
+
+## Architecture
+
+HAMi consists of four components:
+
+- **MutatingWebhook** — intercepts pod creation, injects `schedulerName: hami-scheduler`
+- **Scheduler Extender** — extends kube-scheduler with GPU-aware Filter and Bind logic
+- **Device Plugin** (DaemonSet) — registers vGPU resources via the Kubernetes Device Plugin API
+- **HAMi-core** (`libvgpu.so`) — `LD_PRELOAD` library injected into workload containers, intercepts CUDA API calls to enforce memory and compute isolation
+
+## Prerequisites
+
+- GPU Operator must be enabled (`addons.gpuOperator.enabled: true`)
+- NVIDIA driver >= 440 on host nodes
+- nvidia-container-toolkit configured as the default container runtime
+- GPU nodes labeled with `gpu=on`
+
+## Known Limitations
+
+### glibc < 2.34 requirement for workload containers
+
+HAMi-core uses `LD_PRELOAD` to intercept `dlsym()` for CUDA symbol resolution. The fallback code path relies on `_dl_sym`, a private glibc internal symbol that was removed in glibc 2.34 when libdl and libpthread were merged into libc.so.
+
+**This limitation affects workload containers only**, not the host OS or HAMi's own components.
+
+| Distribution    | glibc | Result                                       |
+| --------------- | ----- | -------------------------------------------- |
+| Ubuntu 18.04    | 2.27  | Full isolation (memory + compute)            |
+| Ubuntu 20.04    | 2.31  | Full isolation (memory + compute)            |
+| Ubuntu 22.04    | 2.35  | Memory isolation works, compute breaks       |
+| Ubuntu 24.04    | 2.39  | Both memory and compute isolation break      |
+| Alpine (musl)   | N/A   | Completely incompatible (`dlvsym` absent)    |
+
+Most modern ML/AI base images (CUDA 12.x, PyTorch 2.x, TensorFlow 2.x) use Ubuntu 22.04+ with glibc >= 2.35, which means compute isolation will not work with these images until the upstream fix is merged.
+
+**Upstream tracking issues:**
+
+- [HAMi-core#174](https://github.com/Project-HAMi/HAMi-core/issues/174) — `_dl_sym` removal in glibc 2.34 breaks HAMi-core's CUDA symbol resolution at the symbol level
+- [HAMi#1190](https://github.com/Project-HAMi/HAMi/issues/1190) — maintainer thread confirming the empirical per-glibc-version isolation behavior shown in the table above
+
+### musl libc (Alpine) incompatibility
+
+HAMi-core is completely incompatible with musl libc. The `dlvsym()` function used by HAMi-core is a glibc extension not available in musl. Only glibc-based container images (Debian, Ubuntu, RHEL, etc.) can use HAMi GPU isolation.
+
+## Usage
+
+Enable HAMi in your tenant Kubernetes cluster values:
+
+```yaml
+addons:
+  gpuOperator:
+    enabled: true
+  hami:
+    enabled: true
+```
+
+When HAMi is enabled, GPU Operator's built-in device plugin is automatically disabled to avoid conflicts. This default is preserved by setting `addons.gpuOperator.valuesOverride.gpu-operator.devicePlugin.enabled: false`; advanced topologies that partition GPU pools (e.g. some nodes use HAMi while others run the standard NVIDIA device plugin via node selectors) can re-enable it explicitly through `valuesOverride`.
+
+### Requesting fractional GPU resources
+
+```yaml
+resources:
+  limits:
+    nvidia.com/gpu: 1
+    nvidia.com/gpumem: 3000     # 3000 MB of GPU memory
+    nvidia.com/gpucores: 30     # 30% of GPU compute cores
+```
+
+## Parameters
+
+Default values shown below are inherited from the upstream HAMi chart and may change with upstream updates.
+
+| Name | Description | Default |
+| --- | --- | --- |
+| `hami.devicePlugin.runtimeClassName` | RuntimeClass for device plugin pods | `nvidia` |
+| `hami.devicePlugin.deviceSplitCount` | Max virtual GPUs per physical GPU | `10` |
+| `hami.devicePlugin.deviceMemoryScaling` | Memory overcommit factor (> 1.0 enables overcommit) | `1` |
+| `hami.scheduler.defaultSchedulerPolicy.nodeSchedulerPolicy` | Node packing strategy (`binpack` or `spread`) | `binpack` |
+| `hami.scheduler.defaultSchedulerPolicy.gpuSchedulerPolicy` | GPU packing strategy (`binpack` or `spread`) | `spread` |
diff --git a/packages/system/hami/charts/hami/Chart.yaml b/packages/system/hami/charts/hami/Chart.yaml
new file mode 100644
index 00000000..55f32ab6
--- /dev/null
+++ b/packages/system/hami/charts/hami/Chart.yaml
@@ -0,0 +1,18 @@
+apiVersion: v2
+appVersion: 2.8.1
+dependencies: []
+description: Heterogeneous AI Computing Virtualization Middleware
+keywords:
+- vgpu
+- gpu
+kubeVersion: '>= 1.18.0-0'
+maintainers:
+- email: archlitchi@gmail.com
+  name: limengxuan
+- email: xiaozhang0210@hotmail.com
+  name: zhangxiao
+name: hami
+sources:
+- https://github.com/Project-HAMi/HAMi
+type: application
+version: 2.8.1
diff --git a/packages/system/hami/charts/hami/README.md b/packages/system/hami/charts/hami/README.md
new file mode 100644
index 00000000..e8217290
--- /dev/null
+++ b/packages/system/hami/charts/hami/README.md
@@ -0,0 +1,237 @@
+# HAMi Helm Chart Values Documentation
+
+This document provides detailed descriptions of all configurable values parameters for the HAMi Helm Chart.
+
+## Global Configuration
+
+| Parameter | Description | Default Value |
+|-----------|-------------|---------------|
+| `global.imageRegistry` | Global Docker image registry | `""` |
+| `global.imagePullSecrets` | Global Docker image pull secrets | `[]` |
+| `global.imageTag` | Image tag | `"v2.8.1"` |
+| `global.gpuHookPath` | GPU Hook path | `/usr/local` |
+| `global.labels` | Global labels | `{}` |
+| `global.annotations` | Global annotations | `{}` |
+| `global.managedNodeSelectorEnable` | Whether to enable managed node selector | `false` |
+| `global.managedNodeSelector.usage` | Managed node selector usage | `"gpu"` |
+| `nameOverride` | Name override | `""` |
+| `fullnameOverride` | Full name override | `""` |
+| `namespaceOverride` | Namespace override | `""` |
+
+## Resource Name Configuration
+
+### NVIDIA GPU Resources
+| Parameter | Description | Default Value |
+|-----------|-------------|---------------|
+| `resourceName` | GPU resource name | `"nvidia.com/gpu"` |
+| `resourceMem` | GPU memory resource name | `"nvidia.com/gpumem"` |
+| `resourceMemPercentage` | GPU memory percentage resource name | `"nvidia.com/gpumem-percentage"` |
+| `resourceCores` | GPU core resource name | `"nvidia.com/gpucores"` |
+| `resourcePriority` | GPU priority resource name | `"nvidia.com/priority"` |
+
+### Cambricon MLU Resources
+| Parameter | Description | Default Value |
+|-----------|-------------|---------------|
+| `mluResourceName` | MLU resource name | `"cambricon.com/vmlu"` |
+| `mluResourceMem` | MLU memory resource name | `"cambricon.com/mlu.smlu.vmemory"` |
+| `mluResourceCores` | MLU core resource name | `"cambricon.com/mlu.smlu.vcore"` |
+
+### Hygon DCU Resources
+| Parameter | Description | Default Value |
+|-----------|-------------|---------------|
+| `dcuResourceName` | DCU resource name | `"hygon.com/dcunum"` |
+| `dcuResourceMem` | DCU memory resource name | `"hygon.com/dcumem"` |
+| `dcuResourceCores` | DCU core resource name | `"hygon.com/dcucores"` |
+
+### Metax GPU Resources
+| Parameter | Description | Default Value |
+|-----------|-------------|---------------|
+| `metaxResourceName` | GPU resource name | `"metax-tech.com/sgpu"` |
+| `metaxResourceCore` | GPU core resource name | `"metax-tech.com/vcore"` |
+| `metaxResourceMem` | GPU memory resource name | `"metax-tech.com/vmemory"` |
+| `metaxsGPUTopologyAware` | GPU topology awareness | `"false"` |
+
+### Enflame GCU Resources
+| Parameter | Description | Default Value |
+|-----------|-------------|---------------|
+| `enflameResourceNameVGCU` | vGCU resource name | `"enflame.com/vgcu"` |
+| `enflameResourceNameVGCUPercentage` | vGCU percentage resource name | `"enflame.com/vgcu-percentage"` |
+
+### Kunlunxin XPU Resources
+| Parameter | Description | Default Value |
+|-----------|-------------|---------------|
+| `kunlunResourceName` | XPU resource name | `"kunlunxin.com/xpu"` |
+
+## Scheduler Configuration
+
+| Parameter | Description | Default Value |
+|-----------|-------------|---------------|
+| `schedulerName` | Scheduler name | `"hami-scheduler"` |
+| `scheduler.nodeName` | Define node name, scheduler will schedule to this node | `""` |
+| `scheduler.overwriteEnv` | Whether to overwrite environment variables | `"false"` |
+| `scheduler.defaultSchedulerPolicy.nodeSchedulerPolicy` | Node scheduler policy | `binpack` |
+| `scheduler.defaultSchedulerPolicy.gpuSchedulerPolicy` | GPU scheduler policy | `spread` |
+| `scheduler.metricsBindAddress` | Metrics bind address | `":9395"` |
+| `scheduler.forceOverwriteDefaultScheduler` | Whether to force overwrite default scheduler | `true` |
+| `scheduler.livenessProbe` | Whether to enable liveness probe | `false` |
+| `scheduler.leaderElect` | Whether to enable leader election | `true` |
+| `scheduler.replicas` | Number of replicas | `1` |
+
+### Kube Scheduler Configuration
+
+| Parameter | Description | Default Value |
+|-----------|-------------|---------------|
+| `scheduler.kubeScheduler.enabled` | Whether to run kube-scheduler container in scheduler pod | `true` |
+| `scheduler.kubeScheduler.image.registry` | Kube scheduler image registry | `"registry.cn-hangzhou.aliyuncs.com"` |
+| `scheduler.kubeScheduler.image.repository` | Kube scheduler image repository | `"google_containers/kube-scheduler"` |
+| `scheduler.kubeScheduler.image.tag` | Kube scheduler image tag | `""` |
+| `scheduler.kubeScheduler.image.pullPolicy` | Kube scheduler image pull policy | `IfNotPresent` |
+| `scheduler.kubeScheduler.image.pullSecrets` | Kube scheduler image pull secrets | `[]` |
+| `scheduler.kubeScheduler.extraNewArgs` | Extra new arguments | `["--config=/config/config.yaml", "-v=4"]` |
+| `scheduler.kubeScheduler.extraArgs` | Extra arguments | `["--policy-config-file=/config/config.json", "-v=4"]` |
+
+### Extender Configuration
+
+| Parameter | Description | Default Value |
+|-----------|-------------|---------------|
+| `scheduler.extender.image.registry` | Scheduler extender image registry | `"docker.io"` |
+| `scheduler.extender.image.repository` | Scheduler extender image repository | `"projecthami/hami"` |
+| `scheduler.extender.image.tag` | Scheduler extender image tag | `""` |
+| `scheduler.extender.image.pullPolicy` | Scheduler extender image pull policy | `IfNotPresent` |
+| `scheduler.extender.image.pullSecrets` | Scheduler extender image pull secrets | `[]` |
+| `scheduler.extender.extraArgs` | Scheduler extender extra arguments | `["--debug", "-v=4"]` |
+
+### Admission Webhook Configuration
+
+| Parameter | Description | Default Value |
+|-----------|-------------|---------------|
+| `scheduler.admissionWebhook.enabled` | Whether to enable admission webhook | `true` |
+| `scheduler.admissionWebhook.customURL.enabled` | Whether to enable custom URL | `false` |
+| `scheduler.admissionWebhook.customURL.host` | Custom URL host | `127.0.0.1` |
+| `scheduler.admissionWebhook.customURL.port` | Custom URL port | `31998` |
+| `scheduler.admissionWebhook.customURL.path` | Custom URL path | `/webhook` |
+| `scheduler.admissionWebhook.reinvocationPolicy` | Reinvocation policy | `Never` |
+| `scheduler.admissionWebhook.failurePolicy` | Failure policy | `Ignore` |
+
+### TLS Certificate Configuration
+
+| Parameter | Description | Default Value |
+|-----------|-------------|---------------|
+| `scheduler.certManager.enabled` | Whether to use cert-manager to generate self-signed certificates | `false` |
+| `scheduler.patch.enabled` | Whether to use kube-webhook-certgen to generate self-signed certificates | `true` |
+| `scheduler.patch.image.registry` | Certgen image registry | `"docker.io"` |
+| `scheduler.patch.image.repository` | Certgen image repository | `"jettech/kube-webhook-certgen"` |
+| `scheduler.patch.image.tag` | Certgen image tag | `"v1.5.2"` |
+| `scheduler.patch.image.pullPolicy` | Certgen image pull policy | `IfNotPresent` |
+| `scheduler.patch.imageNew.registry` | New certgen image registry | `"docker.io"` |
+| `scheduler.patch.imageNew.repository` | New certgen image repository | `"liangjw/kube-webhook-certgen"` |
+| `scheduler.patch.imageNew.tag` | New certgen image tag | `"v1.1.1"` |
+
+### Scheduler Service Configuration
+
+| Parameter | Description | Default Value |
+|-----------|-------------|---------------|
+| `scheduler.service.type` | Service type | `NodePort` |
+| `scheduler.service.httpPort` | HTTP port | `443` |
+| `scheduler.service.schedulerPort` | Scheduler NodePort | `31998` |
+| `scheduler.service.monitorPort` | Monitor port | `31993` |
+| `scheduler.service.monitorTargetPort` | Monitor target port | `9395` |
+
+## Device Plugin Configuration
+
+| Parameter | Description | Default Value |
+|-----------|-------------|---------------|
+| `devicePlugin.image.registry` | Device plugin image registry | `"docker.io"` |
+| `devicePlugin.image.repository` | Device plugin image repository | `"projecthami/hami"` |
+| `devicePlugin.image.tag` | Device plugin image tag | `""` |
+| `devicePlugin.image.pullPolicy` | Device plugin image pull policy | `IfNotPresent` |
+| `devicePlugin.image.pullSecrets` | Device plugin image pull secrets | `[]` |
+
+### Monitor Configuration
+
+| Parameter | Description | Default Value |
+|-----------|-------------|---------------|
+| `devicePlugin.monitor.image.registry` | Monitor image registry | `"docker.io"` |
+| `devicePlugin.monitor.image.repository` | Monitor image repository | `"projecthami/hami"` |
+| `devicePlugin.monitor.image.tag` | Monitor image tag | `""` |
+| `devicePlugin.monitor.image.pullPolicy` | Monitor image pull policy | `IfNotPresent` |
+| `devicePlugin.monitor.image.pullSecrets` | Monitor image pull secrets | `[]` |
+| `devicePlugin.monitor.ctrPath` | Container path | `/usr/local/vgpu/containers` |
+| `devicePlugin.monitor.extraArgs` | Monitor extra arguments | `["-v=4"]` |
+| `devicePlugin.monitor.extraEnvs` | Monitor extra environments | `{}` |
+
+### Device Plugin Other Configuration
+
+| Parameter | Description | Default Value |
+|-----------|-------------|---------------|
+| `devicePlugin.deviceSplitCount` | Integer type, default value: 10. Maximum number of tasks assigned to a single GPU device | `10` |
+| `devicePlugin.deviceMemoryScaling` | Device memory scaling ratio | `1` |
+| `devicePlugin.deviceCoreScaling` | Device core scaling ratio | `1` |
+| `devicePlugin.runtimeClassName` | Runtime class name | `""` |
+| `devicePlugin.createRuntimeClass` | Whether to create runtime class | `false` |
+| `devicePlugin.migStrategy` | String type, "none" means ignore MIG functionality, "mixed" means allocate MIG devices through independent resources | `"none"` |
+| `devicePlugin.disablecorelimit` | String type, "true" means disable core limit, "false" means enable core limit | `"false"` |
+| `devicePlugin.passDeviceSpecsEnabled` | Whether to enable passing device specs | `false` |
+| `devicePlugin.extraArgs` | Device plugin extra arguments | `["-v=4"]` |
+| `devicePlugin.nodeConfiguration.config` | Node configuration for device plugin by json | An example of default configuration. |
+| `devicePlugin.nodeConfiguration.externalConfigName` | Node configuration for device plugin by external congimap | `""` |
+| `devicePlugin.extraEnvs` | Device plugin extra environments | `{}` |
+
+### Device Plugin Service Configuration
+
+| Parameter | Description | Default Value |
+|-----------|-------------|---------------|
+| `devicePlugin.service.type` | Service type | `NodePort` |
+| `devicePlugin.service.httpPort` | HTTP port | `31992` |
+
+### Device Plugin Deployment Configuration
+
+| Parameter | Description | Default Value |
+|-----------|-------------|---------------|
+| `devicePlugin.pluginPath` | Plugin path | `/var/lib/kubelet/device-plugins` |
+| `devicePlugin.libPath` | Library path | `/usr/local/vgpu` |
+| `devicePlugin.nvidiaNodeSelector` | NVIDIA node selector | `{"gpu": "on"}` |
+| `devicePlugin.updateStrategy.type` | Update strategy type | `RollingUpdate` |
+| `devicePlugin.updateStrategy.rollingUpdate.maxUnavailable` | Maximum unavailable count | `1` |
+
+## Device Configuration
+
+### AWS Neuron
+| Parameter | Description | Default Value |
+|-----------|-------------|---------------|
+| `devices.awsneuron.customresources` | Custom resources | `["aws.amazon.com/neuron", "aws.amazon.com/neuroncore"]` |
+
+### Kunlunxin
+| Parameter | Description | Default Value |
+|-----------|-------------|---------------|
+| `devices.kunlun.enabled` | Whether to enable | `true` |
+| `devices.kunlun.customresources` | Custom resources | `["kunlunxin.com/xpu"]` |
+
+### Mthreads
+| Parameter | Description | Default Value |
+|-----------|-------------|---------------|
+| `devices.mthreads.enabled` | Whether to enable | `true` |
+| `devices.mthreads.customresources` | Custom resources | `["mthreads.com/vgpu"]` |
+
+### NVIDIA
+| Parameter | Description | Default Value |
+|-----------|-------------|---------------|
+| `devices.nvidia.gpuCorePolicy` | GPU core policy | `default` |
+| `devices.nvidia.libCudaLogLevel` | CUDA library log level | `1` |
+
+### Huawei Ascend
+| Parameter | Description | Default Value |
+|-----------|-------------|---------------|
+| `devices.ascend.enabled` | Whether to enable | `false` |
+| `devices.ascend.image` | Image | `""` |
+| `devices.ascend.imagePullPolicy` | Image pull policy | `IfNotPresent` |
+| `devices.ascend.extraArgs` | Extra arguments | `[]` |
+| `devices.ascend.nodeSelector` | Node selector | `{"ascend": "on"}` |
+| `devices.ascend.tolerations` | Tolerations | `[]` |
+| `devices.ascend.customresources` | Custom resources | `["huawei.com/Ascend910A", "huawei.com/Ascend910A-memory", ...]` |
+
+### Iluvatar
+| Parameter | Description | Default Value |
+|-----------|-------------|---------------|
+| `devices.iluvatar.enabled` | Whether to enable | `false` |
+| `devices.iluvatar.customresources` | Custom resources | `["iluvatar.ai/BI-V150-vgpu", "iluvatar.ai/BI-V150.vMem","iluvatar.ai/BI-V150.vCore", ...]` |
diff --git a/packages/system/hami/charts/hami/templates/NOTES.txt b/packages/system/hami/charts/hami/templates/NOTES.txt
new file mode 100644
index 00000000..15bb1218
--- /dev/null
+++ b/packages/system/hami/charts/hami/templates/NOTES.txt
@@ -0,0 +1,3 @@
+** Please be patient while the chart is being deployed **
+Resource name: {{ .Values.resourceName }}
+
diff --git a/packages/system/hami/charts/hami/templates/_commons.tpl b/packages/system/hami/charts/hami/templates/_commons.tpl
new file mode 100644
index 00000000..b68018e4
--- /dev/null
+++ b/packages/system/hami/charts/hami/templates/_commons.tpl
@@ -0,0 +1,49 @@
+{{/*
+Return the proper image name
+{{ include "common.images.image" ( dict "imageRoot" .Values.path.to.the.image "global" $) }}
+*/}}
+{{- define "common.images.image" -}}
+{{- $registryName := .imageRoot.registry -}}
+{{- $repositoryName := .imageRoot.repository -}}
+{{- $tag := .imageRoot.tag | toString -}}
+{{- if .global }}
+    {{- if .global.imageRegistry }}
+     {{- $registryName = .global.imageRegistry -}}
+    {{- end -}}
+{{- end -}}
+{{- if .tag }}
+     {{- $tag = .tag | toString -}}
+{{- end -}}
+{{- if $registryName }}
+{{- printf "%s/%s:%s" $registryName $repositoryName $tag -}}
+{{- else -}}
+{{- printf "%s:%s" $repositoryName $tag -}}
+{{- end -}}
+{{- end -}}
+
+{{/*
+Return the proper Docker Image Registry Secret Names (deprecated: use common.images.renderPullSecrets instead)
+{{ include "common.images.pullSecrets" ( dict "images" (list .Values.path.to.the.image1, .Values.path.to.the.image2) "global" .Values.global) }}
+*/}}
+{{- define "common.images.pullSecrets" -}}
+  {{- $pullSecrets := list }}
+
+  {{- if .global }}
+    {{- range .global.imagePullSecrets -}}
+      {{- $pullSecrets = append $pullSecrets . -}}
+    {{- end -}}
+  {{- end -}}
+
+  {{- range .images -}}
+    {{- range .pullSecrets -}}
+      {{- $pullSecrets = append $pullSecrets . -}}
+    {{- end -}}
+  {{- end -}}
+
+  {{- if (not (empty $pullSecrets)) }}
+imagePullSecrets:
+    {{- range $pullSecrets }}
+  - name: {{ . }}
+    {{- end }}
+  {{- end }}
+{{- end -}}
diff --git a/packages/system/hami/charts/hami/templates/_helpers.tpl b/packages/system/hami/charts/hami/templates/_helpers.tpl
new file mode 100644
index 00000000..ffcc61da
--- /dev/null
+++ b/packages/system/hami/charts/hami/templates/_helpers.tpl
@@ -0,0 +1,163 @@
+{{/*
+Expand the name of the chart.
+*/}}
+{{- define "hami-vgpu.name" -}}
+{{- default .Chart.Name .Values.nameOverride | trunc 63 | trimSuffix "-" -}}
+{{- end -}}
+
+{{/*
+Create a default fully qualified app name.
+We truncate at 63 chars because some Kubernetes name fields are limited to this (by the DNS naming spec).
+If release name contains chart name it will be used as a full name.
+*/}}
+{{- define "hami-vgpu.fullname" -}}
+{{- if .Values.fullnameOverride -}}
+{{- .Values.fullnameOverride | trunc 63 | trimSuffix "-" -}}
+{{- else }}
+{{- $name := default .Chart.Name .Values.nameOverride -}}
+{{- if contains $name .Release.Name }}
+{{- .Release.Name | trunc 63 | trimSuffix "-" -}}
+{{- else -}}
+{{- printf "%s-%s" .Release.Name $name | trunc 63 | trimSuffix "-" -}}
+{{- end -}}
+{{- end -}}
+{{- end -}}
+
+{{/*
+Allow the release namespace to be overridden for multi-namespace deployments in combined charts
+*/}}
+{{- define "hami-vgpu.namespace" -}}
+  {{- if .Values.namespaceOverride -}}
+    {{- .Values.namespaceOverride -}}
+  {{- else -}}
+    {{- .Release.Namespace -}}
+  {{- end -}}
+{{- end -}}
+
+{{/*
+The app name for Scheduler
+*/}}
+{{- define "hami-vgpu.scheduler" -}}
+{{- printf "%s-scheduler" ( include "hami-vgpu.fullname" . ) | trunc 63 | trimSuffix "-" -}}
+{{- end -}}
+
+{{/*
+The app name for DevicePlugin
+*/}}
+{{- define "hami-vgpu.device-plugin" -}}
+{{- printf "%s-device-plugin" ( include "hami-vgpu.fullname" . ) | trunc 63 | trimSuffix "-" -}}
+{{- end -}}
+
+{{/*
+  The app name for MockDevicePlugin
+  */}}
+{{- define "hami-vgpu.mock-device-plugin" -}}
+{{- printf "%s-mock-device-plugin" ( include "hami-vgpu.fullname" . ) | trunc 63 | trimSuffix "-" -}}
+{{- end -}}
+
+{{/*
+The tls secret name for Scheduler
+*/}}
+{{- define "hami-vgpu.scheduler.tls" -}}
+{{- printf "%s-scheduler-tls" ( include "hami-vgpu.fullname" . ) | trunc 63 | trimSuffix "-" -}}
+{{- end -}}
+
+{{/*
+The webhook name
+*/}}
+{{- define "hami-vgpu.scheduler.webhook" -}}
+{{- printf "%s-webhook" ( include "hami-vgpu.fullname" . ) | trunc 63 | trimSuffix "-" -}}
+{{- end -}}
+
+{{/*
+Create chart name and version as used by the chart label.
+*/}}
+{{- define "hami-vgpu.chart" -}}
+{{- printf "%s-%s" .Chart.Name .Chart.Version | replace "+" "_" | trunc 63 | trimSuffix "-" }}
+{{- end }}
+
+{{/*
+Common labels
+*/}}
+{{- define "hami-vgpu.labels" -}}
+helm.sh/chart: {{ include "hami-vgpu.chart" . }}
+{{ include "hami-vgpu.selectorLabels" . }}
+{{- if .Chart.AppVersion }}
+app.kubernetes.io/version: {{ .Chart.AppVersion | quote }}
+{{- end }}
+app.kubernetes.io/managed-by: {{ .Release.Service }}
+{{- end }}
+
+{{/*
+Selector labels
+*/}}
+{{- define "hami-vgpu.selectorLabels" -}}
+app.kubernetes.io/name: {{ include "hami-vgpu.name" . }}
+app.kubernetes.io/instance: {{ .Release.Name }}
+{{- end }}
+
+
+{{/*
+    Resolve the tag for kubeScheduler.
+*/}}
+{{- define "resolvedKubeSchedulerTag" -}}
+{{- if .Values.scheduler.kubeScheduler.image.tag }}
+{{- .Values.scheduler.kubeScheduler.image.tag | trim -}}
+{{- else }}
+{{- include "strippedKubeVersion" . | trim -}}
+{{- end }}
+{{- end }}
+
+{{/*
+    Return the stripped Kubernetes version string by removing extra parts after semantic version number.
+    v1.31.1+k3s1 -> v1.31.1
+    v1.30.8-eks-2d5f260 -> v1.30.8
+    v1.31.1 -> v1.31.1
+*/}}
+{{- define "strippedKubeVersion" -}}
+{{ regexReplaceAll "^(v[0-9]+\\.[0-9]+\\.[0-9]+)(.*)$" .Capabilities.KubeVersion.Version "$1" }}
+{{- end -}}
+
+{{- define "hami.scheduler.kubeScheduler.image" -}}
+{{ include "common.images.image" (dict "imageRoot" .Values.scheduler.kubeScheduler.image "global" .Values.global "tag" (include "resolvedKubeSchedulerTag" .)) }}
+{{- end -}}
+
+{{- define "hami.scheduler.extender.image" -}}
+{{ include "common.images.image" (dict "imageRoot" .Values.scheduler.extender.image "global" .Values.global "tag" .Values.global.imageTag) }}
+{{- end -}}
+
+{{- define "hami.devicePlugin.image" -}}
+{{ include "common.images.image" (dict "imageRoot" .Values.devicePlugin.image "global" .Values.global "tag" .Values.global.imageTag) }}
+{{- end -}}
+
+{{- define "hami.mockDevicePlugin.image" -}}
+{{ include "common.images.image" (dict "imageRoot" .Values.mockDevicePlugin.image "global" .Values.global "tag" .Values.mockDevicePlugin.tag) }}
+{{- end -}}
+
+{{- define "hami.devicePlugin.monitor.image" -}}
+{{ include "common.images.image" (dict "imageRoot" .Values.devicePlugin.monitor.image "global" .Values.global "tag" .Values.global.imageTag) }}
+{{- end -}}
+
+{{- define "hami.scheduler.patch.image" -}}
+{{ include "common.images.image" (dict "imageRoot" .Values.scheduler.patch.image "global" .Values.global) }}
+{{- end -}}
+
+{{- define "hami.scheduler.patch.new.image" -}}
+{{ include "common.images.image" (dict "imageRoot" .Values.scheduler.patch.imageNew "global" .Values.global) }}
+{{- end -}}
+
+{{- define "hami.scheduler.extender.imagePullSecrets" -}}
+{{ include "common.images.pullSecrets" (dict "images" (list .Values.scheduler.extender.image) "global" .Values.global) }}
+{{- end -}}
+
+{{- define "hami.devicePlugin.imagePullSecrets" -}}
+{{ include "common.images.pullSecrets" (dict "images" (list .Values.devicePlugin.image) "global" .Values.global) }}
+{{- end -}}
+
+{{- define "hami.scheduler.patch.imagePullSecrets" -}}
+{{ include "common.images.pullSecrets" (dict "images" (list .Values.scheduler.patch.image) "global" .Values.global) }}
+{{- end -}}
+
+{{- define "hami.scheduler.patch.new.imagePullSecrets" -}}
+{{ include "common.images.pullSecrets" (dict "images" (list .Values.scheduler.patch.imageNew) "global" .Values.global) }}
+{{- end -}}
diff --git a/packages/system/hami/charts/hami/templates/device-plugin/configmap.yaml b/packages/system/hami/charts/hami/templates/device-plugin/configmap.yaml
new file mode 100644
index 00000000..74631e24
--- /dev/null
+++ b/packages/system/hami/charts/hami/templates/device-plugin/configmap.yaml
@@ -0,0 +1,13 @@
+{{- if and .Values.devicePlugin.enabled (not .Values.devicePlugin.nodeConfiguration.externalConfigName) -}}
+apiVersion: v1
+kind: ConfigMap
+metadata:
+  name: {{ include "hami-vgpu.device-plugin" . }}
+  namespace: {{ include "hami-vgpu.namespace" . }}
+  labels:
+    app.kubernetes.io/component: hami-device-plugin
+    {{- include "hami-vgpu.labels" . | nindent 4 }}
+data:
+  config.json: |
+{{- .Values.devicePlugin.nodeConfiguration.config | nindent 4 }}
+{{- end }}
diff --git a/packages/system/hami/charts/hami/templates/device-plugin/daemonsetmock.yaml b/packages/system/hami/charts/hami/templates/device-plugin/daemonsetmock.yaml
new file mode 100644
index 00000000..a0dc4ff3
--- /dev/null
+++ b/packages/system/hami/charts/hami/templates/device-plugin/daemonsetmock.yaml
@@ -0,0 +1,55 @@
+{{- if .Values.mockDevicePlugin.enabled }}
+apiVersion: apps/v1
+kind: DaemonSet
+metadata:
+  name: {{ include "hami-vgpu.mock-device-plugin" . }}
+  namespace: {{ include "hami-vgpu.namespace" . }}
+spec:
+  selector:
+    matchLabels:
+      app.kubernetes.io/component: hami-mock-device-plugin
+      {{- include "hami-vgpu.selectorLabels" . | nindent 6 }}
+  template:
+    metadata:
+      annotations:
+        scheduler.alpha.kubernetes.io/critical-pod: ""
+      labels:
+        app.kubernetes.io/component: hami-mock-device-plugin
+        {{- include "hami-vgpu.selectorLabels" . | nindent 8 }}
+    spec:
+      serviceAccountName: {{ include "hami-vgpu.mock-device-plugin" . }}
+      tolerations:
+      - key: CriticalAddonsOnly
+        operator: Exists
+      containers:
+      - image: {{ include "hami.mockDevicePlugin.image" . }}
+        imagePullPolicy: {{ .Values.mockDevicePlugin.image.pullPolicy }}
+        name: hami-mock-dp-cntr
+        env:
+          - name: NODE_NAME
+            valueFrom:
+              fieldRef:
+                fieldPath: spec.nodeName
+        command:
+          - ./k8s-device-plugin
+          - -v=5
+          - --device-config-file=/device-config.yaml
+        volumeMounts:
+          - name: dp
+            mountPath: /var/lib/kubelet/device-plugins
+          - name: sys
+            mountPath: /sys
+          - name: device-config
+            mountPath: /device-config.yaml
+            subPath: device-config.yaml
+      volumes:
+        - name: dp
+          hostPath:
+            path: /var/lib/kubelet/device-plugins
+        - name: sys
+          hostPath:
+            path: /sys
+        - name: device-config
+          configMap:
+            name: {{ include "hami-vgpu.scheduler" . }}-device
+{{- end -}}
diff --git a/packages/system/hami/charts/hami/templates/device-plugin/daemonsetnvidia.yaml b/packages/system/hami/charts/hami/templates/device-plugin/daemonsetnvidia.yaml
new file mode 100644
index 00000000..1f4c24f3
--- /dev/null
+++ b/packages/system/hami/charts/hami/templates/device-plugin/daemonsetnvidia.yaml
@@ -0,0 +1,262 @@
+{{- if .Values.devicePlugin.enabled }}
+apiVersion: apps/v1
+kind: DaemonSet
+metadata:
+  name: {{ include "hami-vgpu.device-plugin" . }}
+  namespace: {{ include "hami-vgpu.namespace" . }}
+  labels:
+    app.kubernetes.io/component: hami-device-plugin
+    {{- include "hami-vgpu.labels" . | nindent 4 }}
+    {{- with .Values.global.labels }}
+    {{- toYaml . | nindent 4 }}
+    {{- end }}
+  {{- if .Values.global.annotations }}
+  annotations: {{ toYaml .Values.global.annotations | nindent 4}}
+  {{- end }}
+spec:
+  updateStrategy:
+    {{- with .Values.devicePlugin.updateStrategy }}
+    {{- toYaml . | nindent 4 }}
+    {{- end }}
+  selector:
+    matchLabels:
+      app.kubernetes.io/component: hami-device-plugin
+      {{- include "hami-vgpu.selectorLabels" . | nindent 6 }}
+  template:
+    metadata:
+      labels:
+        app.kubernetes.io/component: hami-device-plugin
+        hami.io/webhook: ignore
+        {{- include "hami-vgpu.selectorLabels" . | nindent 8 }}
+      annotations:
+        {{- if ge (regexReplaceAll "[^0-9]" .Capabilities.KubeVersion.Minor "" | int) 22 }}
+        checksum/hami-scheduler-newversion-config: {{ include (print $.Template.BasePath "/scheduler/configmapnew.yaml") . | sha256sum }}
+        {{- else }}
+        checksum/hami-scheduler-config: {{ include (print $.Template.BasePath "/scheduler/configmap.yaml") . | sha256sum }}
+        {{- end }}
+        checksum/hami-scheduler-device-config: {{ include (print $.Template.BasePath "/scheduler/device-configmap.yaml") . | sha256sum }}
+      {{- if .Values.devicePlugin.podAnnotations }}
+        {{- toYaml .Values.devicePlugin.podAnnotations | nindent 8 }}
+      {{- end }}
+    spec:
+      {{- if .Values.devicePlugin.runtimeClassName }}
+      runtimeClassName: {{ .Values.devicePlugin.runtimeClassName }}
+      {{- end }}
+      serviceAccountName: {{ include "hami-vgpu.device-plugin" . }}
+      priorityClassName: system-node-critical
+      hostPID: true
+      hostNetwork: true
+      {{- include "hami.devicePlugin.imagePullSecrets" . | nindent 6 }}
+      {{- if .Values.devicePlugin.gpuOperatorToolkitReady.enabled }}
+      initContainers:
+        - name: toolkit-validation
+          image: {{ include "hami.devicePlugin.image" . }}
+          imagePullPolicy: {{ .Values.devicePlugin.image.pullPolicy }}
+          securityContext:
+            privileged: true
+            runAsUser: 0
+          command: ["sh", "-c"]
+          args:
+            - |
+              echo "Waiting for NVIDIA Toolkit to be ready..."
+              until [ -f {{ .Values.devicePlugin.gpuOperatorToolkitReady.hostPath }}/toolkit-ready ]; do 
+                echo "Waiting for {{ .Values.devicePlugin.gpuOperatorToolkitReady.hostPath }}/toolkit-ready..."
+                sleep 5
+              done
+              echo "NVIDIA Toolkit is ready!"
+          volumeMounts:
+            - name: nvidia-validations
+              mountPath: {{ .Values.devicePlugin.gpuOperatorToolkitReady.hostPath | quote }}
+              mountPropagation: HostToContainer
+              readOnly: true
+      {{- end }}
+      containers:
+        - name: device-plugin
+          image: {{ include "hami.devicePlugin.image" . }}
+          imagePullPolicy: {{ .Values.devicePlugin.image.pullPolicy }}
+          lifecycle:
+            postStart:
+              exec:
+                command: ["/bin/sh","-c", {{ printf "/k8s-vgpu/bin/vgpu-init.sh %s/vgpu/" .Values.global.gpuHookPath | quote }}]
+          command:
+            - nvidia-device-plugin
+            - --config-file=/device-config.yaml
+            - --mig-strategy={{ .Values.devicePlugin.migStrategy }}
+            - --disable-core-limit={{ .Values.devicePlugin.disablecorelimit }}
+            {{- range .Values.devicePlugin.extraArgs }}
+            - {{ . }}
+            {{- end }}
+          env:
+            - name: NODE_NAME
+              valueFrom:
+                fieldRef:
+                  fieldPath: spec.nodeName
+            - name: NVIDIA_MIG_MONITOR_DEVICES
+              value: all
+            - name: DEVICE_LIST_STRATEGY
+              value: {{ .Values.devicePlugin.deviceListStrategy }}
+            - name: HOOK_PATH
+              value: {{ .Values.global.gpuHookPath }}
+            {{- if typeIs "bool" .Values.devicePlugin.passDeviceSpecsEnabled }}
+            - name: PASS_DEVICE_SPECS
+              value: {{ .Values.devicePlugin.passDeviceSpecsEnabled | quote }}
+            {{- end }}
+            {{- if typeIs "string" .Values.devicePlugin.nvidiaDriverRoot }}
+            - name: NVIDIA_DRIVER_ROOT
+              value: {{ .Values.devicePlugin.nvidiaDriverRoot }}
+            {{- end }}
+            {{- if typeIs "string" .Values.devicePlugin.nvidiaHookPath }}
+            - name: NVIDIA_CDI_HOOK_PATH
+              value: {{ .Values.devicePlugin.nvidiaHookPath }}
+            {{- end }}
+            {{- if typeIs "bool" .Values.devicePlugin.gdrcopyEnabled }}
+            - name: GDRCOPY_ENABLED
+              value: {{ .Values.devicePlugin.gdrcopyEnabled | quote }}
+            {{- end }}
+            {{- if typeIs "bool" .Values.devicePlugin.gdsEnabled }}
+            - name: GDS_ENABLED
+              value: {{ .Values.devicePlugin.gdsEnabled | quote }}
+            {{- end }}
+            {{- if typeIs "bool" .Values.devicePlugin.mofedEnabled }}
+            - name: MOFED_ENABLED
+              value: {{ .Values.devicePlugin.mofedEnabled | quote }}
+            {{- end }}
+            {{- if eq (.Values.scheduler.defaultSchedulerPolicy.gpuSchedulerPolicy | default "spread")  "topology-aware" }}
+            - name: ENABLE_TOPOLOGY_SCORE
+              value: "true"
+            {{- end }}
+            {{- with .Values.devicePlugin.extraEnvs }}
+            {{- . | toYaml | nindent 12 }}
+            {{- end }}
+          securityContext:
+            privileged: true
+            allowPrivilegeEscalation: true
+            capabilities:
+              drop: ["ALL"]
+              add: ["SYS_ADMIN"]
+          resources:
+          {{- toYaml .Values.devicePlugin.resources | nindent 12 }}
+          volumeMounts:
+            - name: device-plugin
+              mountPath: /var/lib/kubelet/device-plugins
+            - name: lib
+              mountPath: {{ printf "%s%s" .Values.global.gpuHookPath "/vgpu" }}
+            - name: usrbin
+              mountPath: /usrbin
+            - name: deviceconfig
+              mountPath: /config
+            - name: hosttmp
+              mountPath: /tmp
+            - name: device-config
+              mountPath: /device-config.yaml
+              subPath: device-config.yaml
+            - name: cdi-root
+              mountPath: /var/run/cdi
+            {{- if typeIs "string" .Values.devicePlugin.nvidiaDriverRoot }}
+            # We always mount the driver root at /driver-root in the container.
+            # This is required for CDI detection to work correctly.
+            - name: driver-root
+              mountPath: /driver-root
+              readOnly: true
+            {{- end }}
+        - name: vgpu-monitor
+          image: {{ include "hami.devicePlugin.monitor.image" . }}
+          imagePullPolicy: {{ .Values.devicePlugin.monitor.image.pullPolicy }}
+          command:
+            - "vGPUmonitor"
+            {{- range .Values.devicePlugin.monitor.extraArgs }}
+            - {{ . }}
+            {{- end }}
+          securityContext:
+            allowPrivilegeEscalation: false
+            capabilities:
+              drop: ["ALL"]
+              add: ["SYS_ADMIN"]
+          env:
+            - name: NODE_NAME
+              valueFrom:
+                fieldRef:
+                  fieldPath: spec.nodeName
+            - name: NVIDIA_VISIBLE_DEVICES
+              value: "all"
+            - name: NVIDIA_MIG_MONITOR_DEVICES
+              value: "all"
+            - name: HOOK_PATH
+              value: "{{ .Values.global.gpuHookPath }}/vgpu"
+            - name: HAMI_RESYNC_INTERVAL
+              value: {{ .Values.devicePlugin.monitor.resyncInterval | default "5m" | quote }}
+            {{- with .Values.devicePlugin.monitor.extraEnvs }}
+            {{- . | toYaml | nindent 12 }}
+            {{- end }}
+          resources:
+          {{- toYaml .Values.devicePlugin.monitor.resources | nindent 12 }}
+          volumeMounts:
+            - name: ctrs
+              mountPath: {{ .Values.devicePlugin.monitor.ctrPath }}
+            - name: dockers
+              mountPath: /run/docker
+            - name: containerds
+              mountPath: /run/containerd
+            - name: sysinfo
+              mountPath: /sysinfo
+            - name: hostvar
+              mountPath: /hostvar
+            - name: hosttmp
+              mountPath: /tmp
+      volumes:
+        - name: ctrs
+          hostPath:
+            path: {{ .Values.devicePlugin.monitor.ctrPath }}
+        - name: hosttmp
+          hostPath:
+            path: /tmp
+        - name: dockers
+          hostPath:
+            path: /run/docker
+        - name: containerds
+          hostPath:
+            path: /run/containerd
+        - name: device-plugin
+          hostPath:
+            path: {{ .Values.devicePlugin.pluginPath }}
+        - name: lib
+          hostPath:
+            path: {{ .Values.devicePlugin.libPath }}
+        {{- if .Values.devicePlugin.gpuOperatorToolkitReady.enabled }}
+        - name: nvidia-validations
+          hostPath:
+            path: {{ .Values.devicePlugin.gpuOperatorToolkitReady.hostPath }}
+            type: DirectoryOrCreate
+        {{- end }}
+        {{- if typeIs "string" .Values.devicePlugin.nvidiaDriverRoot }}
+        - name: driver-root
+          hostPath:
+            path: {{ .Values.devicePlugin.nvidiaDriverRoot }}
+            type: Directory
+        {{- end }}
+        - name: cdi-root
+          hostPath:
+            path: /var/run/cdi
+            type: DirectoryOrCreate
+        - name: usrbin
+          hostPath:
+            path: /usr/bin
+        - name: sysinfo
+          hostPath:
+            path: /sys
+        - name: hostvar
+          hostPath:
+            path: /var
+        - name: deviceconfig
+          configMap:
+            name: {{ .Values.devicePlugin.nodeConfiguration.externalConfigName | default (include "hami-vgpu.device-plugin" .) }}
+        - name: device-config
+          configMap:
+            name: {{ include "hami-vgpu.scheduler" . }}-device
+      {{- if .Values.devicePlugin.nvidiaNodeSelector }}
+      nodeSelector: {{ toYaml .Values.devicePlugin.nvidiaNodeSelector | nindent 8 }}
+      {{- end }}
+      {{- if .Values.devicePlugin.tolerations }}
+      tolerations: {{ toYaml .Values.devicePlugin.tolerations | nindent 8 }}
+      {{- end }}
+{{- end }}
diff --git a/packages/system/hami/charts/hami/templates/device-plugin/monitorrole.yaml b/packages/system/hami/charts/hami/templates/device-plugin/monitorrole.yaml
new file mode 100644
index 00000000..6ac757f2
--- /dev/null
+++ b/packages/system/hami/charts/hami/templates/device-plugin/monitorrole.yaml
@@ -0,0 +1,28 @@
+{{- if .Values.devicePlugin.enabled -}}
+apiVersion: rbac.authorization.k8s.io/v1
+kind: ClusterRole
+metadata:
+  name:  {{ include "hami-vgpu.device-plugin" . }}-monitor
+rules:
+  - apiGroups:
+      - ""
+    resources:
+      - pods
+    verbs:
+      - get
+      - create
+      - watch
+      - list
+      - update
+      - patch
+  - apiGroups:
+      - ""
+    resources:
+      - nodes
+    verbs:
+      - get
+      - update
+      - list
+      - patch
+{{- end -}}
+
diff --git a/packages/system/hami/charts/hami/templates/device-plugin/monitorrolebinding.yaml b/packages/system/hami/charts/hami/templates/device-plugin/monitorrolebinding.yaml
new file mode 100644
index 00000000..2f0a14ba
--- /dev/null
+++ b/packages/system/hami/charts/hami/templates/device-plugin/monitorrolebinding.yaml
@@ -0,0 +1,17 @@
+{{- if .Values.devicePlugin.enabled -}}
+apiVersion: rbac.authorization.k8s.io/v1
+kind: ClusterRoleBinding
+metadata:
+  name: {{ include "hami-vgpu.device-plugin" . }}
+  labels:
+    app.kubernetes.io/component: "hami-device-plugin"
+    {{- include "hami-vgpu.labels" . | nindent 4 }}
+roleRef:
+  apiGroup: rbac.authorization.k8s.io
+  kind: ClusterRole
+  name: {{ include "hami-vgpu.device-plugin" . }}-monitor
+subjects:
+  - kind: ServiceAccount
+    name: {{ include "hami-vgpu.device-plugin" . }}
+    namespace: {{ include "hami-vgpu.namespace" . }}
+{{- end -}}
diff --git a/packages/system/hami/charts/hami/templates/device-plugin/monitorservice.yaml b/packages/system/hami/charts/hami/templates/device-plugin/monitorservice.yaml
new file mode 100644
index 00000000..88f1b214
--- /dev/null
+++ b/packages/system/hami/charts/hami/templates/device-plugin/monitorservice.yaml
@@ -0,0 +1,29 @@
+{{- if .Values.devicePlugin.enabled -}}
+apiVersion: v1
+kind: Service
+metadata:
+  name: {{ include "hami-vgpu.device-plugin" . }}-monitor
+  namespace: {{ include "hami-vgpu.namespace" . }}
+  labels:
+    app.kubernetes.io/component: hami-device-plugin
+    {{- include "hami-vgpu.labels" . | nindent 4 }}
+    {{- with .Values.devicePlugin.service.labels }}
+    {{- toYaml . | nindent 4 }}
+    {{- end }}
+  {{- if .Values.devicePlugin.service.annotations }}  # Use devicePlugin instead of scheduler
+  annotations: {{ toYaml .Values.devicePlugin.service.annotations | nindent 4 }}
+  {{- end }}
+spec:
+  type: {{ .Values.devicePlugin.service.type | default "NodePort" }}  # Default type is NodePort
+  ports:
+    - name: monitorport
+      port: {{ .Values.devicePlugin.service.httpPort | default 31992 }}  # Default HTTP port is 31992
+      targetPort: 9394
+      {{- if eq (.Values.devicePlugin.service.type | default "NodePort") "NodePort" }}  # If type is NodePort, set nodePort
+      nodePort: {{ .Values.devicePlugin.service.httpPort | default 31992 }}
+      {{- end }}
+      protocol: TCP
+  selector:
+    app.kubernetes.io/component: hami-device-plugin
+    {{- include "hami-vgpu.selectorLabels" . | nindent 4 }}
+{{- end -}}
diff --git a/packages/system/hami/charts/hami/templates/device-plugin/monitorserviceaccount.yaml b/packages/system/hami/charts/hami/templates/device-plugin/monitorserviceaccount.yaml
new file mode 100644
index 00000000..6e3c2a44
--- /dev/null
+++ b/packages/system/hami/charts/hami/templates/device-plugin/monitorserviceaccount.yaml
@@ -0,0 +1,10 @@
+{{- if .Values.devicePlugin.enabled -}}
+apiVersion: v1
+kind: ServiceAccount
+metadata:
+  name: {{ include "hami-vgpu.device-plugin" . }}
+  namespace: {{ include "hami-vgpu.namespace" . }}
+  labels:
+    app.kubernetes.io/component: "hami-device-plugin"
+    {{- include "hami-vgpu.labels" . | nindent 4 }}
+{{- end -}}
diff --git a/packages/system/hami/charts/hami/templates/device-plugin/runtime-class.yaml b/packages/system/hami/charts/hami/templates/device-plugin/runtime-class.yaml
new file mode 100644
index 00000000..ebcc2c9c
--- /dev/null
+++ b/packages/system/hami/charts/hami/templates/device-plugin/runtime-class.yaml
@@ -0,0 +1,9 @@
+{{- if and .Values.devicePlugin.enabled .Values.devicePlugin.createRuntimeClass .Values.devicePlugin.runtimeClassName -}}
+apiVersion: node.k8s.io/v1
+kind: RuntimeClass
+metadata:
+  name: {{ .Values.devicePlugin.runtimeClassName }}
+  annotations:
+    helm.sh/hook: pre-install,pre-upgrade
+handler: nvidia
+{{- end -}}
diff --git a/packages/system/hami/charts/hami/templates/scheduler/certmanager.yaml b/packages/system/hami/charts/hami/templates/scheduler/certmanager.yaml
new file mode 100644
index 00000000..e6d28721
--- /dev/null
+++ b/packages/system/hami/charts/hami/templates/scheduler/certmanager.yaml
@@ -0,0 +1,31 @@
+{{- if .Values.scheduler.admissionWebhook.enabled -}}
+{{- if .Values.scheduler.certManager.enabled }}
+apiVersion: cert-manager.io/v1
+kind: Certificate
+metadata:
+  name: {{ include "hami-vgpu.scheduler" . }}-serving-cert
+  namespace: {{ include "hami-vgpu.namespace" . }}
+  labels:
+    app.kubernetes.io/component: hami-scheduler
+    {{- include "hami-vgpu.labels" . | nindent 4 }}
+spec:
+  dnsNames:
+    - {{ include "hami-vgpu.scheduler" . }}.{{ include "hami-vgpu.namespace" . }}.svc
+    - {{ include "hami-vgpu.scheduler" . }}.{{ include "hami-vgpu.namespace" . }}.svc.cluster.local
+  issuerRef:
+    kind: Issuer
+    name: {{ include "hami-vgpu.scheduler" . }}-selfsigned-issuer
+  secretName: {{ include "hami-vgpu.scheduler.tls" . }}
+---
+apiVersion: cert-manager.io/v1
+kind: Issuer
+metadata:
+  name: {{ include "hami-vgpu.scheduler" . }}-selfsigned-issuer
+  namespace: {{ include "hami-vgpu.namespace" . }}
+  labels:
+    app.kubernetes.io/component: hami-scheduler
+    {{- include "hami-vgpu.labels" . | nindent 4 }}
+spec:
+  selfSigned: {}
+{{- end }}
+{{- end }}
diff --git a/packages/system/hami/charts/hami/templates/scheduler/clusterrole.yaml b/packages/system/hami/charts/hami/templates/scheduler/clusterrole.yaml
new file mode 100644
index 00000000..81c4fddf
--- /dev/null
+++ b/packages/system/hami/charts/hami/templates/scheduler/clusterrole.yaml
@@ -0,0 +1,34 @@
+apiVersion: rbac.authorization.k8s.io/v1
+kind: ClusterRole
+metadata:
+  name: {{ include "hami-vgpu.scheduler" . }}
+  labels:
+    app.kubernetes.io/component: "hami-scheduler"
+    {{- include "hami-vgpu.labels" . | nindent 4 }}
+rules:
+  - apiGroups: [""]
+    resources: ["pods", "configmaps"]
+    verbs: ["get", "list", "watch", "patch"]
+  - apiGroups: [""]
+    resources: ["pods/binding"]
+    verbs: ["create"]
+  - apiGroups: [""]
+    resources: ["nodes"]
+    verbs: ["get", "list", "patch", "watch"]
+  - apiGroups: [""]
+    resources: ["events"]
+    verbs: ["create", "get", "list"]
+  - apiGroups: [""]
+    resources: ["resourcequotas"]
+    verbs: ["get", "list", "watch"]
+---
+{{- if .Values.mockDevicePlugin.enabled }}
+apiVersion: rbac.authorization.k8s.io/v1
+kind: ClusterRole
+metadata:
+  name: {{ include "hami-vgpu.mock-device-plugin" . }}
+rules:
+  - apiGroups: [""]
+    resources: ["nodes"]
+    verbs: ["get", "update", "list", "patch"]
+{{- end -}}
diff --git a/packages/system/hami/charts/hami/templates/scheduler/clusterrolebinding.yaml b/packages/system/hami/charts/hami/templates/scheduler/clusterrolebinding.yaml
new file mode 100644
index 00000000..a81d425c
--- /dev/null
+++ b/packages/system/hami/charts/hami/templates/scheduler/clusterrolebinding.yaml
@@ -0,0 +1,47 @@
+apiVersion: rbac.authorization.k8s.io/v1
+kind: ClusterRoleBinding
+metadata:
+  name: {{ include "hami-vgpu.scheduler" . }}-kube
+  labels:
+    app.kubernetes.io/component: "hami-scheduler"
+    {{- include "hami-vgpu.labels" . | nindent 4 }}
+roleRef:
+  apiGroup: rbac.authorization.k8s.io
+  kind: ClusterRole
+  name: system:kube-scheduler
+subjects:
+  - kind: ServiceAccount
+    name: {{ include "hami-vgpu.scheduler" . }}
+    namespace: {{ include "hami-vgpu.namespace" . }}
+---
+apiVersion: rbac.authorization.k8s.io/v1
+kind: ClusterRoleBinding
+metadata:
+  name: {{ include "hami-vgpu.scheduler" . }}-volume
+  labels:
+    app.kubernetes.io/component: "hami-scheduler"
+    {{- include "hami-vgpu.labels" . | nindent 4 }}
+roleRef:
+  apiGroup: rbac.authorization.k8s.io
+  kind: ClusterRole
+  name: system:volume-scheduler
+subjects:
+  - kind: ServiceAccount
+    name: {{ include "hami-vgpu.scheduler" . }}
+    namespace: {{ include "hami-vgpu.namespace" . }}
+---
+apiVersion: rbac.authorization.k8s.io/v1
+kind: ClusterRoleBinding
+metadata:
+  name: {{ include "hami-vgpu.scheduler" . }}
+  labels:
+    app.kubernetes.io/component: "hami-scheduler"
+    {{- include "hami-vgpu.labels" . | nindent 4 }}
+roleRef:
+  apiGroup: rbac.authorization.k8s.io
+  kind: ClusterRole
+  name: {{ include "hami-vgpu.scheduler" . }}
+subjects:
+  - kind: ServiceAccount
+    name: {{ include "hami-vgpu.scheduler" . }}
+    namespace: {{ include "hami-vgpu.namespace" . }}
diff --git a/packages/system/hami/charts/hami/templates/scheduler/configmap.yaml b/packages/system/hami/charts/hami/templates/scheduler/configmap.yaml
new file mode 100644
index 00000000..109ecbce
--- /dev/null
+++ b/packages/system/hami/charts/hami/templates/scheduler/configmap.yaml
@@ -0,0 +1,142 @@
+{{- if .Values.scheduler.kubeScheduler.enabled -}}
+apiVersion: v1
+kind: ConfigMap
+metadata:
+  name: {{ include "hami-vgpu.scheduler" . }}
+  namespace: {{ include "hami-vgpu.namespace" . }}
+  labels:
+    app.kubernetes.io/component: hami-scheduler
+    {{- include "hami-vgpu.labels" . | nindent 4 }}
+data:
+  config.json: |
+    {
+        "kind": "Policy",
+        "apiVersion": "v1",
+        "extenders": [
+            {
+                {{- if .Values.scheduler.admissionWebhook.enabled }}
+                "urlPrefix": "https://127.0.0.1:443",
+                "enableHttps": true,
+                "tlsConfig": {
+                    "insecure": true
+                },
+                {{- else }}
+                "urlPrefix": "http://127.0.0.1:80",
+                "enableHttps": false,
+                {{- end }}
+                "filterVerb": "filter",
+                "bindVerb": "bind",
+                "weight": 1,
+                "nodeCacheCapable": true,
+                "httpTimeout": 30000000000,
+                "managedResources": [
+                    {{- range .Values.devices.amd.customresources }}
+                    {
+                      "name": "{{ . }}",
+                      "ignoredByScheduler": true
+                    },
+                    {{- end }}
+                    {{- if .Values.devices.ascend.enabled }}
+                    {{- range .Values.devices.ascend.customresources }}
+                    {
+                      "name": "{{ . }}",
+                      "ignoredByScheduler": true
+                    },
+                    {{- end }}
+                    {{- end }}
+                    {{- if .Values.devices.mthreads.enabled }}
+                    {{- range .Values.devices.mthreads.customresources }}
+                    {
+                      "name": "{{ . }}",
+                      "ignoredByScheduler": true
+                    },
+                    {{- end }}
+                    {{- end }}
+                    {{- if .Values.devices.enflame.enabled }}
+                    {{- range .Values.devices.enflame.customresources }}
+                    {
+                      "name": "{{ . }}",
+                      "ignoredByScheduler": true
+                    },
+                    {{- end }}
+                    {{- end }}
+                    {{- if .Values.devices.kunlun.enabled }}
+                    {{- range .Values.devices.kunlun.customresources }}
+                    {
+                      "name": "{{ . }}",
+                      "ignoredByScheduler": true
+                    },
+                    {{- end }}
+                    {{- end }}
+                    {{- range .Values.devices.awsneuron.customresources }}
+                    {
+                      "name": "{{ . }}",
+                      "ignoredByScheduler": true
+                    },
+                    {{- end }}
+                    {{- if .Values.devices.iluvatar.enabled }}
+                    {{- range .Values.devices.iluvatar.customresources }}
+                    {
+                      "name": "{{ . }}",
+                      "ignoredByScheduler": true
+                    },
+                    {{- end }}
+                    {{- end }}
+                    {
+                        "name": "{{ .Values.resourceName }}",
+                        "ignoredByScheduler": true
+                    },
+                    {
+                        "name": "{{ .Values.resourceMem }}",
+                        "ignoredByScheduler": true
+                    },
+                    {
+                        "name": "{{ .Values.resourceCores }}",
+                        "ignoredByScheduler": true
+                    },
+                    {
+                        "name": "{{ .Values.resourceMemPercentage }}",
+                        "ignoredByScheduler": true
+                    },
+                    {
+                        "name": "{{ .Values.resourcePriority }}",
+                        "ignoredByScheduler": true
+                    },
+                    {
+                        "name": "{{ .Values.mluResourceName }}",
+                        "ignoredByScheduler": true
+                    },
+                    {
+                        "name": "{{ .Values.dcuResourceName }}",
+                        "ignoredByScheduler": true
+                    },
+                    {
+                        "name": "{{ .Values.dcuResourceMem }}",
+                        "ignoredByScheduler": true 
+                    },
+                    {
+                        "name": "{{ .Values.dcuResourceCores }}",
+                        "ignoredByScheduler": true
+                    },
+                    {
+                        "name": "metax-tech.com/gpu",
+                        "ignoredByScheduler": true
+                    },
+                    {
+                        "name": "{{ .Values.metaxResourceName }}",
+                        "ignoredByScheduler": true
+                    },
+                    {
+                        "name": "{{ .Values.metaxResourceCore }}",
+                        "ignoredByScheduler": true
+                    },
+                    {
+                        "name": "{{ .Values.metaxResourceMem }}",
+                        "ignoredByScheduler": true
+                    }
+                ],
+                "ignoreable": false
+            }
+        ]
+    }
+{{- end }}
diff --git a/packages/system/hami/charts/hami/templates/scheduler/configmapnew.yaml b/packages/system/hami/charts/hami/templates/scheduler/configmapnew.yaml
new file mode 100644
index 00000000..6f6db097
--- /dev/null
+++ b/packages/system/hami/charts/hami/templates/scheduler/configmapnew.yaml
@@ -0,0 +1,102 @@
+{{- if .Values.scheduler.kubeScheduler.enabled -}}
+apiVersion: v1
+kind: ConfigMap
+metadata:
+  name: {{ include "hami-vgpu.scheduler" . }}-newversion
+  namespace: {{ include "hami-vgpu.namespace" . }}
+  labels:
+    app.kubernetes.io/component: hami-scheduler
+    {{- include "hami-vgpu.labels" . | nindent 4 }}
+data:
+  config.yaml: |
+    {{- if gt (regexReplaceAll "[^0-9]" .Capabilities.KubeVersion.Minor "" | int) 25}}
+    apiVersion: kubescheduler.config.k8s.io/v1
+    {{- else }}
+    apiVersion: kubescheduler.config.k8s.io/v1beta2
+    {{- end }}
+    kind: KubeSchedulerConfiguration
+    leaderElection:
+      leaderElect: false
+    profiles:
+    - schedulerName: {{ .Values.schedulerName }}
+    extenders:
+    {{- if .Values.scheduler.admissionWebhook.enabled }}
+    - urlPrefix: "https://127.0.0.1:443"
+      enableHTTPS: true
+      tlsConfig:
+        insecure: true
+    {{- else }}
+    - urlPrefix: "http://127.0.0.1:80"
+      enableHTTPS: false
+    {{- end }}
+      filterVerb: filter
+      bindVerb: bind
+      nodeCacheCapable: true
+      weight: 1
+      httpTimeout: 30s
+      managedResources:
+      - name: {{ .Values.resourceName }}
+        ignoredByScheduler: true
+      - name: {{ .Values.resourceMem }}
+        ignoredByScheduler: true
+      - name: {{ .Values.resourceCores }}
+        ignoredByScheduler: true
+      - name: {{ .Values.resourceMemPercentage }}
+        ignoredByScheduler: true
+      - name: {{ .Values.resourcePriority }}
+        ignoredByScheduler: true
+      - name: {{ .Values.mluResourceName }}
+        ignoredByScheduler: true
+      - name: {{ .Values.dcuResourceName }}
+        ignoredByScheduler: true
+      - name: {{ .Values.dcuResourceMem }}
+        ignoredByScheduler: true
+      - name: {{ .Values.dcuResourceCores }}
+        ignoredByScheduler: true
+      - name: "metax-tech.com/gpu"
+        ignoredByScheduler: true
+      - name: {{ .Values.metaxResourceName }}
+        ignoredByScheduler: true
+      - name: {{ .Values.metaxResourceCore }}
+        ignoredByScheduler: true
+      - name: {{ .Values.metaxResourceMem }}
+        ignoredByScheduler: true
+      {{- if .Values.devices.ascend.enabled }}
+      {{- range .Values.devices.ascend.customresources }}
+      - name: {{ . }}
+        ignoredByScheduler: true
+      {{- end }}
+      {{- end }}
+      {{- if .Values.devices.mthreads.enabled }}
+      {{- range .Values.devices.mthreads.customresources }}
+      - name: {{ . }}
+        ignoredByScheduler: true
+      {{- end }}
+      {{- end }}
+      {{- if .Values.devices.enflame.enabled }}
+      {{- range .Values.devices.enflame.customresources }}
+      - name: {{ . }}
+        ignoredByScheduler: true
+      {{- end }}
+      {{- end }}
+      {{- if .Values.devices.kunlun.enabled }}
+      {{- range .Values.devices.kunlun.customresources }}
+      - name: {{ . }}
+        ignoredByScheduler: true
+      {{- end }}
+      {{- end }}
+      {{- range .Values.devices.awsneuron.customresources }}
+      - name: {{ . }}
+        ignoredByScheduler: true
+      {{- end }}
+      {{- if .Values.devices.iluvatar.enabled }}
+      {{- range .Values.devices.iluvatar.customresources }}
+      - name: {{ . }}
+        ignoredByScheduler: true
+      {{- end }}
+      {{- end }}
+      {{- range .Values.devices.amd.customresources }}
+      - name: {{ . }}
+        ignoredByScheduler: true
+      {{- end }}
+{{- end }}
diff --git a/packages/system/hami/charts/hami/templates/scheduler/deployment.yaml b/packages/system/hami/charts/hami/templates/scheduler/deployment.yaml
new file mode 100644
index 00000000..1d89e189
--- /dev/null
+++ b/packages/system/hami/charts/hami/templates/scheduler/deployment.yaml
@@ -0,0 +1,225 @@
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: {{ include "hami-vgpu.scheduler" . }}
+  namespace: {{ include "hami-vgpu.namespace" . }}
+  labels:
+    app.kubernetes.io/component: hami-scheduler
+    {{- include "hami-vgpu.labels" . | nindent 4 }}
+    {{- with .Values.global.labels }}
+    {{- toYaml . | nindent 4 }}
+    {{- end }}
+  {{- if .Values.global.annotations }}
+  annotations: {{ toYaml .Values.global.annotations | nindent 4}}
+  {{- end }}
+spec:
+  {{- if .Values.scheduler.leaderElect }}
+  replicas: {{ .Values.scheduler.replicas }}
+  {{- else }}
+  replicas: 1
+  {{- end }}
+  selector:
+    matchLabels:
+      app.kubernetes.io/component: hami-scheduler
+      {{- include "hami-vgpu.selectorLabels" . | nindent 6 }}
+  template:
+    metadata:
+      labels:
+        app.kubernetes.io/component: hami-scheduler
+        {{- include "hami-vgpu.selectorLabels" . | nindent 8 }}
+        hami.io/webhook: ignore
+      annotations:
+        {{- if ge (regexReplaceAll "[^0-9]" .Capabilities.KubeVersion.Minor "" | int) 22 }}
+        checksum/hami-scheduler-newversion-config: {{ include (print $.Template.BasePath "/scheduler/configmapnew.yaml") . | sha256sum }}
+        {{- else }}
+        checksum/hami-scheduler-config: {{ include (print $.Template.BasePath "/scheduler/configmap.yaml") . | sha256sum }}
+        {{- end }}
+        checksum/hami-scheduler-device-config: {{ include (print $.Template.BasePath "/scheduler/device-configmap.yaml") . | sha256sum }}
+      {{- if .Values.scheduler.podAnnotations }}
+        {{- toYaml .Values.scheduler.podAnnotations | nindent 8 }}
+      {{- end }}
+    spec:
+      serviceAccountName: {{ include "hami-vgpu.scheduler" . }}
+      priorityClassName: system-node-critical
+      {{- include "hami.scheduler.extender.imagePullSecrets" . | nindent 6 }}
+      containers:
+        {{- if .Values.scheduler.kubeScheduler.enabled }}
+        - name: kube-scheduler
+          image: {{ include "hami.scheduler.kubeScheduler.image" . }}
+          imagePullPolicy: {{ .Values.scheduler.kubeScheduler.image.pullPolicy }}
+          command:
+            - kube-scheduler
+            {{- if ge (regexReplaceAll "[^0-9]" .Capabilities.KubeVersion.Minor "" | int) 22 }}
+            {{- range .Values.scheduler.kubeScheduler.extraNewArgs }}
+            - {{ . }}
+            {{- end }}
+            {{- else }}
+            - --scheduler-name={{ .Values.schedulerName }}
+            {{- range .Values.scheduler.kubeScheduler.extraArgs }}
+            - {{ . }}
+            {{- end }}
+            {{- end }}
+            - --leader-elect={{ .Values.scheduler.leaderElect }}
+            - --leader-elect-resource-name={{ .Values.schedulerName }}
+            - --leader-elect-resource-namespace={{ include "hami-vgpu.namespace" . }}
+          resources:
+          {{- toYaml .Values.scheduler.kubeScheduler.resources | nindent 12 }}
+          volumeMounts:
+            - name: scheduler-config
+              mountPath: /config
+        {{- end }}
+          {{- if .Values.scheduler.livenessProbe }}
+          livenessProbe:
+            failureThreshold: 8
+            httpGet:
+              path: /healthz
+              port: 10259
+              scheme: HTTPS
+            initialDelaySeconds: 10
+            periodSeconds: 10
+            successThreshold: 1
+            timeoutSeconds: 15
+          {{- end }}
+        - name: vgpu-scheduler-extender
+          image: {{ include "hami.scheduler.extender.image" . }}
+          imagePullPolicy: {{ .Values.scheduler.extender.image.pullPolicy }}
+          env:
+          {{- if .Values.scheduler.nodeLockExpire }}
+            - name: HAMI_NODELOCK_EXPIRE
+              value: "{{ .Values.scheduler.nodeLockExpire }}"
+          {{- end }}
+          {{- if .Values.global.managedNodeSelectorEnable }}
+          {{- range $key, $value := .Values.global.managedNodeSelector }}
+            - name: NODE_SELECTOR_{{ $key | upper | replace "-" "_" }}
+              value: "{{ $value }}"
+          {{- end }}
+          {{- end }}
+          command:
+            - scheduler
+            {{- if .Values.scheduler.admissionWebhook.enabled }}
+            - --http_bind=0.0.0.0:443
+            - --cert_file=/tls/tls.crt
+            - --key_file=/tls/tls.key
+            {{- else }}
+            - --http_bind=0.0.0.0:80
+            {{- end }}
+            - --scheduler-name={{ .Values.schedulerName }}
+            - --metrics-bind-address={{ .Values.scheduler.metricsBindAddress }}
+            - --node-scheduler-policy={{ .Values.scheduler.defaultSchedulerPolicy.nodeSchedulerPolicy }}
+            - --gpu-scheduler-policy={{ .Values.scheduler.defaultSchedulerPolicy.gpuSchedulerPolicy }}
+            - --force-overwrite-default-scheduler={{ .Values.scheduler.forceOverwriteDefaultScheduler}}
+            - --device-config-file=/device-config.yaml
+            - --leader-elect={{ .Values.scheduler.leaderElect }}
+            - --leader-elect-resource-name={{ .Values.schedulerName }}
+            - --leader-elect-resource-namespace={{ include "hami-vgpu.namespace" . }}
+            {{- if .Values.devices.ascend.enabled }}
+            - --enable-ascend=true
+            {{- end }}
+            {{- if .Values.devices.iluvatar.enabled }}
+            - --enable-iluvatar=true
+            {{- end }}
+            {{- if .Values.scheduler.nodeLabelSelector }}
+            - --node-label-selector={{- $first := true -}}
+              {{- range $key, $value := .Values.scheduler.nodeLabelSelector -}}
+              {{- if not $first }},{{ end -}}
+              {{- $key }}={{ $value -}}
+              {{- $first = false -}}
+              {{- end -}}
+            {{- end }}
+            {{- range .Values.scheduler.extender.extraArgs }}
+            - {{ . }}
+            {{- end }}
+          ports:
+            {{- if .Values.scheduler.admissionWebhook.enabled }}
+            - name: https
+              containerPort: 443
+              protocol: TCP
+            {{- else }}
+            - name: http
+              containerPort: 80
+              protocol: TCP
+            {{- end }}
+            - name: metrics
+              containerPort: {{ last (splitList ":" .Values.scheduler.metricsBindAddress) | int }}
+              protocol: TCP
+          resources:
+          {{- toYaml .Values.scheduler.extender.resources | nindent 12 }}
+          volumeMounts:
+            - name: device-config
+              mountPath: /device-config.yaml
+              subPath: device-config.yaml
+          {{- if .Values.scheduler.admissionWebhook.enabled }}
+            - name: tls-config
+              mountPath: /tls
+          {{- end }}
+          {{- if .Values.scheduler.livenessProbe }}
+          livenessProbe:
+            httpGet:
+              path: /healthz
+              {{- if .Values.scheduler.admissionWebhook.enabled }}
+              port: https
+              scheme: HTTPS
+              {{- else }}
+              port: http
+              scheme: HTTP
+              {{- end }}
+            initialDelaySeconds: 10
+            periodSeconds: 10
+            failureThreshold: 3
+            timeoutSeconds: 5
+          {{- end }}
+          {{- if .Values.scheduler.leaderElect }}
+          readinessProbe:
+            httpGet:
+              path: /readyz
+              {{- if .Values.scheduler.admissionWebhook.enabled }}
+              port: https
+              scheme: HTTPS
+              {{- else }}
+              port: http
+              scheme: HTTP
+              {{- end }}
+            initialDelaySeconds: 10
+            periodSeconds: 10
+            failureThreshold: 3
+            timeoutSeconds: 5
+          {{- end }}
+      {{- if .Values.scheduler.leaderElect }}
+      affinity:
+        podAntiAffinity:
+          requiredDuringSchedulingIgnoredDuringExecution:
+          - labelSelector:
+              matchExpressions:
+              - key: app.kubernetes.io/component
+                operator: In
+                values:
+                - hami-scheduler
+            topologyKey: "kubernetes.io/hostname"
+      {{- end }}
+      volumes:
+        {{- if .Values.scheduler.admissionWebhook.enabled }}
+        - name: tls-config
+          secret:
+            secretName: {{ template "hami-vgpu.scheduler.tls" . }}
+        {{- end }}
+        {{- if .Values.scheduler.kubeScheduler.enabled }}
+        - name: scheduler-config
+          configMap:
+            {{- if ge (regexReplaceAll "[^0-9]" .Capabilities.KubeVersion.Minor "" | int) 22 }}
+            name: {{ template "hami-vgpu.scheduler" . }}-newversion
+            {{- else }}
+            name: {{ template "hami-vgpu.scheduler" . }}
+            {{- end }}
+        {{- end }}
+        - name: device-config
+          configMap:
+            name: {{ include "hami-vgpu.scheduler" . }}-device
+      {{- if .Values.scheduler.nodeSelector }}
+      nodeSelector: {{ toYaml .Values.scheduler.nodeSelector | nindent 8 }}
+      {{- end }}
+      {{- if .Values.scheduler.tolerations }}
+      tolerations: {{ toYaml .Values.scheduler.tolerations | nindent 8 }}
+      {{- end }}
+      {{- if .Values.scheduler.nodeName }}
+      nodeName: {{ .Values.scheduler.nodeName }}
+      {{- end }}
diff --git a/packages/system/hami/charts/hami/templates/scheduler/device-configmap.yaml b/packages/system/hami/charts/hami/templates/scheduler/device-configmap.yaml
new file mode 100644
index 00000000..873b813b
--- /dev/null
+++ b/packages/system/hami/charts/hami/templates/scheduler/device-configmap.yaml
@@ -0,0 +1,408 @@
+apiVersion: v1
+kind: ConfigMap
+metadata:
+  name: {{ include "hami-vgpu.scheduler" . }}-device
+  namespace: {{ include "hami-vgpu.namespace" . }}
+  labels:
+    app.kubernetes.io/component: hami-scheduler
+    {{- include "hami-vgpu.labels" . | nindent 4 }}
+data:
+  device-config.yaml: |-
+  {{- if .Files.Glob "files/device-config.yaml" }}
+  {{- .Files.Get "files/device-config.yaml" | nindent 4}}
+  {{- else }}
+    nvidia:
+      resourceCountName: {{ .Values.resourceName }}
+      resourceMemoryName: {{ .Values.resourceMem }}
+      resourceMemoryPercentageName: {{ .Values.resourceMemPercentage }}
+      resourceCoreName: {{ .Values.resourceCores }}
+      resourcePriorityName: {{ .Values.resourcePriority }}
+      overwriteEnv: false
+      defaultMemory: 0
+      defaultCores: 0
+      defaultGPUNum: 1
+      memoryFactor: 1
+      deviceSplitCount: {{ .Values.devicePlugin.deviceSplitCount }}
+      deviceMemoryScaling: {{ .Values.devicePlugin.deviceMemoryScaling }}
+      deviceCoreScaling: {{ .Values.devicePlugin.deviceCoreScaling }}
+      gpuCorePolicy: {{ .Values.devices.nvidia.gpuCorePolicy }}
+      libCudaLogLevel: {{ .Values.devices.nvidia.libCudaLogLevel }}
+      runtimeClassName: "{{ .Values.devicePlugin.runtimeClassName }}"
+      knownMigGeometries:
+      - models: [ "A30" ]
+        allowedGeometries:
+          - 
+            - name: 1g.6gb
+              core: 25
+              memory: 6144
+              count: 4
+          - 
+            - name: 2g.12gb
+              core: 50
+              memory: 12288
+              count: 2
+          - 
+            - name: 4g.24gb
+              core: 100
+              memory: 24576
+              count: 1
+      - models: [ "A100-SXM4-40GB", "A100-40GB-PCIe", "A100-PCIE-40GB"]
+        allowedGeometries:
+          - 
+            - name: 1g.5gb
+              core: 14
+              memory: 5120
+              count: 7
+          - 
+            - name: 1g.5gb
+              core: 14
+              memory: 5120
+              count: 1
+            - name: 2g.10gb
+              core: 28
+              memory: 10240
+              count: 3
+          - 
+            - name: 3g.20gb
+              core: 42
+              memory: 20480
+              count: 2
+          - 
+            - name: 7g.40gb
+              core: 100
+              memory: 40960
+              count: 1
+      - models: [ "A100-SXM4-80GB", "A100-80GB-PCIe", "A100-PCIE-80GB"]
+        allowedGeometries:
+          - 
+            - name: 1g.10gb
+              core: 14
+              memory: 10240
+              count: 7
+          - 
+            - name: 1g.10gb
+              core: 14
+              memory: 10240
+              count: 1
+            - name: 2g.20gb
+              core: 28
+              memory: 20480
+              count: 3
+          - 
+            - name: 3g.40gb
+              core: 42
+              memory: 40960
+              count: 2
+          - 
+            - name: 7g.79gb
+              core: 100
+              memory: 80896
+              count: 1
+      - models: [ "H100-PCIE-80GB", "H100-SXM5-80GB"]
+        allowedGeometries:
+          - 
+            - name: 1g.10gb
+              core: 14
+              memory: 10240
+              count: 7
+          - 
+            - name: 1g.10gb
+              core: 14
+              memory: 10240
+              count: 1
+            - name: 2g.20gb
+              core: 28
+              memory: 20480
+              count: 3
+          - 
+            - name: 3g.40gb
+              core: 42
+              memory: 40960
+              count: 2
+          - 
+            - name: 7g.80gb
+              core: 100
+              memory: 81920
+              count: 1
+      - models: [ "H100-PCIE-94GB", "H100-SXM5-94GB"]
+        allowedGeometries:
+          - 
+            - name: 1g.12gb
+              core: 14
+              memory: 12288
+              count: 7
+          - 
+            - name: 1g.12gb
+              core: 14
+              memory: 12288
+              count: 1
+            - name: 2g.24gb
+              core: 28
+              memory: 24576
+              count: 3
+          - 
+            - name: 3g.47gb
+              core: 42
+              memory: 48128
+              count: 2
+          - 
+            - name: 7g.94gb
+              core: 100
+              memory: 96256
+              count: 1
+      - models: [ "H20", "H100 on GH200"]
+        allowedGeometries:
+          - 
+            - name: 1g.12gb
+              core: 14
+              memory: 12288
+              count: 7
+          - 
+            - name: 1g.12gb
+              core: 14
+              memory: 12288
+              count: 1
+            - name: 2g.24gb
+              core: 28
+              memory: 24576
+              count: 3 
+          - 
+            - name: 3g.48gb
+              core: 42
+              memory: 49152
+              count: 2
+          - 
+            - name: 7g.96gb
+              core: 100
+              memory: 98304
+              count: 1
+      - models: [ "H200 NVL", "H200-SXM5"]
+        allowedGeometries:
+          - 
+            - name: 1g.18gb
+              core: 14
+              memory: 18432
+              count: 7
+          - 
+            - name: 1g.18gb
+              core: 14
+              memory: 18432
+              count: 1
+            - name: 2g.35gb
+              core: 28
+              memory: 35840
+              count: 3
+          - 
+            - name: 3g.71gb
+              core: 42
+              memory: 72704
+              count: 2
+          - 
+            - name: 7g.141gb
+              core: 100
+              memory: 144384
+              count: 1
+      - models: [ "B200" ]
+        allowedGeometries:
+          - 
+            - name: 1g.23gb
+              core: 14
+              memory: 23552
+              count: 7
+          - 
+            - name: 1g.23gb
+              core: 14
+              memory: 23552
+              count: 1
+            - name: 2g.45gb
+              core: 28
+              memory: 46080
+              count: 3
+          - 
+            - name: 3g.90gb
+              core: 42
+              memory: 92160
+              count: 2
+          - 
+            - name: 7g.180gb
+              core: 100
+              memory: 184320
+              count: 1
+    cambricon:
+      resourceCountName: {{ .Values.mluResourceName }}
+      resourceMemoryName: {{ .Values.mluResourceMem }}
+      resourceCoreName: {{ .Values.mluResourceCores }}
+    hygon:
+      resourceCountName: {{ .Values.dcuResourceName }}
+      resourceMemoryName: {{ .Values.dcuResourceMem }}
+      resourceCoreName: {{ .Values.dcuResourceCores }}
+      memoryFactor: 1
+    metax:
+      resourceCountName: "metax-tech.com/gpu"
+      resourceVCountName: {{ .Values.metaxResourceName }}
+      resourceVMemoryName: {{ .Values.metaxResourceMem }}
+      resourceVCoreName: {{ .Values.metaxResourceCore }}
+      sgpuTopologyAware: {{ .Values.metaxsGPUTopologyAware }}
+    enflame:
+      resourceNameGCU: "enflame.com/gcu"
+      resourceNameVGCU: {{ .Values.enflameResourceNameVGCU }}
+      resourceNameVGCUPercentage: {{ .Values.enflameResourceNameVGCUPercentage }}
+    mthreads:
+      resourceCountName: "mthreads.com/vgpu"
+      resourceMemoryName: "mthreads.com/sgpu-memory"
+      resourceCoreName: "mthreads.com/sgpu-core"
+    iluvatars:
+    - chipName: MR-V100
+      commonWord: MR-V100
+      resourceCountName: iluvatar.ai/MR-V100-vgpu
+      resourceMemoryName: iluvatar.ai/MR-V100.vMem
+      resourceCoreName: iluvatar.ai/MR-V100.vCore
+    - chipName: MR-V50
+      commonWord: MR-V50
+      resourceCountName: iluvatar.ai/MR-V50-vgpu
+      resourceMemoryName: iluvatar.ai/MR-V50.vMem
+      resourceCoreName: iluvatar.ai/MR-V50.vCore
+    - chipName: BI-V150
+      commonWord: BI-V150
+      resourceCountName: iluvatar.ai/BI-V150-vgpu
+      resourceMemoryName: iluvatar.ai/BI-V150.vMem
+      resourceCoreName: iluvatar.ai/BI-V150.vCore
+    - chipName: BI-V100
+      commonWord: BI-V100
+      resourceCountName: iluvatar.ai/BI-V100-vgpu
+      resourceMemoryName: iluvatar.ai/BI-V100.vMem
+      resourceCoreName: iluvatar.ai/BI-V100.vCore
+    kunlun:
+      resourceCountName: {{ .Values.kunlunResourceName }}
+      resourceVCountName: {{ .Values.kunlunResourceVCountName }}
+      resourceVMemoryName: {{ .Values.kunlunResourceVMemoryName }}
+    awsneuron:
+      resourceCountName: "aws.amazon.com/neuron"
+      resourceCoreName: "aws.amazon.com/neuroncore"
+    amd:
+      resourceCountName: "amd.com/gpu"
+    vnpus:
+    - chipName: 910A
+      commonWord: Ascend910A
+      resourceName: huawei.com/Ascend910A
+      resourceMemoryName: huawei.com/Ascend910A-memory
+      memoryAllocatable: 32768
+      memoryCapacity: 32768
+      memoryFactor: 1
+      aiCore: 30
+      templates:
+        - name: vir02
+          memory: 2184
+          aiCore: 2
+        - name: vir04
+          memory: 4369
+          aiCore: 4
+        - name: vir08
+          memory: 8738
+          aiCore: 8
+        - name: vir16
+          memory: 17476
+          aiCore: 16
+    - chipName: 910B2
+      commonWord: Ascend910B2
+      resourceName: huawei.com/Ascend910B2
+      resourceMemoryName: huawei.com/Ascend910B2-memory
+      memoryAllocatable: 65536
+      memoryCapacity: 65536
+      memoryFactor: 1
+      aiCore: 24
+      aiCPU: 6
+      templates:
+        - name: vir03_1c_8g
+          memory: 8192
+          aiCore: 3
+          aiCPU: 1
+        - name: vir06_1c_16g
+          memory: 16384
+          aiCore: 6
+          aiCPU: 1
+        - name: vir12_3c_32g
+          memory: 32768
+          aiCore: 12
+          aiCPU: 3
+    - chipName: 910B3
+      commonWord: Ascend910B3
+      resourceName: huawei.com/Ascend910B3
+      resourceMemoryName: huawei.com/Ascend910B3-memory
+      memoryAllocatable: 65536
+      memoryCapacity: 65536
+      memoryFactor: 1
+      aiCore: 20
+      aiCPU: 7
+      templates:
+        - name: vir05_1c_16g
+          memory: 16384
+          aiCore: 5
+          aiCPU: 1
+        - name: vir10_3c_32g
+          memory: 32768
+          aiCore: 10
+          aiCPU: 3
+    - chipName: 910B4-1
+      commonWord: Ascend910B4-1
+      resourceName: huawei.com/Ascend910B4-1
+      resourceMemoryName: huawei.com/Ascend910B4-1-memory
+      memoryAllocatable: 65536
+      memoryCapacity: 65536
+      memoryFactor: 1
+      aiCore: 20
+      aiCPU: 7
+      templates:
+        # NOTE: Names of vnpu template for 910B4-1 are fixed by Ascend runtime and must not be changed.
+        # The memory is used for scheduling so the correct values must be set.
+        # Template vir05_1c_8g actually provides 16GB memory,
+        - name: vir05_1c_8g
+          memory: 16384
+          aiCore: 5
+          aiCPU: 1
+        # Template vir10_3c_16g actually provides 32GB memory
+        - name: vir10_3c_16g
+          memory: 32768
+          aiCore: 10
+          aiCPU: 3
+    - chipName: 910B4
+      commonWord: Ascend910B4
+      resourceName: huawei.com/Ascend910B4
+      resourceMemoryName: huawei.com/Ascend910B4-memory
+      memoryAllocatable: 32768
+      memoryCapacity: 32768
+      memoryFactor: 1
+      aiCore: 20
+      aiCPU: 7
+      templates:
+        - name: vir05_1c_8g
+          memory: 8192
+          aiCore: 5
+          aiCPU: 1
+        - name: vir10_3c_16g
+          memory: 16384
+          aiCore: 10
+          aiCPU: 3
+    - chipName: 310P3
+      commonWord: Ascend310P
+      resourceName: huawei.com/Ascend310P
+      resourceMemoryName: huawei.com/Ascend310P-memory
+      memoryAllocatable: 21527
+      memoryCapacity: 24576
+      memoryFactor: 1
+      aiCore: 8
+      aiCPU: 7
+      templates:
+        - name: vir01
+          memory: 3072
+          aiCore: 1
+          aiCPU: 1
+        - name: vir02
+          memory: 6144
+          aiCore: 2
+          aiCPU: 2
+        - name: vir04
+          memory: 12288
+          aiCore: 4
+          aiCPU: 4
+  {{ end }}
diff --git a/packages/system/hami/charts/hami/templates/scheduler/job-patch/clusterrole.yaml b/packages/system/hami/charts/hami/templates/scheduler/job-patch/clusterrole.yaml
new file mode 100644
index 00000000..77e891cc
--- /dev/null
+++ b/packages/system/hami/charts/hami/templates/scheduler/job-patch/clusterrole.yaml
@@ -0,0 +1,21 @@
+{{- if and .Values.scheduler.admissionWebhook.enabled (.Values.scheduler.patch.enabled) (not .Values.scheduler.certManager.enabled) -}}
+apiVersion: rbac.authorization.k8s.io/v1
+kind: ClusterRole
+metadata:
+  name: {{ include "hami-vgpu.fullname" . }}-admission
+  annotations:
+    "helm.sh/hook": pre-install,pre-upgrade,post-install,post-upgrade
+    "helm.sh/hook-delete-policy": before-hook-creation,hook-succeeded
+  labels:
+    {{- include "hami-vgpu.labels" . | nindent 4 }}
+    app.kubernetes.io/component: admission-webhook
+rules:
+  - apiGroups:
+      - admissionregistration.k8s.io
+    resources:
+      #- validatingwebhookconfigurations
+      - mutatingwebhookconfigurations
+    verbs:
+      - get
+      - update
+{{- end }}
diff --git a/packages/system/hami/charts/hami/templates/scheduler/job-patch/clusterrolebinding.yaml b/packages/system/hami/charts/hami/templates/scheduler/job-patch/clusterrolebinding.yaml
new file mode 100644
index 00000000..2b82f926
--- /dev/null
+++ b/packages/system/hami/charts/hami/templates/scheduler/job-patch/clusterrolebinding.yaml
@@ -0,0 +1,20 @@
+{{- if and .Values.scheduler.admissionWebhook.enabled (.Values.scheduler.patch.enabled) (not .Values.scheduler.certManager.enabled) -}}
+apiVersion: rbac.authorization.k8s.io/v1
+kind: ClusterRoleBinding
+metadata:
+  name:  {{ include "hami-vgpu.fullname" . }}-admission
+  annotations:
+    "helm.sh/hook": pre-install,pre-upgrade,post-install,post-upgrade
+    "helm.sh/hook-delete-policy": before-hook-creation,hook-succeeded
+  labels:
+    {{- include "hami-vgpu.labels" . | nindent 4 }}
+    app.kubernetes.io/component: admission-webhook
+roleRef:
+  apiGroup: rbac.authorization.k8s.io
+  kind: ClusterRole
+  name: {{ include "hami-vgpu.fullname" . }}-admission
+subjects:
+  - kind: ServiceAccount
+    name: {{ include "hami-vgpu.fullname" . }}-admission
+    namespace: {{ include "hami-vgpu.namespace" . }}
+{{- end }}
diff --git a/packages/system/hami/charts/hami/templates/scheduler/job-patch/job-createSecret.yaml b/packages/system/hami/charts/hami/templates/scheduler/job-patch/job-createSecret.yaml
new file mode 100644
index 00000000..0e36d95f
--- /dev/null
+++ b/packages/system/hami/charts/hami/templates/scheduler/job-patch/job-createSecret.yaml
@@ -0,0 +1,68 @@
+{{- if and .Values.scheduler.admissionWebhook.enabled (.Values.scheduler.patch.enabled) (not .Values.scheduler.certManager.enabled) -}}
+apiVersion: batch/v1
+kind: Job
+metadata:
+  name: {{ include "hami-vgpu.fullname" . }}-admission-create
+  namespace: {{ include "hami-vgpu.namespace" . }}
+  annotations:
+    "helm.sh/hook": pre-install,pre-upgrade
+    "helm.sh/hook-delete-policy": before-hook-creation,hook-succeeded
+  labels:
+    {{- include "hami-vgpu.labels" . | nindent 4 }}
+    app.kubernetes.io/component: admission-webhook
+spec:
+  {{- if .Capabilities.APIVersions.Has "batch/v1alpha1" }}
+  # Alpha feature since k8s 1.12
+  ttlSecondsAfterFinished: 0
+  {{- end }}
+  template:
+    metadata:
+      name: {{ include "hami-vgpu.fullname" . }}-admission-create
+      {{- if .Values.scheduler.patch.podAnnotations }}
+      annotations: {{ toYaml .Values.scheduler.patch.podAnnotations | nindent 8 }}
+      {{- end }}
+      labels:
+        {{- include "hami-vgpu.labels" . | nindent 8 }}
+        app.kubernetes.io/component: admission-webhook
+        hami.io/webhook: ignore
+    spec:
+      {{- if .Values.scheduler.patch.priorityClassName }}
+      priorityClassName: {{ .Values.scheduler.patch.priorityClassName }}
+      {{- end }}
+      {{- if ge (regexReplaceAll "[^0-9]" .Capabilities.KubeVersion.Minor "" | int) 22 }}
+        {{- include "hami.scheduler.patch.new.imagePullSecrets" . | nindent 6 }}
+      {{- else }}
+        {{- include "hami.scheduler.patch.imagePullSecrets" . | nindent 6 }}
+      {{- end }}
+      containers:
+        - name: create
+          {{- if ge (regexReplaceAll "[^0-9]" .Capabilities.KubeVersion.Minor "" | int) 22 }}
+          image: {{ include "hami.scheduler.patch.new.image" . }}
+          imagePullPolicy: {{ .Values.scheduler.patch.imageNew.pullPolicy }}
+          {{- else }}
+          image: {{ include "hami.scheduler.patch.image" . }}
+          imagePullPolicy: {{ .Values.scheduler.patch.image.pullPolicy }}
+          {{- end }}
+          args:
+            - create
+            - --cert-name=tls.crt
+            - --key-name=tls.key
+            {{- if .Values.scheduler.admissionWebhook.customURL.enabled }}
+            - --host={{ printf "%s.%s.svc,127.0.0.1,%s" (include "hami-vgpu.scheduler" .) (include "hami-vgpu.namespace" .) .Values.scheduler.admissionWebhook.customURL.host}}
+            {{- else }}
+            - --host={{ printf "%s.%s.svc,127.0.0.1" (include "hami-vgpu.scheduler" .) (include "hami-vgpu.namespace" .) }}
+            {{- end }}
+            - --namespace={{ include "hami-vgpu.namespace" . }}
+            - --secret-name={{ include "hami-vgpu.scheduler.tls" . }}
+      restartPolicy: OnFailure
+      serviceAccountName: {{ include "hami-vgpu.fullname" . }}-admission
+      {{- if .Values.scheduler.patch.nodeSelector }}
+      nodeSelector: {{ toYaml .Values.scheduler.patch.nodeSelector | nindent 8 }}
+      {{- end }}
+      {{- if .Values.scheduler.patch.tolerations }}
+      tolerations: {{ toYaml .Values.scheduler.patch.tolerations | nindent 8 }}
+      {{- end }}
+      securityContext:
+        runAsNonRoot: true
+        runAsUser: {{ .Values.scheduler.patch.runAsUser }}
+{{- end }}
diff --git a/packages/system/hami/charts/hami/templates/scheduler/job-patch/job-patchWebhook.yaml b/packages/system/hami/charts/hami/templates/scheduler/job-patch/job-patchWebhook.yaml
new file mode 100644
index 00000000..ce52042c
--- /dev/null
+++ b/packages/system/hami/charts/hami/templates/scheduler/job-patch/job-patchWebhook.yaml
@@ -0,0 +1,63 @@
+{{- if and .Values.scheduler.admissionWebhook.enabled (.Values.scheduler.patch.enabled) (not .Values.scheduler.certManager.enabled) -}}
+apiVersion: batch/v1
+kind: Job
+metadata:
+  name: {{ include "hami-vgpu.fullname" . }}-admission-patch
+  namespace: {{ include "hami-vgpu.namespace" . }}
+  annotations:
+    "helm.sh/hook": post-install,post-upgrade
+    "helm.sh/hook-delete-policy": before-hook-creation,hook-succeeded
+  labels:
+    {{- include "hami-vgpu.labels" . | nindent 4 }}
+    app.kubernetes.io/component: admission-webhook
+spec:
+  {{- if .Capabilities.APIVersions.Has "batch/v1alpha1" }}
+  # Alpha feature since k8s 1.12
+  ttlSecondsAfterFinished: 0
+  {{- end }}
+  template:
+    metadata:
+      name: {{ include "hami-vgpu.fullname" . }}-admission-patch
+      {{- if .Values.scheduler.patch.podAnnotations }}
+      annotations: {{ toYaml .Values.scheduler.patch.podAnnotations | nindent 8 }}
+      {{- end }}
+      labels:
+        {{- include "hami-vgpu.labels" . | nindent 8 }}
+        app.kubernetes.io/component: admission-webhook
+        hami.io/webhook: ignore
+    spec:
+      {{- if .Values.scheduler.patch.priorityClassName }}
+      priorityClassName: {{ .Values.scheduler.patch.priorityClassName }}
+      {{- end }}
+      {{- if ge (regexReplaceAll "[^0-9]" .Capabilities.KubeVersion.Minor "" | int) 22 }}
+        {{- include "hami.scheduler.patch.new.imagePullSecrets" . | nindent 6 }}
+      {{- else }}
+        {{- include "hami.scheduler.patch.imagePullSecrets" . | nindent 6 }}
+      {{- end }}
+      containers:
+        - name: patch
+          {{- if ge (regexReplaceAll "[^0-9]" .Capabilities.KubeVersion.Minor "" | int) 22 }}
+          image: {{ include "hami.scheduler.patch.new.image" . }}
+          imagePullPolicy: {{ .Values.scheduler.patch.imageNew.pullPolicy }}
+          {{- else }}
+          image: {{ include "hami.scheduler.patch.image" . }}
+          imagePullPolicy: {{ .Values.scheduler.patch.image.pullPolicy }}
+          {{- end }}
+          args:
+            - patch
+            - --webhook-name={{ include "hami-vgpu.scheduler.webhook" . }}
+            - --namespace={{ include "hami-vgpu.namespace" . }}
+            - --patch-validating=false
+            - --secret-name={{ include "hami-vgpu.scheduler.tls" . }}
+      restartPolicy: OnFailure
+      serviceAccountName: {{ include "hami-vgpu.fullname" . }}-admission
+      {{- if .Values.scheduler.patch.nodeSelector }}
+      nodeSelector: {{ toYaml .Values.scheduler.patch.nodeSelector | nindent 8 }}
+      {{- end }}
+      {{- if .Values.scheduler.patch.tolerations }}
+      tolerations: {{ toYaml .Values.scheduler.patch.tolerations | nindent 8 }}
+      {{- end }}
+      securityContext:
+        runAsNonRoot: true
+        runAsUser: {{ .Values.scheduler.patch.runAsUser }}
+{{- end }}
diff --git a/packages/system/hami/charts/hami/templates/scheduler/job-patch/role.yaml b/packages/system/hami/charts/hami/templates/scheduler/job-patch/role.yaml
new file mode 100644
index 00000000..56682f8e
--- /dev/null
+++ b/packages/system/hami/charts/hami/templates/scheduler/job-patch/role.yaml
@@ -0,0 +1,21 @@
+{{- if and .Values.scheduler.admissionWebhook.enabled (.Values.scheduler.patch.enabled) (not .Values.scheduler.certManager.enabled) -}}
+apiVersion: rbac.authorization.k8s.io/v1
+kind: Role
+metadata:
+  name:  {{ include "hami-vgpu.fullname" . }}-admission
+  namespace: {{ include "hami-vgpu.namespace" . }}
+  annotations:
+    "helm.sh/hook": pre-install,pre-upgrade,post-install,post-upgrade
+    "helm.sh/hook-delete-policy": before-hook-creation,hook-succeeded
+  labels:
+    {{- include "hami-vgpu.labels" . | nindent 4 }}
+    app.kubernetes.io/component: admission-webhook
+rules:
+  - apiGroups:
+      - ""
+    resources:
+      - secrets
+    verbs:
+      - get
+      - create
+{{- end }}
diff --git a/packages/system/hami/charts/hami/templates/scheduler/job-patch/rolebinding.yaml b/packages/system/hami/charts/hami/templates/scheduler/job-patch/rolebinding.yaml
new file mode 100644
index 00000000..7239b128
--- /dev/null
+++ b/packages/system/hami/charts/hami/templates/scheduler/job-patch/rolebinding.yaml
@@ -0,0 +1,21 @@
+{{- if and .Values.scheduler.admissionWebhook.enabled (.Values.scheduler.patch.enabled) (not .Values.scheduler.certManager.enabled) -}}
+apiVersion: rbac.authorization.k8s.io/v1
+kind: RoleBinding
+metadata:
+  name: {{ include "hami-vgpu.fullname" . }}-admission
+  namespace: {{ include "hami-vgpu.namespace" . }}
+  annotations:
+    "helm.sh/hook": pre-install,pre-upgrade,post-install,post-upgrade
+    "helm.sh/hook-delete-policy": before-hook-creation,hook-succeeded
+  labels:
+    {{- include "hami-vgpu.labels" . | nindent 4 }}
+    app.kubernetes.io/component: admission-webhook
+roleRef:
+  apiGroup: rbac.authorization.k8s.io
+  kind: Role
+  name: {{ include "hami-vgpu.fullname" . }}-admission
+subjects:
+  - kind: ServiceAccount
+    name: {{ include "hami-vgpu.fullname" . }}-admission
+    namespace: {{ include "hami-vgpu.namespace" . }}
+{{- end }}
diff --git a/packages/system/hami/charts/hami/templates/scheduler/job-patch/serviceaccount.yaml b/packages/system/hami/charts/hami/templates/scheduler/job-patch/serviceaccount.yaml
new file mode 100644
index 00000000..857c6a8d
--- /dev/null
+++ b/packages/system/hami/charts/hami/templates/scheduler/job-patch/serviceaccount.yaml
@@ -0,0 +1,13 @@
+{{- if and .Values.scheduler.admissionWebhook.enabled (.Values.scheduler.patch.enabled) (not .Values.scheduler.certManager.enabled) }}
+apiVersion: v1
+kind: ServiceAccount
+metadata:
+  name: {{ include "hami-vgpu.fullname" . }}-admission
+  namespace: {{ include "hami-vgpu.namespace" . }}
+  annotations:
+    "helm.sh/hook": pre-install,pre-upgrade,post-install,post-upgrade
+    "helm.sh/hook-delete-policy": before-hook-creation,hook-succeeded
+  labels:
+    {{- include "hami-vgpu.labels" . | nindent 4 }}
+    app.kubernetes.io/component: admission-webhook
+{{- end }}
diff --git a/packages/system/hami/charts/hami/templates/scheduler/role.yaml b/packages/system/hami/charts/hami/templates/scheduler/role.yaml
new file mode 100644
index 00000000..5f7d7e44
--- /dev/null
+++ b/packages/system/hami/charts/hami/templates/scheduler/role.yaml
@@ -0,0 +1,12 @@
+apiVersion: rbac.authorization.k8s.io/v1
+kind: Role
+metadata:
+  name: {{ include "hami-vgpu.scheduler" . }}
+  namespace: {{ include "hami-vgpu.namespace" . }}
+  labels:
+    app.kubernetes.io/component: "hami-scheduler"
+    {{- include "hami-vgpu.labels" . | nindent 4 }}
+rules:
+  - apiGroups: ["coordination.k8s.io"]
+    resources: ["leases"]
+    verbs: ["create", "list", "watch", "get", "update", "patch"]
diff --git a/packages/system/hami/charts/hami/templates/scheduler/rolebinding.yaml b/packages/system/hami/charts/hami/templates/scheduler/rolebinding.yaml
new file mode 100644
index 00000000..96e175a1
--- /dev/null
+++ b/packages/system/hami/charts/hami/templates/scheduler/rolebinding.yaml
@@ -0,0 +1,31 @@
+apiVersion: rbac.authorization.k8s.io/v1
+kind: RoleBinding
+metadata:
+  name: {{ include "hami-vgpu.scheduler" . }}
+  namespace: {{ include "hami-vgpu.namespace" . }}
+  labels:
+    app.kubernetes.io/component: "hami-scheduler"
+    {{- include "hami-vgpu.labels" . | nindent 4 }}
+roleRef:
+  apiGroup: rbac.authorization.k8s.io
+  kind: Role
+  name: {{ include "hami-vgpu.scheduler" . }}
+subjects:
+  - kind: ServiceAccount
+    name: {{ include "hami-vgpu.scheduler" . }}
+    namespace: {{ include "hami-vgpu.namespace" . }}
+---
+{{- if .Values.mockDevicePlugin.enabled }}
+apiVersion: rbac.authorization.k8s.io/v1
+kind: ClusterRoleBinding
+metadata:
+  name: {{ include "hami-vgpu.mock-device-plugin" . }}
+roleRef:
+  apiGroup: rbac.authorization.k8s.io
+  kind: ClusterRole
+  name: {{ include "hami-vgpu.mock-device-plugin" . }}
+subjects:
+  - kind: ServiceAccount
+    name: {{ include "hami-vgpu.mock-device-plugin" . }}
+    namespace: {{ include "hami-vgpu.namespace" . }}
+{{- end -}}
diff --git a/packages/system/hami/charts/hami/templates/scheduler/service.yaml b/packages/system/hami/charts/hami/templates/scheduler/service.yaml
new file mode 100644
index 00000000..d7538fed
--- /dev/null
+++ b/packages/system/hami/charts/hami/templates/scheduler/service.yaml
@@ -0,0 +1,34 @@
+apiVersion: v1
+kind: Service
+metadata:
+  name: {{ include "hami-vgpu.scheduler" . }}
+  namespace: {{ include "hami-vgpu.namespace" . }}
+  labels:
+    app.kubernetes.io/component: hami-scheduler
+    {{- include "hami-vgpu.labels" . | nindent 4 }}
+    {{- if .Values.scheduler.service.labels }}
+    {{ toYaml .Values.scheduler.service.labels | indent 4 }}
+    {{- end }}
+  {{- if .Values.scheduler.service.annotations }}
+  annotations: {{ toYaml .Values.scheduler.service.annotations | nindent 4 }}
+  {{- end }}
+spec:
+  type: {{ .Values.scheduler.service.type | default "NodePort" }}  # Default type is NodePort
+  ports:
+    - name: http
+      port: {{ .Values.scheduler.service.httpPort | default 443 }}  # Default HTTP port is 443
+      targetPort: {{ .Values.scheduler.service.httpTargetPort | default 443 }}
+      {{- if eq (.Values.scheduler.service.type | default "NodePort") "NodePort" }}  # If type is NodePort, set nodePort
+      nodePort: {{ .Values.scheduler.service.schedulerPort | default 31998 }}
+      {{- end }}
+      protocol: TCP
+    - name: monitor
+      port: {{ .Values.scheduler.service.monitorPort | default 31993 }}  # Default monitoring port is 31993
+      targetPort: {{ .Values.scheduler.service.monitorTargetPort | default 9395 }}
+      {{- if eq (.Values.scheduler.service.type | default "NodePort") "NodePort" }}  # If type is NodePort, set nodePort
+      nodePort: {{ .Values.scheduler.service.monitorPort | default 31993 }}
+      {{- end }}
+      protocol: TCP
+  selector:
+    app.kubernetes.io/component: hami-scheduler
+    {{- include "hami-vgpu.selectorLabels" . | nindent 4 }}
diff --git a/packages/system/hami/charts/hami/templates/scheduler/serviceaccount.yaml b/packages/system/hami/charts/hami/templates/scheduler/serviceaccount.yaml
new file mode 100644
index 00000000..0435c003
--- /dev/null
+++ b/packages/system/hami/charts/hami/templates/scheduler/serviceaccount.yaml
@@ -0,0 +1,16 @@
+apiVersion: v1
+kind: ServiceAccount
+metadata:
+  name: {{ include "hami-vgpu.scheduler" . }}
+  namespace: {{ include "hami-vgpu.namespace" . }}
+  labels:
+    app.kubernetes.io/component: "hami-scheduler"
+    {{- include "hami-vgpu.labels" . | nindent 4 }}
+---
+{{- if .Values.mockDevicePlugin.enabled }}
+apiVersion: v1
+kind: ServiceAccount
+metadata:
+  name: {{ include "hami-vgpu.mock-device-plugin" . }}
+  namespace: {{ include "hami-vgpu.namespace" . }}
+{{- end -}}
diff --git a/packages/system/hami/charts/hami/templates/scheduler/webhook.yaml b/packages/system/hami/charts/hami/templates/scheduler/webhook.yaml
new file mode 100644
index 00000000..db9f8029
--- /dev/null
+++ b/packages/system/hami/charts/hami/templates/scheduler/webhook.yaml
@@ -0,0 +1,57 @@
+{{- if .Values.scheduler.admissionWebhook.enabled -}}
+apiVersion: admissionregistration.k8s.io/v1
+kind: MutatingWebhookConfiguration
+metadata:
+  {{- if .Values.scheduler.certManager.enabled }}
+  annotations:
+    cert-manager.io/inject-ca-from: {{ include "hami-vgpu.namespace" . }}/{{ include "hami-vgpu.scheduler" . }}-serving-cert
+  {{- end }}
+  name: {{ include "hami-vgpu.scheduler.webhook" . }}
+webhooks:
+  - admissionReviewVersions:
+    - v1beta1
+    clientConfig:
+      {{- if .Values.scheduler.admissionWebhook.customURL.enabled }}
+      url: https://{{ .Values.scheduler.admissionWebhook.customURL.host}}:{{.Values.scheduler.admissionWebhook.customURL.port}}{{.Values.scheduler.admissionWebhook.customURL.path}}
+      {{- else }}
+      service:
+        name: {{ include "hami-vgpu.scheduler" . }}
+        namespace: {{ include "hami-vgpu.namespace" . }}
+        path: /webhook
+        port: {{ .Values.scheduler.service.httpPort }}
+      {{- end }}
+    failurePolicy: {{ .Values.scheduler.admissionWebhook.failurePolicy }}
+    matchPolicy: Equivalent
+    name: vgpu.hami.io
+    namespaceSelector:
+      matchExpressions:
+      - key: hami.io/webhook
+        operator: NotIn
+        values:
+        - ignore
+      {{- if .Values.scheduler.admissionWebhook.whitelistNamespaces }}
+      - key: kubernetes.io/metadata.name
+        operator: NotIn
+        values:
+        {{- toYaml .Values.scheduler.admissionWebhook.whitelistNamespaces | nindent 10 }}
+      {{- end }}
+    objectSelector:
+      matchExpressions:
+      - key: hami.io/webhook
+        operator: NotIn
+        values:
+        - ignore
+    reinvocationPolicy: {{ .Values.scheduler.admissionWebhook.reinvocationPolicy }}
+    rules:
+      - apiGroups:
+          - ""
+        apiVersions:
+          - v1
+        operations:
+          - CREATE
+        resources:
+          - pods
+        scope: '*'
+    sideEffects: None
+    timeoutSeconds: 10
+{{- end }}
diff --git a/packages/system/hami/charts/hami/values.yaml b/packages/system/hami/charts/hami/values.yaml
new file mode 100644
index 00000000..01caca0b
--- /dev/null
+++ b/packages/system/hami/charts/hami/values.yaml
@@ -0,0 +1,455 @@
+## @section Global configuration
+## @param global.imageRegistry Global Docker image registry
+## @param global.imagePullSecrets Global Docker image pull secrets
+global:
+  ## @param global.imageRegistry Global Docker image registry
+  imageRegistry: ""
+  ## E.g.
+  ## imagePullSecrets:
+  ##   - myRegistryKeySecretName
+  ## @param global.imagePullSecrets Global Docker image pull secrets
+  imagePullSecrets: []
+  imageTag: "v2.8.1"
+  gpuHookPath: /usr/local
+  labels: {}
+  annotations: {}
+  managedNodeSelectorEnable: false
+  managedNodeSelector:
+    usage: "gpu"
+
+nameOverride: ""
+fullnameOverride: ""
+namespaceOverride: ""
+
+#Nvidia GPU Parameters
+resourceName: "nvidia.com/gpu"
+resourceMem: "nvidia.com/gpumem"
+resourceMemPercentage: "nvidia.com/gpumem-percentage"
+resourceCores: "nvidia.com/gpucores"
+resourcePriority: "nvidia.com/priority"
+
+#MLU Parameters
+mluResourceName: "cambricon.com/vmlu"
+mluResourceMem: "cambricon.com/mlu.smlu.vmemory"
+mluResourceCores: "cambricon.com/mlu.smlu.vcore"
+
+#Hygon DCU Parameters
+dcuResourceName: "hygon.com/dcunum"
+dcuResourceMem: "hygon.com/dcumem"
+dcuResourceCores: "hygon.com/dcucores"
+
+#Metax sGPU Parameters
+metaxResourceName: "metax-tech.com/sgpu"
+metaxResourceCore: "metax-tech.com/vcore"
+metaxResourceMem: "metax-tech.com/vmemory"
+metaxsGPUTopologyAware: "false"
+
+#Enflame VGCU Parameters
+enflameResourceNameVGCU: "enflame.com/vgcu"
+enflameResourceNameVGCUPercentage: "enflame.com/vgcu-percentage"
+
+#Kunlun XPU Parameters
+kunlunResourceName: "kunlunxin.com/xpu"
+kunlunResourceVCountName: "kunlunxin.com/vxpu"
+kunlunResourceVMemoryName: "kunlunxin.com/vxpu-memory"
+
+schedulerName: "hami-scheduler"
+
+scheduler:
+  # @param nodeName defines the node name and the nvidia-vgpu-scheduler-scheduler will schedule to the node.
+  # if we install the nvidia-vgpu-scheduler-scheduler as default scheduler, we need to remove the k8s default
+  # scheduler pod from the cluster first, we must specify node name to skip the schedule workflow.
+  nodeName: ""
+  #nodeLabelSelector:
+  #  "gpu": "on"
+  overwriteEnv: "false"
+  defaultSchedulerPolicy:
+    nodeSchedulerPolicy: binpack
+    gpuSchedulerPolicy: spread
+  metricsBindAddress: ":9395"
+  # If set to false, When Pod.Spec.SchedulerName equals to the const DefaultSchedulerName in k8s.io/api/core/v1 package, webhook will not overwrite it
+  forceOverwriteDefaultScheduler: true
+  livenessProbe: false
+  leaderElect: true
+  # when leaderElect is true, replicas is available, otherwise replicas is 1.
+  replicas: 1
+  kubeScheduler:
+    # @param enabled indicate whether to run kube-scheduler container in the scheduler pod, it's true by default.
+    enabled: true
+    ## @param image.registry kube scheduler image registry
+    ## @param image.repository kube scheduler image repository
+    ## @param image.tag kube scheduler image tag (immutable tags are recommended)
+    ## @param image.pullPolicy kube scheduler image pull policy
+    ## @param image.pullSecrets Specify docker-registry secret names as an array
+    image:
+      registry: "registry.cn-hangzhou.aliyuncs.com"
+      repository: "google_containers/kube-scheduler"
+      tag: ""
+      ## Specify a imagePullPolicy
+      ## Defaults to 'Always' if image tag is 'latest', else set to 'IfNotPresent'
+      ## ref: https://kubernetes.io/docs/user-guide/images/#pre-pulling-images
+      ##
+      pullPolicy: IfNotPresent
+      ## Optionally specify an array of imagePullSecrets.
+      ## Secrets must be manually created in the namespace.
+      ## Example:
+      ## pullSecrets:
+      ##   - myRegistryKeySecretName
+      ##
+      pullSecrets: []
+    resources: {}
+      # If you do want to specify resources, uncomment the following lines, adjust them as necessary.
+      # and remove the curly braces after 'resources:'.
+#      limits:
+#        cpu: 1000m
+#        memory: 1000Mi
+#      requests:
+#        cpu: 100m
+#        memory: 100Mi
+    extraNewArgs:
+      - --config=/config/config.yaml
+      - -v=4
+    extraArgs:
+      - --policy-config-file=/config/config.json
+      - -v=4
+  extender:
+    ## @param image.registry scheduler extender image registry
+    ## @param image.repository scheduler extender image repository
+    ## @param image.tag scheduler extender image tag (immutable tags are recommended)
+    ## @param image.pullPolicy scheduler extender image pull policy
+    ## @param image.pullSecrets Specify docker-registry secret names as an array
+    image:
+      registry: "docker.io"
+      repository: "projecthami/hami"
+      tag: ""
+      ## Specify a imagePullPolicy
+      ## Defaults to 'Always' if image tag is 'latest', else set to 'IfNotPresent'
+      ## ref: https://kubernetes.io/docs/user-guide/images/#pre-pulling-images
+      ##
+      pullPolicy: IfNotPresent
+      ## Optionally specify an array of imagePullSecrets.
+      ## Secrets must be manually created in the namespace.
+      ## Example:
+      ## pullSecrets:
+      ##   - myRegistryKeySecretName
+      ##
+      pullSecrets: []
+    resources: {}
+      # If you do want to specify resources, uncomment the following lines, adjust them as necessary,
+      # and remove the curly braces after 'resources:'.
+#      limits:
+#        cpu: 1000m
+#        memory: 1000Mi
+#      requests:
+#        cpu: 100m
+#        memory: 100Mi
+    extraArgs:
+      - --debug
+      - -v=4
+  nodeLockExpire: "5m"
+  podAnnotations: {}
+  tolerations: []
+  #serviceAccountName: "hami-vgpu-scheduler-sa"
+  admissionWebhook:
+    # If set to false, the admission webhook is not installed and any pods that should use HAMi must be
+    # configured to use the right 'schedulerName' and other device-specific configurations.
+    enabled: true
+    customURL:
+      enabled: false
+      # must be an endpoint using https.
+      # should generate host certs here
+      host: 127.0.0.1 # hostname or ip, can be your node'IP if you want to use https://<nodeIP>:<schedulerPort>/<path>
+      port: 31998
+      path: /webhook
+    whitelistNamespaces:
+    # Specify the namespaces that the webhook will not be applied to.
+      # - default
+      # - kube-system
+      # - istio-system
+    reinvocationPolicy: Never
+    failurePolicy: Ignore
+  ## TLS Certificate Option 1: Use cert-manager to generate self-signed certificate.
+  ## If enabled, always takes precedence over options 2.
+  certManager:
+    enabled: false
+  ## TLS Certificate Option 2: Use kube-webhook-certgen to generate self-signed certificate.
+  ## If true and certManager.enabled is false, Helm will automatically create a self-signed cert and secret for you.
+  patch:
+    enabled: true
+    ## @param image.registry scheduler extender image registry
+    ## @param image.repository scheduler extender image repository
+    ## @param image.tag scheduler extender image tag (immutable tags are recommended)
+    ## @param image.pullPolicy scheduler extender image pull policy
+    ## @param image.pullSecrets Specify docker-registry secret names as an array
+    image:
+      registry: "docker.io"
+      repository: "jettech/kube-webhook-certgen"
+      tag: "v1.5.2"
+      ## Specify a imagePullPolicy
+      ## Defaults to 'Always' if image tag is 'latest', else set to 'IfNotPresent'
+      ## ref: https://kubernetes.io/docs/user-guide/images/#pre-pulling-images
+      ##
+      pullPolicy: IfNotPresent
+      ## Optionally specify an array of imagePullSecrets.
+      ## Secrets must be manually created in the namespace.
+      ## Example:
+      ## pullSecrets:
+      ##   - myRegistryKeySecretName
+      ##
+      pullSecrets: []
+    ## @param image.registry scheduler extender image registry
+    ## @param image.repository scheduler extender image repository
+    ## @param image.tag scheduler extender image tag (immutable tags are recommended)
+    ## @param image.pullPolicy scheduler extender image pull policy
+    ## @param image.pullSecrets Specify docker-registry secret names as an array
+    imageNew: 
+      registry: "docker.io"
+      repository: "liangjw/kube-webhook-certgen"
+      tag: "v1.1.1"
+      ## Specify a imagePullPolicy
+      ## Defaults to 'Always' if image tag is 'latest', else set to 'IfNotPresent'
+      ## ref: https://kubernetes.io/docs/user-guide/images/#pre-pulling-images
+      ##
+      pullPolicy: IfNotPresent
+      ## Optionally specify an array of imagePullSecrets.
+      ## Secrets must be manually created in the namespace.
+      ## Example:
+      ## pullSecrets:
+      ##   - myRegistryKeySecretName
+      ##
+      pullSecrets: []
+    priorityClassName: ""
+    podAnnotations: {}
+    nodeSelector: {}
+    tolerations: []
+    runAsUser: 2000
+  service:
+    type: NodePort  # Default type is NodePort, can be changed to ClusterIP
+    httpPort: 443   # HTTP port
+    schedulerPort: 31998  # NodePort for HTTP
+    monitorPort: 31993    # Monitoring port
+    monitorTargetPort: 9395
+    httpTargetPort: 443
+    labels: {}
+    annotations: {}
+
+devicePlugin:
+  enabled: true
+  gpuOperatorToolkitReady:
+    enabled: false
+    hostPath: "/run/nvidia/validations"
+  ## @param image.registry devicePlugin image registry
+  ## @param image.repository devicePlugin image repository
+  ## @param image.tag devicePlugin image tag (immutable tags are recommended)
+  ## @param image.pullPolicy devicePlugin image pull policy
+  ## @param image.pullSecrets Specify docker-registry secret names as an array
+  image:
+    registry: "docker.io"
+    repository: "projecthami/hami"
+    tag: ""
+    ## Specify a imagePullPolicy
+    ## Defaults to 'Always' if image tag is 'latest', else set to 'IfNotPresent'
+    ## ref: https://kubernetes.io/docs/user-guide/images/#pre-pulling-images
+    ##
+    pullPolicy: IfNotPresent
+    ## Optionally specify an array of imagePullSecrets.
+    ## Secrets must be manually created in the namespace.
+    ## Example:
+    ## pullSecrets:
+    ##   - myRegistryKeySecretName
+    ##
+    pullSecrets: []
+  monitor:
+    ## @param image.registry monitor image registry
+    ## @param image.repository monitor image repository
+    ## @param image.tag monitor image tag (immutable tags are recommended)
+    ## @param image.pullPolicy monitor image pull policy
+    ## @param image.pullSecrets Specify docker-registry secret names as an array
+    image:
+      registry: "docker.io"
+      repository: "projecthami/hami"
+      tag: ""
+      ## Specify a imagePullPolicy
+      ## Defaults to 'Always' if image tag is 'latest', else set to 'IfNotPresent'
+      ## ref: https://kubernetes.io/docs/user-guide/images/#pre-pulling-images
+      ##
+      pullPolicy: IfNotPresent
+      ## Optionally specify an array of imagePullSecrets.
+      ## Secrets must be manually created in the namespace.
+      ## Example:
+      ## pullSecrets:
+      ##   - myRegistryKeySecretName
+      ##
+      pullSecrets: []
+    ctrPath: /usr/local/vgpu/containers
+    resyncInterval: "5m"
+    extraArgs:
+      - -v=4
+    extraEnvs: {}
+    resources: {}
+      # If you do want to specify resources, uncomment the following lines, adjust them as necessary.
+      # and remove the curly braces after 'resources:'.
+#      limits:
+#        cpu: 1000m
+#        memory: 1000Mi
+#      requests:
+#        cpu: 100m
+#        memory: 100Mi
+
+  deviceSplitCount: 10
+  deviceMemoryScaling: 1
+  deviceCoreScaling: 1
+  # Node configuration for device plugin, Priority: externalConfigName > config > default config
+  nodeConfiguration:
+    # If you want to use a custom config.json, you can set the content here.
+    # If this is set, it will override the default config.json(An example is as follows).
+    config: |
+      {
+        "nodeconfig": [
+          {
+            "name": "your-node-name",
+            "operatingmode": "hami-core",
+            "devicememoryscaling": 1,
+            "devicesplitcount": 10,
+            "migstrategy": "none",
+            "filterdevices": {
+              "uuid": [],
+              "index": []
+            }
+          }
+        ]
+      }
+    # If you want to use an existing ConfigMap, you can set the name here.
+    # If this is set, the chart will not create the ConfigMap and will use the existing one.
+    externalConfigName: ""
+  # The runtime class name to be used by the device plugin, and added to the pod.spec.runtimeClassName of applications utilizing NVIDIA GPUs
+  runtimeClassName: ""
+  # Whether to create runtime class, name comes from runtimeClassName when it is set
+  createRuntimeClass: false
+  migStrategy: "none"
+  disablecorelimit: "false"
+  passDeviceSpecsEnabled: false
+  deviceListStrategy: "envvar"
+  nvidiaHookPath: null
+  nvidiaDriverRoot: null
+  gdrcopyEnabled: null
+  gdsEnabled: null
+  mofedEnabled: null
+
+  extraArgs:
+    - -v=4
+  extraEnvs: {}
+  service:
+    type: NodePort  # Default type is NodePort, can be changed to ClusterIP
+    httpPort: 31992
+    labels: {}
+    annotations: {}
+
+  pluginPath: /var/lib/kubelet/device-plugins
+  libPath: /usr/local/vgpu
+
+  podAnnotations: {}
+  nvidiaNodeSelector:
+    gpu: "on"
+  tolerations: []
+  # The updateStrategy for DevicePlugin DaemonSet.
+  # If you want to update the DaemonSet by manual, set type as "OnDelete".
+  # We recommend use OnDelete update strategy because DevicePlugin pod restart will cause business pod restart, this behavior is destructive.
+  # Otherwise, you can use RollingUpdate update strategy to rolling update DevicePlugin pod.
+  updateStrategy:
+    type: RollingUpdate
+    rollingUpdate:
+      maxUnavailable: 1
+
+  resources: {}
+    # If you do want to specify resources, uncomment the following lines, adjust them as necessary.
+    # and remove the curly braces after 'resources:'.
+#    limits:
+#       cpu: 1000m
+#       memory: 1000Mi
+#    requests:
+#      cpu: 100m
+#      memory: 100Mi
+
+mockDevicePlugin:
+  enabled: false
+  image:
+    registry: "docker.io"
+    repository: "projecthami/mock-device-plugin"
+    tag: "1.0.1"
+    ## Specify a imagePullPolicy
+    ## Defaults to 'Always' if image tag is 'latest', else set to 'IfNotPresent'
+    ## ref: https://kubernetes.io/docs/user-guide/images/#pre-pulling-images
+    ##
+    pullPolicy: IfNotPresent
+    ## Optionally specify an array of imagePullSecrets.
+    ## Secrets must be manually created in the namespace.
+    ## Example:
+    ## pullSecrets:
+    ##   - myRegistryKeySecretName
+    ##
+    pullSecrets: []
+devices:
+  amd:
+    customresources:
+      - amd.com/gpu
+      - amd.com/gpu-memory
+  awsneuron:
+    customresources:
+      - aws.amazon.com/neuron
+      - aws.amazon.com/neuroncore
+  kunlun:
+    enabled: true
+    customresources:
+      - kunlunxin.com/xpu
+      - kunlunxin.com/vxpu
+      - kunlunxin.com/vxpu-memory
+  enflame:
+    enabled: true
+    customresources:
+      - enflame.com/vgcu
+      - enflame.com/vgcu-percentage
+      - enflame.com/gcu
+  mthreads:
+    enabled: true
+    customresources:
+      - mthreads.com/vgpu
+  nvidia:
+    gpuCorePolicy: default
+    libCudaLogLevel: 1
+  ascend:
+    enabled: false
+    image: ""
+    imagePullPolicy: IfNotPresent
+    extraArgs: []
+    nodeSelector:
+      ascend: "on"
+    tolerations: []
+    customresources:
+      - huawei.com/Ascend910A
+      - huawei.com/Ascend910A-memory
+      - huawei.com/Ascend910B2
+      - huawei.com/Ascend910B2-memory
+      - huawei.com/Ascend910B3
+      - huawei.com/Ascend910B3-memory
+      - huawei.com/Ascend910B4
+      - huawei.com/Ascend910B4-memory
+      - huawei.com/Ascend910B4-1
+      - huawei.com/Ascend910B4-1-memory
+      - huawei.com/Ascend310P
+      - huawei.com/Ascend310P-memory
+  iluvatar:
+    enabled: false
+    customresources:
+      - iluvatar.ai/BI-V100-vgpu
+      - iluvatar.ai/BI-V100.vCore
+      - iluvatar.ai/BI-V100.vMem
+      - iluvatar.ai/BI-V150-vgpu
+      - iluvatar.ai/BI-V150.vCore
+      - iluvatar.ai/BI-V150.vMem
+      - iluvatar.ai/MR-V100-vgpu
+      - iluvatar.ai/MR-V100.vCore
+      - iluvatar.ai/MR-V100.vMem
+      - iluvatar.ai/MR-V50-vgpu
+      - iluvatar.ai/MR-V50.vCore
+      - iluvatar.ai/MR-V50.vMem
diff --git a/packages/system/hami/values.yaml b/packages/system/hami/values.yaml
new file mode 100644
index 00000000..76e8d631
--- /dev/null
+++ b/packages/system/hami/values.yaml
@@ -0,0 +1,17 @@
+hami:
+  scheduler:
+    kubeScheduler:
+      image:
+        registry: registry.k8s.io
+        repository: kube-scheduler
+  devicePlugin:
+    runtimeClassName: nvidia
+    updateStrategy:
+      type: RollingUpdate
+    # See https://github.com/Project-HAMi/HAMi/blob/v2.8.1/deployments/charts/hami/values.yaml
+    # Each entry: {"name": "node-name", "operatingmode": "hami-core", "devicesplitcount": N, ...}
+    nodeConfiguration:
+      config: |
+        {
+          "nodeconfig": []
+        }
diff --git a/packages/system/harbor-rd/cozyrds/harbor.yaml b/packages/system/harbor-rd/cozyrds/harbor.yaml
index 5abc725f..5682c053 100644
--- a/packages/system/harbor-rd/cozyrds/harbor.yaml
+++ b/packages/system/harbor-rd/cozyrds/harbor.yaml
@@ -13,7 +13,6 @@ spec:
     prefix: "harbor-"
     labels:
       sharding.fluxcd.io/key: tenants
-      internal.cozystack.io/tenantmodule: "true"
     chartRef:
       kind: ExternalArtifact
       name: cozystack-harbor-application-default-harbor
@@ -24,7 +23,6 @@ spec:
     plural: Harbor
     name: harbor
     description: Managed Harbor container registry
-    module: true
     tags:
       - registry
       - container
diff --git a/packages/system/harbor/templates/database.yaml b/packages/system/harbor/templates/database.yaml
index 02e11faa..9a948601 100644
--- a/packages/system/harbor/templates/database.yaml
+++ b/packages/system/harbor/templates/database.yaml
@@ -5,7 +5,9 @@ metadata:
   name: {{ .Values.harbor.fullnameOverride }}-db
 spec:
   instances: {{ .Values.db.replicas }}
-  imageName: ghcr.io/cloudnative-pg/postgresql:17.7
+  {{- $existingCluster := lookup "postgresql.cnpg.io/v1" "Cluster" .Release.Namespace (printf "%s-db" .Values.harbor.fullnameOverride) }}
+  {{- $image := dig "spec" "imageName" "ghcr.io/cloudnative-pg/postgresql:17.7-standard-trixie" $existingCluster }}
+  imageName: {{ if regexMatch ":17\\." $image }}ghcr.io/cloudnative-pg/postgresql:17.7-standard-trixie{{ else if regexMatch ":[0-9]+\\.[0-9]+$" $image }}{{ printf "%s-standard-trixie" $image }}{{ else }}{{ $image }}{{ end }}
   storage:
     size: {{ .Values.db.size }}
     {{- with .Values.db.storageClass }}
diff --git a/packages/system/hetzner-robotlb/charts/robotlb/Chart.yaml b/packages/system/hetzner-robotlb/charts/robotlb/Chart.yaml
index 743f255d..192470dd 100644
--- a/packages/system/hetzner-robotlb/charts/robotlb/Chart.yaml
+++ b/packages/system/hetzner-robotlb/charts/robotlb/Chart.yaml
@@ -1,5 +1,5 @@
 apiVersion: v2
-appVersion: 0.0.5
+appVersion: 0.0.6
 description: A Helm chart for robotlb (loadbalancer on hetzner cloud).
 name: robotlb
 type: application
diff --git a/packages/system/hetzner-robotlb/charts/robotlb/templates/deployment.yaml b/packages/system/hetzner-robotlb/charts/robotlb/templates/deployment.yaml
index 4fd71366..41b4661a 100644
--- a/packages/system/hetzner-robotlb/charts/robotlb/templates/deployment.yaml
+++ b/packages/system/hetzner-robotlb/charts/robotlb/templates/deployment.yaml
@@ -5,7 +5,7 @@ metadata:
   labels:
     {{- include "robotlb.labels" . | nindent 4 }}
 spec:
-  replicas: {{ .Values.replicas }}
+  replicas: 1
   selector:
     matchLabels:
       {{- include "robotlb.selectorLabels" . | nindent 6 }}
diff --git a/packages/system/hetzner-robotlb/charts/robotlb/templates/role.yaml b/packages/system/hetzner-robotlb/charts/robotlb/templates/role.yaml
index 3a7b9334..76bac249 100644
--- a/packages/system/hetzner-robotlb/charts/robotlb/templates/role.yaml
+++ b/packages/system/hetzner-robotlb/charts/robotlb/templates/role.yaml
@@ -3,7 +3,8 @@ apiVersion: rbac.authorization.k8s.io/v1
 kind: ClusterRole
 metadata:
   name: {{ include "robotlb.fullname" . }}-cr
-rules: {{- toYaml .Values.serviceAccount.permissions | nindent 2 }}
+rules:
+ {{- toYaml .Values.serviceAccount.permissions | nindent 2 }}
 ---
 apiVersion: rbac.authorization.k8s.io/v1
 kind: ClusterRoleBinding
diff --git a/packages/system/hetzner-robotlb/charts/robotlb/values.yaml b/packages/system/hetzner-robotlb/charts/robotlb/values.yaml
index 739c7b73..4365f677 100644
--- a/packages/system/hetzner-robotlb/charts/robotlb/values.yaml
+++ b/packages/system/hetzner-robotlb/charts/robotlb/values.yaml
@@ -36,6 +36,9 @@ serviceAccount:
   - apiGroups: [""]
     resources: [nodes, pods]
     verbs: [get, list, watch]
+  - apiGroups: [discovery.k8s.io]
+    resources: [endpointslices]
+    verbs: [get, list, watch]
 
 podAnnotations: {}
 podLabels: {}
diff --git a/packages/system/kamaji/values.yaml b/packages/system/kamaji/values.yaml
index 9c6eea0e..6ba649f0 100644
--- a/packages/system/kamaji/values.yaml
+++ b/packages/system/kamaji/values.yaml
@@ -3,14 +3,37 @@ kamaji:
     deploy: false
   image:
     pullPolicy: IfNotPresent
-    tag: v1.2.1@sha256:34b9af510bbad3e64ea5cd7afdffa2b31c0751009c467355c8f67f8f9c5b3513
+    tag: v1.3.0@sha256:8c9af083b60600c0800eb56a2cda75f26007b7272a1cf019140de003bfce1a4d
     repository: ghcr.io/cozystack/cozystack/kamaji
   resources:
     limits:
       cpu: 200m
-      memory: 500Mi
+      memory: 512Mi
     requests:
       cpu: 100m
-      memory: 100Mi
+      memory: 256Mi
+  startupProbe:
+    httpGet:
+      path: /healthz
+      port: healthcheck
+    initialDelaySeconds: 5
+    periodSeconds: 5
+    timeoutSeconds: 3
+    successThreshold: 1
+    failureThreshold: 12
+  livenessProbe:
+    httpGet:
+      path: /healthz
+      port: healthcheck
+    initialDelaySeconds: 30
+    periodSeconds: 20
+    timeoutSeconds: 1
+  readinessProbe:
+    httpGet:
+      path: /readyz
+      port: healthcheck
+    initialDelaySeconds: 30
+    periodSeconds: 10
+    timeoutSeconds: 1
   extraArgs:
-    - --migrate-image=ghcr.io/cozystack/cozystack/kamaji:v1.2.1@sha256:34b9af510bbad3e64ea5cd7afdffa2b31c0751009c467355c8f67f8f9c5b3513
+    - --migrate-image=ghcr.io/cozystack/cozystack/kamaji:v1.3.0@sha256:8c9af083b60600c0800eb56a2cda75f26007b7272a1cf019140de003bfce1a4d
diff --git a/packages/system/keycloak/templates/db.yaml b/packages/system/keycloak/templates/db.yaml
index fb649a43..6a57e93a 100644
--- a/packages/system/keycloak/templates/db.yaml
+++ b/packages/system/keycloak/templates/db.yaml
@@ -4,7 +4,9 @@ metadata:
   name: keycloak-db
 spec:
   instances: 2
-  imageName: ghcr.io/cloudnative-pg/postgresql:17.7
+  {{- $existingCluster := lookup "postgresql.cnpg.io/v1" "Cluster" .Release.Namespace "keycloak-db" }}
+  {{- $image := dig "spec" "imageName" "ghcr.io/cloudnative-pg/postgresql:17.7-standard-trixie" $existingCluster }}
+  imageName: {{ if regexMatch ":17\\." $image }}ghcr.io/cloudnative-pg/postgresql:17.7-standard-trixie{{ else if regexMatch ":[0-9]+\\.[0-9]+$" $image }}{{ printf "%s-standard-trixie" $image }}{{ else }}{{ $image }}{{ end }}
   storage:
     size: 20Gi
   {{- if .Values._cluster.scheduling }}
diff --git a/packages/system/keycloak/templates/ingress.yaml b/packages/system/keycloak/templates/ingress.yaml
index aa282f71..7f9fd476 100644
--- a/packages/system/keycloak/templates/ingress.yaml
+++ b/packages/system/keycloak/templates/ingress.yaml
@@ -11,7 +11,7 @@ metadata:
   {{- with .Values.ingress.annotations }}
   annotations:
     {{- if eq $solver "http01" }}
-    acme.cert-manager.io/http01-ingress-class: {{ $exposeIngress }}
+    acme.cert-manager.io/http01-ingress-ingressclassname: {{ $exposeIngress }}
     {{- end }}
     cert-manager.io/cluster-issuer: {{ $clusterIssuer }}
     {{- toYaml . | nindent 4 }}
diff --git a/packages/system/kubeovn-plunger/templates/deployment.yaml b/packages/system/kubeovn-plunger/templates/deployment.yaml
index 183e8c5e..37b49620 100644
--- a/packages/system/kubeovn-plunger/templates/deployment.yaml
+++ b/packages/system/kubeovn-plunger/templates/deployment.yaml
@@ -29,6 +29,8 @@ spec:
         {{- end }}
         - --metrics-bind-address=:8080
         - --metrics-secure=false
+        - --kube-ovn-namespace={{ .Release.Namespace }}
+        - --ovn-central-name={{ .Values.ovnCentralName }}
         ports:
         - name: metrics
           containerPort: 8080
diff --git a/packages/system/kubeovn-plunger/templates/rbac.yaml b/packages/system/kubeovn-plunger/templates/rbac.yaml
index a0c5527b..3ca22740 100644
--- a/packages/system/kubeovn-plunger/templates/rbac.yaml
+++ b/packages/system/kubeovn-plunger/templates/rbac.yaml
@@ -22,8 +22,6 @@ rules:
   - get
   - list
   - watch
-  resourceNames:
-  - {{ .Values.ovnCentralName }}
 ---
 apiVersion: rbac.authorization.k8s.io/v1
 kind: RoleBinding
diff --git a/packages/system/kubeovn-plunger/values.yaml b/packages/system/kubeovn-plunger/values.yaml
index 7691803e..02620593 100644
--- a/packages/system/kubeovn-plunger/values.yaml
+++ b/packages/system/kubeovn-plunger/values.yaml
@@ -1,4 +1,4 @@
 portSecurity: true
 routes: ""
-image: ghcr.io/cozystack/cozystack/kubeovn-plunger:v1.2.1@sha256:71d487b9da211061921c7183d5a1e19a32c64dffe33a3fac0557c4cb25764dfe
+image: ghcr.io/cozystack/cozystack/kubeovn-plunger:v1.3.0@sha256:b75a0facb99c3b0fe8090414b3425f5c4858fff632d1de122ce9944c4daa89c0
 ovnCentralName: ovn-central
diff --git a/packages/system/kubeovn-webhook/values.yaml b/packages/system/kubeovn-webhook/values.yaml
index df5215d1..0c6ce81a 100644
--- a/packages/system/kubeovn-webhook/values.yaml
+++ b/packages/system/kubeovn-webhook/values.yaml
@@ -1,3 +1,3 @@
 portSecurity: true
 routes: ""
-image: ghcr.io/cozystack/cozystack/kubeovn-webhook:v1.2.1@sha256:e6334c29d3aaf0dea766c88e3e05b53ad623d1bb497b3c836e6f76adade45b29
+image: ghcr.io/cozystack/cozystack/kubeovn-webhook:v1.3.0@sha256:e18f9fd679e38f65362a8d0042f25468272f6d081136ad47027168d8e7e07a4a
diff --git a/packages/system/kubeovn/charts/kube-ovn/Chart.yaml b/packages/system/kubeovn/charts/kube-ovn/Chart.yaml
index f0295b87..9c01c123 100644
--- a/packages/system/kubeovn/charts/kube-ovn/Chart.yaml
+++ b/packages/system/kubeovn/charts/kube-ovn/Chart.yaml
@@ -15,12 +15,12 @@ type: application
 # This is the chart version. This version number should be incremented each time you make changes
 # to the chart and its templates, including the app version.
 # Versions are expected to follow Semantic Versioning (https://semver.org/)
-version: v1.15.3
+version: v1.15.10
 
 # This is the version number of the application being deployed. This version number should be
 # incremented each time you make changes to the application. Versions are not expected to
 # follow Semantic Versioning. They should reflect the version the application is using.
 # It is recommended to use it with quotes.
-appVersion: "1.15.3"
+appVersion: "1.15.10"
 
 kubeVersion: ">= 1.29.0-0"
diff --git a/packages/system/kubeovn/charts/kube-ovn/templates/kube-ovn-crd.yaml b/packages/system/kubeovn/charts/kube-ovn/templates/kube-ovn-crd.yaml
index 5b093be6..62ae328e 100644
--- a/packages/system/kubeovn/charts/kube-ovn/templates/kube-ovn-crd.yaml
+++ b/packages/system/kubeovn/charts/kube-ovn/templates/kube-ovn-crd.yaml
@@ -1353,6 +1353,77 @@ spec:
                         type: string
                     type: object
                   type: array
+                resources:
+                  description: |-
+                    `resources` are the compute resources required by this container.
+                    For more information, see https://kubernetes.io/docs/concepts/configuration/manage-resources-containers/
+                  properties:
+                    claims:
+                      description: |-
+                        Claims lists the names of resources, defined in spec.resourceClaims, that are used by this container.
+
+                        This field depends on the DynamicResourceAllocation feature gate.
+
+                        This field is immutable. It can only be set for containers.
+                      items:
+                        description: ResourceClaim references one entry in PodSpec.ResourceClaims.
+                        properties:
+                          name:
+                            description: |-
+                              Name must match the name of one entry in pod.spec.resourceClaims of
+                              the Pod where this field is used. It makes that resource available
+                              inside a container.
+                            type: string
+                          request:
+                            description: |-
+                              Request is the name chosen for a request in the referenced claim.
+                              If empty, everything from the claim is made available, otherwise
+                              only the result of this request.
+                            type: string
+                        required:
+                          - name
+                        type: object
+                      type: array
+                      x-kubernetes-list-map-keys:
+                        - name
+                      x-kubernetes-list-type: map
+                    limits:
+                      additionalProperties:
+                        anyOf:
+                        - type: integer
+                        - type: string
+                        pattern: ^(\+|-)?(([0-9]+(\.[0-9]*)?)|(\.[0-9]+))(([KMGTPE]i)|[numkMGTPE]|([eE](\+|-)?(([0-9]+(\.[0-9]*)?)|(\.[0-9]+))))?$
+                        x-kubernetes-int-or-string: true
+                      description: |-
+                        Limits describes the maximum amount of compute resources allowed.
+                        More info: https://kubernetes.io/docs/concepts/configuration/manage-resources-containers/
+                      type: object
+                    requests:
+                      additionalProperties:
+                        anyOf:
+                          - type: integer
+                          - type: string
+                        pattern: ^(\+|-)?(([0-9]+(\.[0-9]*)?)|(\.[0-9]+))(([KMGTPE]i)|[numkMGTPE]|([eE](\+|-)?(([0-9]+(\.[0-9]*)?)|(\.[0-9]+))))?$
+                        x-kubernetes-int-or-string: true
+                      description: |-
+                        Requests describes the minimum amount of compute resources required.
+                        If Requests is omitted for a container, it defaults to Limits if that is explicitly specified,
+                        otherwise to an implementation-defined value. Requests cannot exceed Limits.
+                        More info: https://kubernetes.io/docs/concepts/configuration/manage-resources-containers/
+                      type: object
+                  type: object
+                bandwidth:
+                  type: object
+                  description: Optional bandwidth limit for each egress gateway instance in both ingress and egress directions.
+                  properties:
+                    ingress:
+                      type: integer
+                      format: int64
+                      description: ingress bandwidth limit in Mbps
+                    egress:
+                      type: integer
+                      format: int64
+                      description: egress bandwidth limit in Mbps
 ---
 apiVersion: apiextensions.k8s.io/v1
 kind: CustomResourceDefinition
diff --git a/packages/system/kubeovn/charts/kube-ovn/templates/ovncni-ds.yaml b/packages/system/kubeovn/charts/kube-ovn/templates/ovncni-ds.yaml
index 8be8eba6..06971d01 100644
--- a/packages/system/kubeovn/charts/kube-ovn/templates/ovncni-ds.yaml
+++ b/packages/system/kubeovn/charts/kube-ovn/templates/ovncni-ds.yaml
@@ -126,6 +126,9 @@ spec:
           - --enable-ovn-ipsec={{- .Values.func.ENABLE_OVN_IPSEC }}
           - --set-vxlan-tx-off={{- .Values.func.SET_VXLAN_TX_OFF }}
           - --non-primary-cni-mode={{- .Values.cni_conf.NON_PRIMARY_CNI }}
+          {{- with .Values.mtu }}
+          - --mtu={{ . }}
+          {{- end }}
         securityContext:
           runAsUser: 0
           privileged: false
diff --git a/packages/system/kubeovn/charts/kube-ovn/templates/pre-upgrade-ovs-ovn.yaml b/packages/system/kubeovn/charts/kube-ovn/templates/pre-upgrade-ovs-ovn.yaml
new file mode 100644
index 00000000..a18a4111
--- /dev/null
+++ b/packages/system/kubeovn/charts/kube-ovn/templates/pre-upgrade-ovs-ovn.yaml
@@ -0,0 +1,129 @@
+{{- if include "kubeovn.ovn.versionCompatibility" . -}}
+---
+apiVersion: v1
+kind: ServiceAccount
+metadata:
+  name: ovs-ovn-pre-upgrade
+  namespace: {{ .Values.namespace }}
+  annotations:
+    "helm.sh/hook": pre-upgrade
+    "helm.sh/hook-weight": "1"
+    "helm.sh/hook-delete-policy": hook-succeeded
+automountServiceAccountToken: false
+---
+apiVersion: rbac.authorization.k8s.io/v1
+kind: ClusterRole
+metadata:
+  annotations:
+    rbac.authorization.k8s.io/system-only: "true"
+    "helm.sh/hook": pre-upgrade
+    "helm.sh/hook-weight": "2"
+    "helm.sh/hook-delete-policy": hook-succeeded
+  name: system:ovs-ovn-pre-upgrade
+rules:
+  - apiGroups:
+      - ""
+    resources:
+      - pods
+    verbs:
+      - list
+      - get
+  - apiGroups:
+      - ""
+    resources:
+      - pods/exec
+    verbs:
+      - create
+---
+apiVersion: rbac.authorization.k8s.io/v1
+kind: ClusterRoleBinding
+metadata:
+  name: ovs-ovn-pre-upgrade
+  annotations:
+    "helm.sh/hook": pre-upgrade
+    "helm.sh/hook-weight": "3"
+    "helm.sh/hook-delete-policy": hook-succeeded
+roleRef:
+  name: system:ovs-ovn-pre-upgrade
+  kind: ClusterRole
+  apiGroup: rbac.authorization.k8s.io
+subjects:
+  - kind: ServiceAccount
+    name: ovs-ovn-pre-upgrade
+    namespace: {{ .Values.namespace }}
+---
+apiVersion: batch/v1
+kind: Job
+metadata:
+  name: "{{ .Chart.Name }}-pre-upgrade-hook"
+  namespace: {{ .Values.namespace }}
+  labels:
+    app.kubernetes.io/managed-by: {{ .Release.Service | quote }}
+    app.kubernetes.io/instance: {{ .Release.Name | quote }}
+    app.kubernetes.io/version: {{ .Chart.AppVersion }}
+    helm.sh/chart: "{{ .Chart.Name }}-{{ .Chart.Version }}"
+  annotations:
+    "helm.sh/hook": pre-upgrade
+    "helm.sh/hook-weight": "4"
+    "helm.sh/hook-delete-policy": hook-succeeded
+spec:
+  completions: 1
+  template:
+    metadata:
+      name: "{{ .Release.Name }}"
+      labels:
+        app.kubernetes.io/managed-by: {{ .Release.Service | quote }}
+        app.kubernetes.io/instance: {{ .Release.Name | quote }}
+        helm.sh/chart: "{{ .Chart.Name }}-{{ .Chart.Version }}"
+        app: pre-upgrade
+        component: job
+    spec:
+      tolerations:
+        - key: ""
+          operator: "Exists"
+          effect: "NoSchedule"
+      affinity:
+        podAntiAffinity:
+          requiredDuringSchedulingIgnoredDuringExecution:
+            - topologyKey: kubernetes.io/hostname
+              labelSelector:
+                matchExpressions:
+                  - key: app
+                    operator: In
+                    values:
+                      - pre-upgrade
+                  - key: component
+                    operator: In
+                    values:
+                      - job
+      restartPolicy: Never
+      hostNetwork: true
+      nodeSelector:
+        kubernetes.io/os: "linux"
+      serviceAccountName: ovs-ovn-pre-upgrade
+      automountServiceAccountToken: true
+      securityContext:
+        seccompProfile:
+          type: RuntimeDefault
+      containers:
+        - name: ovs-ovn-pre-upgrade
+          image: "{{ .Values.global.registry.address}}/{{ .Values.global.images.kubeovn.repository }}:{{ .Values.global.images.kubeovn.tag }}"
+          env:
+            - name: POD_NAMESPACE
+              valueFrom:
+                fieldRef:
+                  fieldPath: metadata.namespace
+          command:
+            - bash
+            - -eo
+            - pipefail
+            - -c
+            - bash /kube-ovn/pre-upgrade-ovs.sh 2>&1 | tee -a /var/log/kube-ovn/pre-upgrade-ovs.log
+          volumeMounts:
+            - mountPath: /var/log/kube-ovn
+              name: kube-ovn-log
+      volumes:
+        - name: kube-ovn-log
+          hostPath:
+            path: {{ .Values.log_conf.LOG_DIR }}/kube-ovn
+{{- end -}}
diff --git a/packages/system/kubeovn/charts/kube-ovn/values.yaml b/packages/system/kubeovn/charts/kube-ovn/values.yaml
index 3ffb7750..fd3472f6 100644
--- a/packages/system/kubeovn/charts/kube-ovn/values.yaml
+++ b/packages/system/kubeovn/charts/kube-ovn/values.yaml
@@ -8,11 +8,11 @@ global:
   images:
     kubeovn:
       repository: kube-ovn
-      tag: v1.15.3
+      tag: v1.15.10
     natgateway:
       repository: vpc-nat-gateway
       # Falls back to the same tag as kubeovn if empty
-      tag: v1.15.3
+      tag: v1.15.10
 
 image:
   pullPolicy: IfNotPresent
diff --git a/packages/system/kubeovn/values.yaml b/packages/system/kubeovn/values.yaml
index 6911f8aa..e9d2d10f 100644
--- a/packages/system/kubeovn/values.yaml
+++ b/packages/system/kubeovn/values.yaml
@@ -65,4 +65,4 @@ global:
   images:
     kubeovn:
       repository: kubeovn
-      tag: v1.15.3@sha256:fa53d5f254f640cb626329ad35d9e7aad647dd8e1e645e68f3f13c3659472a30
+      tag: v1.15.10@sha256:741299cbd0081a786a6b60c460fa3156b3a42a38141c559dd8ac031f50c5504f
diff --git a/packages/system/kubernetes-rd/cozyrds/kubernetes.yaml b/packages/system/kubernetes-rd/cozyrds/kubernetes.yaml
index 5e9e8f94..f81eaff0 100644
--- a/packages/system/kubernetes-rd/cozyrds/kubernetes.yaml
+++ b/packages/system/kubernetes-rd/cozyrds/kubernetes.yaml
@@ -2,13 +2,27 @@ apiVersion: cozystack.io/v1alpha1
 kind: ApplicationDefinition
 metadata:
   name: kubernetes
+  annotations:
+    # Kubernetes tenants boot a Kamaji control plane whose admin-kubeconfig
+    # Secret is provisioned asynchronously. Cold Kamaji start (image pull +
+    # etcd + apiserver Ready) plus admin-kubeconfig generation can exceed
+    # Flux helm-controller's default wait budget, causing remediation loops
+    # that uninstall the Cluster CR. This override is applied by cozystack-api
+    # to the HelmRelease Spec.Install.Timeout and Spec.Upgrade.Timeout.
+    #
+    # Coupling: the wait-for-kubeconfig init container in
+    # packages/apps/kubernetes/templates/_helpers.tpl hard-codes a 10m
+    # deadline chosen to stay strictly below this value so the pod's
+    # CrashLoopBackOff surfaces before flux remediation fires. If this
+    # annotation is raised, update that init deadline correspondingly.
+    release.cozystack.io/helm-install-timeout: "15m"
 spec:
   application:
     kind: Kubernetes
     singular: kubernetes
     plural: kuberneteses
     openAPISchema: |-
-      {"title":"Chart Values","type":"object","properties":{"storageClass":{"description":"StorageClass used to store the data.","type":"string","default":"replicated"},"nodeGroups":{"description":"Worker nodes configuration map.","type":"object","default":{"md0":{"ephemeralStorage":"20Gi","gpus":[],"instanceType":"u1.medium","maxReplicas":10,"minReplicas":0,"resources":{},"roles":["ingress-nginx"]}},"additionalProperties":{"type":"object","required":["ephemeralStorage","instanceType","maxReplicas","minReplicas","resources"],"properties":{"ephemeralStorage":{"description":"Ephemeral storage size.","default":"20Gi","pattern":"^(\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))(([KMGTPE]i)|[numkMGTPE]|([eE](\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))))?$","anyOf":[{"type":"integer"},{"type":"string"}],"x-kubernetes-int-or-string":true},"gpus":{"description":"List of GPUs to attach (NVIDIA driver requires at least 4 GiB RAM).","type":"array","items":{"type":"object","required":["name"],"properties":{"name":{"description":"Name of GPU, such as \"nvidia.com/AD102GL_L40S\".","type":"string"}}}},"instanceType":{"description":"Virtual machine instance type.","type":"string","default":"u1.medium"},"maxReplicas":{"description":"Maximum number of replicas.","type":"integer","default":10},"minReplicas":{"description":"Minimum number of replicas.","type":"integer","default":0},"resources":{"description":"CPU and memory resources for each worker node.","type":"object","properties":{"cpu":{"description":"CPU available.","pattern":"^(\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))(([KMGTPE]i)|[numkMGTPE]|([eE](\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))))?$","anyOf":[{"type":"integer"},{"type":"string"}],"x-kubernetes-int-or-string":true},"memory":{"description":"Memory (RAM) available.","pattern":"^(\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))(([KMGTPE]i)|[numkMGTPE]|([eE](\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))))?$","anyOf":[{"type":"integer"},{"type":"string"}],"x-kubernetes-int-or-string":true}}},"roles":{"description":"List of node roles.","type":"array","items":{"type":"string"}}}}},"version":{"description":"Kubernetes major.minor version to deploy","type":"string","default":"v1.35","enum":["v1.35","v1.34","v1.33","v1.32","v1.31","v1.30"]},"host":{"description":"External hostname for Kubernetes cluster. Defaults to `<cluster-name>.<tenant-host>` if empty.","type":"string","default":""},"addons":{"description":"Cluster addons configuration.","type":"object","default":{},"required":["certManager","cilium","coredns","fluxcd","gatewayAPI","gpuOperator","ingressNginx","monitoringAgents","velero","verticalPodAutoscaler"],"properties":{"certManager":{"description":"Cert-manager addon.","type":"object","default":{},"required":["enabled","valuesOverride"],"properties":{"enabled":{"description":"Enable cert-manager.","type":"boolean","default":false},"valuesOverride":{"description":"Custom Helm values overrides.","type":"object","default":{},"x-kubernetes-preserve-unknown-fields":true}}},"cilium":{"description":"Cilium CNI plugin.","type":"object","default":{},"required":["valuesOverride"],"properties":{"valuesOverride":{"description":"Custom Helm values overrides.","type":"object","default":{},"x-kubernetes-preserve-unknown-fields":true}}},"coredns":{"description":"CoreDNS addon.","type":"object","default":{},"required":["valuesOverride"],"properties":{"valuesOverride":{"description":"Custom Helm values overrides.","type":"object","default":{},"x-kubernetes-preserve-unknown-fields":true}}},"fluxcd":{"description":"FluxCD GitOps operator.","type":"object","default":{},"required":["enabled","valuesOverride"],"properties":{"enabled":{"description":"Enable FluxCD.","type":"boolean","default":false},"valuesOverride":{"description":"Custom Helm values overrides.","type":"object","default":{},"x-kubernetes-preserve-unknown-fields":true}}},"gatewayAPI":{"description":"Gateway API addon.","type":"object","default":{},"required":["enabled"],"properties":{"enabled":{"description":"Enable Gateway API.","type":"boolean","default":false}}},"gpuOperator":{"description":"NVIDIA GPU Operator.","type":"object","default":{},"required":["enabled","valuesOverride"],"properties":{"enabled":{"description":"Enable GPU Operator.","type":"boolean","default":false},"valuesOverride":{"description":"Custom Helm values overrides.","type":"object","default":{},"x-kubernetes-preserve-unknown-fields":true}}},"ingressNginx":{"description":"Ingress-NGINX controller.","type":"object","default":{},"required":["enabled","exposeMethod","valuesOverride"],"properties":{"enabled":{"description":"Enable the controller (requires nodes labeled `ingress-nginx`).","type":"boolean","default":false},"exposeMethod":{"description":"Method to expose the controller. Allowed values: `Proxied`, `LoadBalancer`.","type":"string","default":"Proxied","enum":["Proxied","LoadBalancer"]},"hosts":{"description":"Domains routed to this tenant cluster when `exposeMethod` is `Proxied`.","type":"array","default":[],"items":{"type":"string"}},"valuesOverride":{"description":"Custom Helm values overrides.","type":"object","default":{},"x-kubernetes-preserve-unknown-fields":true}}},"monitoringAgents":{"description":"Monitoring agents.","type":"object","default":{},"required":["enabled","valuesOverride"],"properties":{"enabled":{"description":"Enable monitoring agents.","type":"boolean","default":false},"valuesOverride":{"description":"Custom Helm values overrides.","type":"object","default":{},"x-kubernetes-preserve-unknown-fields":true}}},"velero":{"description":"Velero backup/restore addon.","type":"object","default":{},"required":["enabled","valuesOverride"],"properties":{"enabled":{"description":"Enable Velero.","type":"boolean","default":false},"valuesOverride":{"description":"Custom Helm values overrides.","type":"object","default":{},"x-kubernetes-preserve-unknown-fields":true}}},"verticalPodAutoscaler":{"description":"Vertical Pod Autoscaler.","type":"object","default":{},"required":["valuesOverride"],"properties":{"valuesOverride":{"description":"Custom Helm values overrides.","type":"object","default":{},"x-kubernetes-preserve-unknown-fields":true}}}}},"controlPlane":{"description":"Kubernetes control-plane configuration.","type":"object","default":{},"required":["apiServer","controllerManager","konnectivity","replicas","scheduler"],"properties":{"apiServer":{"description":"API Server configuration.","type":"object","default":{},"required":["resources","resourcesPreset"],"properties":{"resources":{"description":"CPU and memory resources for API Server.","type":"object","default":{},"properties":{"cpu":{"description":"CPU available.","pattern":"^(\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))(([KMGTPE]i)|[numkMGTPE]|([eE](\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))))?$","anyOf":[{"type":"integer"},{"type":"string"}],"x-kubernetes-int-or-string":true},"memory":{"description":"Memory (RAM) available.","pattern":"^(\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))(([KMGTPE]i)|[numkMGTPE]|([eE](\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))))?$","anyOf":[{"type":"integer"},{"type":"string"}],"x-kubernetes-int-or-string":true}}},"resourcesPreset":{"description":"Preset if `resources` omitted.","type":"string","default":"large","enum":["nano","micro","small","medium","large","xlarge","2xlarge"]}}},"controllerManager":{"description":"Controller Manager configuration.","type":"object","default":{},"required":["resources","resourcesPreset"],"properties":{"resources":{"description":"CPU and memory resources for Controller Manager.","type":"object","default":{},"properties":{"cpu":{"description":"CPU available.","pattern":"^(\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))(([KMGTPE]i)|[numkMGTPE]|([eE](\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))))?$","anyOf":[{"type":"integer"},{"type":"string"}],"x-kubernetes-int-or-string":true},"memory":{"description":"Memory (RAM) available.","pattern":"^(\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))(([KMGTPE]i)|[numkMGTPE]|([eE](\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))))?$","anyOf":[{"type":"integer"},{"type":"string"}],"x-kubernetes-int-or-string":true}}},"resourcesPreset":{"description":"Preset if `resources` omitted.","type":"string","default":"micro","enum":["nano","micro","small","medium","large","xlarge","2xlarge"]}}},"konnectivity":{"description":"Konnectivity configuration.","type":"object","default":{},"required":["server"],"properties":{"server":{"description":"Konnectivity Server configuration.","type":"object","default":{},"required":["resources","resourcesPreset"],"properties":{"resources":{"description":"CPU and memory resources for Konnectivity.","type":"object","default":{},"properties":{"cpu":{"description":"CPU available.","pattern":"^(\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))(([KMGTPE]i)|[numkMGTPE]|([eE](\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))))?$","anyOf":[{"type":"integer"},{"type":"string"}],"x-kubernetes-int-or-string":true},"memory":{"description":"Memory (RAM) available.","pattern":"^(\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))(([KMGTPE]i)|[numkMGTPE]|([eE](\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))))?$","anyOf":[{"type":"integer"},{"type":"string"}],"x-kubernetes-int-or-string":true}}},"resourcesPreset":{"description":"Preset if `resources` omitted.","type":"string","default":"micro","enum":["nano","micro","small","medium","large","xlarge","2xlarge"]}}}}},"replicas":{"description":"Number of control-plane replicas.","type":"integer","default":2},"scheduler":{"description":"Scheduler configuration.","type":"object","default":{},"required":["resources","resourcesPreset"],"properties":{"resources":{"description":"CPU and memory resources for Scheduler.","type":"object","default":{},"properties":{"cpu":{"description":"CPU available.","pattern":"^(\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))(([KMGTPE]i)|[numkMGTPE]|([eE](\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))))?$","anyOf":[{"type":"integer"},{"type":"string"}],"x-kubernetes-int-or-string":true},"memory":{"description":"Memory (RAM) available.","pattern":"^(\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))(([KMGTPE]i)|[numkMGTPE]|([eE](\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))))?$","anyOf":[{"type":"integer"},{"type":"string"}],"x-kubernetes-int-or-string":true}}},"resourcesPreset":{"description":"Preset if `resources` omitted.","type":"string","default":"micro","enum":["nano","micro","small","medium","large","xlarge","2xlarge"]}}}}}}}
+      {"title":"Chart Values","type":"object","properties":{"storageClass":{"description":"StorageClass used to store the data.","type":"string","default":"replicated"},"nodeGroups":{"description":"Worker nodes configuration map.","type":"object","default":{"md0":{"ephemeralStorage":"20Gi","gpus":[],"instanceType":"u1.medium","maxReplicas":10,"minReplicas":0,"resources":{},"roles":["ingress-nginx"]}},"additionalProperties":{"type":"object","required":["ephemeralStorage","instanceType","maxReplicas","minReplicas","resources"],"properties":{"ephemeralStorage":{"description":"Ephemeral storage size.","default":"20Gi","pattern":"^(\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))(([KMGTPE]i)|[numkMGTPE]|([eE](\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))))?$","anyOf":[{"type":"integer"},{"type":"string"}],"x-kubernetes-int-or-string":true},"gpus":{"description":"List of GPUs to attach (NVIDIA driver requires at least 4 GiB RAM).","type":"array","items":{"type":"object","required":["name"],"properties":{"name":{"description":"Name of GPU, such as \"nvidia.com/AD102GL_L40S\".","type":"string"}}}},"instanceType":{"description":"Virtual machine instance type.","type":"string","default":"u1.medium"},"maxReplicas":{"description":"Maximum number of replicas.","type":"integer","default":10},"minReplicas":{"description":"Minimum number of replicas.","type":"integer","default":0},"resources":{"description":"CPU and memory resources for each worker node.","type":"object","properties":{"cpu":{"description":"CPU available.","pattern":"^(\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))(([KMGTPE]i)|[numkMGTPE]|([eE](\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))))?$","anyOf":[{"type":"integer"},{"type":"string"}],"x-kubernetes-int-or-string":true},"memory":{"description":"Memory (RAM) available.","pattern":"^(\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))(([KMGTPE]i)|[numkMGTPE]|([eE](\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))))?$","anyOf":[{"type":"integer"},{"type":"string"}],"x-kubernetes-int-or-string":true}}},"roles":{"description":"List of node roles.","type":"array","items":{"type":"string"}}}}},"version":{"description":"Kubernetes major.minor version to deploy","type":"string","default":"v1.35","enum":["v1.35","v1.34","v1.33","v1.32","v1.31","v1.30"]},"host":{"description":"External hostname for Kubernetes cluster. Defaults to `<cluster-name>.<tenant-host>` if empty.","type":"string","default":""},"addons":{"description":"Cluster addons configuration.","type":"object","default":{},"required":["certManager","cilium","coredns","fluxcd","gatewayAPI","gpuOperator","hami","ingressNginx","monitoringAgents","velero","verticalPodAutoscaler"],"properties":{"certManager":{"description":"Cert-manager addon.","type":"object","default":{},"required":["enabled","valuesOverride"],"properties":{"enabled":{"description":"Enable cert-manager.","type":"boolean","default":false},"valuesOverride":{"description":"Custom Helm values overrides.","type":"object","default":{},"x-kubernetes-preserve-unknown-fields":true}}},"cilium":{"description":"Cilium CNI plugin.","type":"object","default":{},"required":["valuesOverride"],"properties":{"valuesOverride":{"description":"Custom Helm values overrides.","type":"object","default":{},"x-kubernetes-preserve-unknown-fields":true}}},"coredns":{"description":"CoreDNS addon.","type":"object","default":{},"required":["valuesOverride"],"properties":{"valuesOverride":{"description":"Custom Helm values overrides.","type":"object","default":{},"x-kubernetes-preserve-unknown-fields":true}}},"fluxcd":{"description":"FluxCD GitOps operator.","type":"object","default":{},"required":["enabled","valuesOverride"],"properties":{"enabled":{"description":"Enable FluxCD.","type":"boolean","default":false},"valuesOverride":{"description":"Custom Helm values overrides.","type":"object","default":{},"x-kubernetes-preserve-unknown-fields":true}}},"gatewayAPI":{"description":"Gateway API addon.","type":"object","default":{},"required":["enabled"],"properties":{"enabled":{"description":"Enable Gateway API.","type":"boolean","default":false}}},"gpuOperator":{"description":"NVIDIA GPU Operator.","type":"object","default":{},"required":["enabled","valuesOverride"],"properties":{"enabled":{"description":"Enable GPU Operator.","type":"boolean","default":false},"valuesOverride":{"description":"Custom Helm values overrides.","type":"object","default":{},"x-kubernetes-preserve-unknown-fields":true}}},"hami":{"description":"HAMi GPU virtualization middleware.","type":"object","default":{},"required":["enabled","valuesOverride"],"properties":{"enabled":{"description":"Enable HAMi (requires GPU Operator).","type":"boolean","default":false},"valuesOverride":{"description":"Custom Helm values overrides.","type":"object","default":{},"x-kubernetes-preserve-unknown-fields":true}}},"ingressNginx":{"description":"Ingress-NGINX controller.","type":"object","default":{},"required":["enabled","exposeMethod","valuesOverride"],"properties":{"enabled":{"description":"Enable the controller (requires nodes labeled `ingress-nginx`).","type":"boolean","default":false},"exposeMethod":{"description":"Method to expose the controller. Allowed values: `Proxied`, `LoadBalancer`.","type":"string","default":"Proxied","enum":["Proxied","LoadBalancer"]},"hosts":{"description":"Domains routed to this tenant cluster when `exposeMethod` is `Proxied`.","type":"array","default":[],"items":{"type":"string"}},"valuesOverride":{"description":"Custom Helm values overrides.","type":"object","default":{},"x-kubernetes-preserve-unknown-fields":true}}},"monitoringAgents":{"description":"Monitoring agents.","type":"object","default":{},"required":["enabled","valuesOverride"],"properties":{"enabled":{"description":"Enable monitoring agents.","type":"boolean","default":false},"valuesOverride":{"description":"Custom Helm values overrides.","type":"object","default":{},"x-kubernetes-preserve-unknown-fields":true}}},"velero":{"description":"Velero backup/restore addon.","type":"object","default":{},"required":["enabled","valuesOverride"],"properties":{"enabled":{"description":"Enable Velero.","type":"boolean","default":false},"valuesOverride":{"description":"Custom Helm values overrides.","type":"object","default":{},"x-kubernetes-preserve-unknown-fields":true}}},"verticalPodAutoscaler":{"description":"Vertical Pod Autoscaler.","type":"object","default":{},"required":["valuesOverride"],"properties":{"valuesOverride":{"description":"Custom Helm values overrides.","type":"object","default":{},"x-kubernetes-preserve-unknown-fields":true}}}}},"controlPlane":{"description":"Kubernetes control-plane configuration.","type":"object","default":{},"required":["apiServer","controllerManager","konnectivity","replicas","scheduler"],"properties":{"apiServer":{"description":"API Server configuration.","type":"object","default":{},"required":["resources","resourcesPreset"],"properties":{"resources":{"description":"CPU and memory resources for API Server.","type":"object","default":{},"properties":{"cpu":{"description":"CPU available.","pattern":"^(\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))(([KMGTPE]i)|[numkMGTPE]|([eE](\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))))?$","anyOf":[{"type":"integer"},{"type":"string"}],"x-kubernetes-int-or-string":true},"memory":{"description":"Memory (RAM) available.","pattern":"^(\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))(([KMGTPE]i)|[numkMGTPE]|([eE](\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))))?$","anyOf":[{"type":"integer"},{"type":"string"}],"x-kubernetes-int-or-string":true}}},"resourcesPreset":{"description":"Preset if `resources` omitted.","type":"string","default":"large","enum":["nano","micro","small","medium","large","xlarge","2xlarge"]}}},"controllerManager":{"description":"Controller Manager configuration.","type":"object","default":{},"required":["resources","resourcesPreset"],"properties":{"resources":{"description":"CPU and memory resources for Controller Manager.","type":"object","default":{},"properties":{"cpu":{"description":"CPU available.","pattern":"^(\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))(([KMGTPE]i)|[numkMGTPE]|([eE](\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))))?$","anyOf":[{"type":"integer"},{"type":"string"}],"x-kubernetes-int-or-string":true},"memory":{"description":"Memory (RAM) available.","pattern":"^(\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))(([KMGTPE]i)|[numkMGTPE]|([eE](\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))))?$","anyOf":[{"type":"integer"},{"type":"string"}],"x-kubernetes-int-or-string":true}}},"resourcesPreset":{"description":"Preset if `resources` omitted.","type":"string","default":"micro","enum":["nano","micro","small","medium","large","xlarge","2xlarge"]}}},"konnectivity":{"description":"Konnectivity configuration.","type":"object","default":{},"required":["server"],"properties":{"server":{"description":"Konnectivity Server configuration.","type":"object","default":{},"required":["resources","resourcesPreset"],"properties":{"resources":{"description":"CPU and memory resources for Konnectivity.","type":"object","default":{},"properties":{"cpu":{"description":"CPU available.","pattern":"^(\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))(([KMGTPE]i)|[numkMGTPE]|([eE](\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))))?$","anyOf":[{"type":"integer"},{"type":"string"}],"x-kubernetes-int-or-string":true},"memory":{"description":"Memory (RAM) available.","pattern":"^(\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))(([KMGTPE]i)|[numkMGTPE]|([eE](\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))))?$","anyOf":[{"type":"integer"},{"type":"string"}],"x-kubernetes-int-or-string":true}}},"resourcesPreset":{"description":"Preset if `resources` omitted.","type":"string","default":"micro","enum":["nano","micro","small","medium","large","xlarge","2xlarge"]}}}}},"replicas":{"description":"Number of control-plane replicas.","type":"integer","default":2},"scheduler":{"description":"Scheduler configuration.","type":"object","default":{},"required":["resources","resourcesPreset"],"properties":{"resources":{"description":"CPU and memory resources for Scheduler.","type":"object","default":{},"properties":{"cpu":{"description":"CPU available.","pattern":"^(\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))(([KMGTPE]i)|[numkMGTPE]|([eE](\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))))?$","anyOf":[{"type":"integer"},{"type":"string"}],"x-kubernetes-int-or-string":true},"memory":{"description":"Memory (RAM) available.","pattern":"^(\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))(([KMGTPE]i)|[numkMGTPE]|([eE](\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))))?$","anyOf":[{"type":"integer"},{"type":"string"}],"x-kubernetes-int-or-string":true}}},"resourcesPreset":{"description":"Preset if `resources` omitted.","type":"string","default":"micro","enum":["nano","micro","small","medium","large","xlarge","2xlarge"]}}}}},"images":{"description":"Optional image overrides for air-gapped or rate-limited registries.","type":"object","default":{},"properties":{"waitForKubeconfig":{"description":"Image used by the wait-for-kubeconfig init container. Empty falls back to images/busybox.tag.","type":"string","default":""}}}}}
   release:
     prefix: kubernetes-
     labels:
@@ -26,7 +40,7 @@ spec:
     tags:
       - compute
     icon: <svg width="144" height="144" viewBox="0 0 144 144" fill="none" xmlns="http://www.w3.org/2000/svg">
<rect width="144" height="144" rx="24" fill="url(#paint0_linear_681_2845)"/>
<path d="M71.9968 19C70.3039 19.0002 68.9312 20.5322 68.9314 22.4221C68.9314 22.4511 68.9373 22.4788 68.9379 22.5076C68.9354 22.7644 68.9231 23.0737 68.9314 23.2973C68.9717 24.3873 69.2082 25.2216 69.3506 26.2258C69.6084 28.3752 69.8245 30.1569 69.6912 31.813C69.5615 32.4375 69.1037 33.0086 68.6956 33.4056L68.6235 34.7086C66.7839 34.8617 64.9319 35.1421 63.0821 35.5641C55.1226 37.3798 48.2695 41.4991 43.052 47.0609C42.7134 46.8288 42.1211 46.4019 41.945 46.2712C41.3977 46.3454 40.8445 46.5151 40.1241 46.0935C38.7522 45.1657 37.5028 43.8851 35.991 42.3424C35.2982 41.6044 34.7966 40.9018 33.9735 40.1904C33.7866 40.0289 33.5014 39.8104 33.2923 39.6442C32.6489 39.1288 31.89 38.86 31.157 38.8348C30.2147 38.8024 29.3075 39.1725 28.7138 39.9206C27.6584 41.2506 27.9963 43.2833 29.4671 44.4614C29.482 44.4734 29.4979 44.4827 29.5129 44.4943C29.715 44.6589 29.9625 44.8698 30.1483 45.0076C31.0217 45.6555 31.8195 45.9872 32.6897 46.5015C34.5231 47.6391 36.043 48.5823 37.2486 49.7196C37.7194 50.2237 37.8016 51.1122 37.8643 51.4964L38.8468 52.3782C33.5872 60.3308 31.153 70.1539 32.5915 80.1627L31.3077 80.5378C30.9693 80.9768 30.4912 81.6676 29.9911 81.8738C28.4138 82.3729 26.6387 82.5562 24.4956 82.7819C23.4894 82.866 22.6213 82.8158 21.5546 83.0188C21.3198 83.0635 20.9927 83.1491 20.7358 83.2097C20.7269 83.2116 20.7186 83.2142 20.7096 83.2162C20.6956 83.2195 20.6772 83.2263 20.6638 83.2294C18.857 83.668 17.6963 85.3365 18.0699 86.9805C18.4437 88.6248 20.2086 89.6248 22.0262 89.2312C22.0393 89.2282 22.0584 89.2277 22.072 89.2246C22.0926 89.2199 22.1106 89.2099 22.131 89.2049C22.3844 89.149 22.7019 89.0868 22.9236 89.0272C23.9723 88.7451 24.7318 88.3306 25.6746 87.9677C27.7029 87.2368 29.3828 86.6262 31.0195 86.3883C31.703 86.3345 32.4232 86.812 32.7814 87.0134L34.1177 86.7831C37.1926 96.3613 43.6366 104.103 51.7963 108.961L51.2396 110.303C51.4403 110.824 51.6616 111.53 51.5121 112.045C50.9171 113.595 49.898 115.231 48.7374 117.055C48.1755 117.898 47.6004 118.552 47.0934 119.516C46.972 119.747 46.8175 120.102 46.7004 120.346C45.9125 122.039 46.4904 123.99 48.0038 124.722C49.5268 125.459 51.4171 124.682 52.2352 122.985C52.2364 122.982 52.2406 122.98 52.2417 122.978C52.2426 122.976 52.2409 122.973 52.2417 122.971C52.3582 122.731 52.5233 122.415 52.6216 122.188C53.056 121.189 53.2005 120.332 53.5059 119.365C54.317 117.318 54.7626 115.17 55.8791 113.832C56.1849 113.466 56.6833 113.325 57.2001 113.186L57.8944 111.922C65.008 114.665 72.9705 115.402 80.9245 113.587C82.7391 113.173 84.4908 112.637 86.1843 111.994C86.3794 112.342 86.742 113.011 86.8393 113.179C87.3644 113.351 87.9377 113.439 88.4047 114.133C89.2401 115.567 89.8114 117.263 90.5073 119.312C90.8128 120.279 90.9638 121.136 91.3981 122.136C91.4971 122.363 91.6614 122.684 91.778 122.925C92.5944 124.628 94.4907 125.407 96.0159 124.669C97.5292 123.937 98.1077 121.986 97.3194 120.293C97.2023 120.049 97.0412 119.695 96.9198 119.464C96.4127 118.499 95.8377 117.852 95.2758 117.009C94.1152 115.185 93.1526 113.67 92.5575 112.12C92.3087 111.32 92.5995 110.823 92.7933 110.303C92.6772 110.17 92.4288 109.414 92.2824 109.059C100.762 104.029 107.017 95.9985 109.955 86.7239C110.351 86.7865 111.041 86.9091 111.265 86.9542C111.726 86.6487 112.149 86.2501 112.981 86.3159C114.617 86.5537 116.297 87.1645 118.326 87.8953C119.268 88.2581 120.028 88.6793 121.077 88.9614C121.298 89.021 121.616 89.0766 121.869 89.1325C121.889 89.1375 121.908 89.1475 121.928 89.1522C121.942 89.1553 121.961 89.1558 121.974 89.1588C123.792 89.552 125.557 88.5526 125.93 86.9081C126.303 85.2641 125.143 83.5952 123.336 83.157C123.074 83.097 122.701 82.995 122.446 82.9465C121.379 82.7435 120.511 82.7935 119.505 82.7095C117.361 82.4839 115.586 82.3004 114.009 81.8014C113.366 81.5507 112.908 80.7819 112.686 80.4655L111.448 80.1035C112.09 75.438 111.917 70.5825 110.806 65.7243C109.685 60.8208 107.704 56.3361 105.062 52.3848C105.379 52.0948 105.979 51.5612 106.149 51.4043C106.199 50.8517 106.156 50.2722 106.725 49.6603C107.931 48.523 109.451 47.5799 111.284 46.4423C112.154 45.9279 112.959 45.5964 113.832 44.9484C114.03 44.8019 114.299 44.5699 114.507 44.4022C115.977 43.2237 116.316 41.1911 115.26 39.8614C114.204 38.5317 112.159 38.4065 110.688 39.585C110.479 39.7516 110.195 39.9688 110.007 40.1312C109.184 40.8426 108.676 41.5452 107.983 42.2832C106.471 43.8259 105.222 45.113 103.85 46.0409C103.255 46.3885 102.385 46.2682 101.99 46.2449L100.824 47.0806C94.1753 40.0763 85.1235 35.5982 75.3766 34.7283C75.3494 34.3179 75.3137 33.5761 75.3046 33.3529C74.9056 32.9693 74.4235 32.6418 74.3024 31.813C74.1691 30.1569 74.3917 28.3752 74.6496 26.2258C74.7919 25.2216 75.0284 24.3873 75.0688 23.2973C75.078 23.0495 75.0632 22.69 75.0622 22.4221C75.062 20.5322 73.6898 18.9998 71.9968 19ZM68.1585 42.8886L67.248 59.0447L67.1825 59.0776C67.1214 60.5229 65.9375 61.677 64.4839 61.677C63.8884 61.677 63.3388 61.4849 62.8922 61.1571L62.866 61.1703L49.6807 51.7794C53.7331 47.7759 58.9164 44.8172 64.89 43.4546C65.9812 43.2056 67.0719 43.0209 68.1585 42.8886ZM75.8417 42.8886C82.8159 43.7504 89.2657 46.9232 94.2081 51.786L81.108 61.1176L81.0621 61.0979C79.8994 61.9512 78.2611 61.7394 77.3548 60.5978C76.9835 60.1301 76.7887 59.5801 76.7653 59.0249L76.7522 59.0184L75.8417 42.8886ZM44.8991 57.814L56.9382 68.633L56.9251 68.6988C58.0117 69.6479 58.172 71.2949 57.2657 72.4368C56.8944 72.9045 56.3975 73.2182 55.8639 73.3647L55.8508 73.4173L40.4188 77.8923C39.6334 70.6765 41.3261 63.6621 44.8991 57.814ZM99.0094 57.8206C100.798 60.7336 102.153 63.9871 102.959 67.5143C103.756 70.9991 103.956 74.4778 103.627 77.8397L88.1166 73.3515L88.1035 73.2857C86.7145 72.9043 85.8609 71.4848 86.1843 70.0611C86.3168 69.4778 86.6249 68.9844 87.0423 68.6198L87.0358 68.5869L99.0094 57.8206ZM69.5274 69.4688H74.4596L77.5251 73.3186L76.4247 78.1226L71.9968 80.2614L67.5558 78.1161L66.4554 73.312L69.5274 69.4688ZM85.3393 82.6437C85.5489 82.6331 85.7576 82.652 85.9616 82.6898L85.9878 82.6569L101.95 85.3682C99.6142 91.9624 95.144 97.675 89.1711 101.498L82.9747 86.4606L82.9944 86.4343C82.4252 85.1055 82.9948 83.5472 84.3044 82.9135C84.6397 82.7513 84.99 82.6614 85.3393 82.6437ZM58.5298 82.7095C59.748 82.7267 60.8406 83.5761 61.1237 84.822C61.2562 85.4052 61.1917 85.9831 60.973 86.4935L61.0189 86.5527L54.888 101.439C49.1559 97.7432 44.5904 92.2099 42.1481 85.4208L57.9731 82.7227L57.9993 82.7556C58.1763 82.7229 58.3558 82.7071 58.5298 82.7095ZM71.8986 89.2312C72.3229 89.2155 72.7534 89.303 73.1627 89.501C73.6992 89.7606 74.1136 90.1692 74.3745 90.6592H74.4334L82.2346 104.821C81.2221 105.162 80.1813 105.454 79.1167 105.697C73.1505 107.058 67.2032 106.645 61.818 104.802L69.5995 90.6658H69.6126C70.0795 89.7888 70.965 89.2656 71.8986 89.2312Z" fill="white" stroke="white" stroke-width="0.25"/>
<defs>
<linearGradient id="paint0_linear_681_2845" x1="10" y1="15.5" x2="144" y2="131.5" gradientUnits="userSpaceOnUse">
<stop stop-color="#4D87FF"/>
<stop offset="1" stop-color="#0547D0"/>
</linearGradient>
</defs>
</svg>

-    keysOrder: [["apiVersion"], ["appVersion"], ["kind"], ["metadata"], ["metadata", "name"], ["spec", "storageClass"], ["spec", "nodeGroups"], ["spec", "nodeGroups", "md0"], ["spec", "nodeGroups", "md0", "minReplicas"], ["spec", "nodeGroups", "md0", "maxReplicas"], ["spec", "nodeGroups", "md0", "instanceType"], ["spec", "nodeGroups", "md0", "ephemeralStorage"], ["spec", "nodeGroups", "md0", "roles"], ["spec", "nodeGroups", "md0", "resources"], ["spec", "nodeGroups", "md0", "gpus"], ["spec", "version"], ["spec", "host"], ["spec", "addons"], ["spec", "addons", "certManager"], ["spec", "addons", "certManager", "enabled"], ["spec", "addons", "certManager", "valuesOverride"], ["spec", "addons", "cilium"], ["spec", "addons", "cilium", "valuesOverride"], ["spec", "addons", "gatewayAPI"], ["spec", "addons", "gatewayAPI", "enabled"], ["spec", "addons", "ingressNginx"], ["spec", "addons", "ingressNginx", "enabled"], ["spec", "addons", "ingressNginx", "exposeMethod"], ["spec", "addons", "ingressNginx", "hosts"], ["spec", "addons", "ingressNginx", "valuesOverride"], ["spec", "addons", "gpuOperator"], ["spec", "addons", "gpuOperator", "enabled"], ["spec", "addons", "gpuOperator", "valuesOverride"], ["spec", "addons", "fluxcd"], ["spec", "addons", "fluxcd", "enabled"], ["spec", "addons", "fluxcd", "valuesOverride"], ["spec", "addons", "monitoringAgents"], ["spec", "addons", "monitoringAgents", "enabled"], ["spec", "addons", "monitoringAgents", "valuesOverride"], ["spec", "addons", "verticalPodAutoscaler"], ["spec", "addons", "verticalPodAutoscaler", "valuesOverride"], ["spec", "addons", "velero"], ["spec", "addons", "velero", "enabled"], ["spec", "addons", "velero", "valuesOverride"], ["spec", "addons", "coredns"], ["spec", "addons", "coredns", "valuesOverride"], ["spec", "controlPlane"], ["spec", "controlPlane", "replicas"], ["spec", "controlPlane", "apiServer"], ["spec", "controlPlane", "apiServer", "resources"], ["spec", "controlPlane", "apiServer", "resourcesPreset"], ["spec", "controlPlane", "controllerManager"], ["spec", "controlPlane", "controllerManager", "resources"], ["spec", "controlPlane", "controllerManager", "resourcesPreset"], ["spec", "controlPlane", "scheduler"], ["spec", "controlPlane", "scheduler", "resources"], ["spec", "controlPlane", "scheduler", "resourcesPreset"], ["spec", "controlPlane", "konnectivity"], ["spec", "controlPlane", "konnectivity", "server"], ["spec", "controlPlane", "konnectivity", "server", "resources"], ["spec", "controlPlane", "konnectivity", "server", "resourcesPreset"]]
+    keysOrder: [["apiVersion"], ["appVersion"], ["kind"], ["metadata"], ["metadata", "name"], ["spec", "storageClass"], ["spec", "nodeGroups"], ["spec", "nodeGroups", "md0"], ["spec", "nodeGroups", "md0", "minReplicas"], ["spec", "nodeGroups", "md0", "maxReplicas"], ["spec", "nodeGroups", "md0", "instanceType"], ["spec", "nodeGroups", "md0", "ephemeralStorage"], ["spec", "nodeGroups", "md0", "roles"], ["spec", "nodeGroups", "md0", "resources"], ["spec", "nodeGroups", "md0", "gpus"], ["spec", "version"], ["spec", "host"], ["spec", "addons"], ["spec", "addons", "certManager"], ["spec", "addons", "certManager", "enabled"], ["spec", "addons", "certManager", "valuesOverride"], ["spec", "addons", "cilium"], ["spec", "addons", "cilium", "valuesOverride"], ["spec", "addons", "gatewayAPI"], ["spec", "addons", "gatewayAPI", "enabled"], ["spec", "addons", "ingressNginx"], ["spec", "addons", "ingressNginx", "enabled"], ["spec", "addons", "ingressNginx", "exposeMethod"], ["spec", "addons", "ingressNginx", "hosts"], ["spec", "addons", "ingressNginx", "valuesOverride"], ["spec", "addons", "gpuOperator"], ["spec", "addons", "gpuOperator", "enabled"], ["spec", "addons", "gpuOperator", "valuesOverride"], ["spec", "addons", "hami"], ["spec", "addons", "hami", "enabled"], ["spec", "addons", "hami", "valuesOverride"], ["spec", "addons", "fluxcd"], ["spec", "addons", "fluxcd", "enabled"], ["spec", "addons", "fluxcd", "valuesOverride"], ["spec", "addons", "monitoringAgents"], ["spec", "addons", "monitoringAgents", "enabled"], ["spec", "addons", "monitoringAgents", "valuesOverride"], ["spec", "addons", "verticalPodAutoscaler"], ["spec", "addons", "verticalPodAutoscaler", "valuesOverride"], ["spec", "addons", "velero"], ["spec", "addons", "velero", "enabled"], ["spec", "addons", "velero", "valuesOverride"], ["spec", "addons", "coredns"], ["spec", "addons", "coredns", "valuesOverride"], ["spec", "controlPlane"], ["spec", "controlPlane", "replicas"], ["spec", "controlPlane", "apiServer"], ["spec", "controlPlane", "apiServer", "resources"], ["spec", "controlPlane", "apiServer", "resourcesPreset"], ["spec", "controlPlane", "controllerManager"], ["spec", "controlPlane", "controllerManager", "resources"], ["spec", "controlPlane", "controllerManager", "resourcesPreset"], ["spec", "controlPlane", "scheduler"], ["spec", "controlPlane", "scheduler", "resources"], ["spec", "controlPlane", "scheduler", "resourcesPreset"], ["spec", "controlPlane", "konnectivity"], ["spec", "controlPlane", "konnectivity", "server"], ["spec", "controlPlane", "konnectivity", "server", "resources"], ["spec", "controlPlane", "konnectivity", "server", "resourcesPreset"], ["spec", "images"], ["spec", "images", "waitForKubeconfig"]]
   secrets:
     exclude: []
     include:
diff --git a/packages/system/kubevirt-csi-node/values.yaml b/packages/system/kubevirt-csi-node/values.yaml
index ad7e6f01..56e28543 100644
--- a/packages/system/kubevirt-csi-node/values.yaml
+++ b/packages/system/kubevirt-csi-node/values.yaml
@@ -1,3 +1,3 @@
 storageClass: replicated
 csiDriver:
-  image: ghcr.io/cozystack/cozystack/kubevirt-csi-driver:0.0.0@sha256:5b209248acba3d8cf0d999edd8a915915a2e565a1c8ac1f5dde0dffec20fa02e
+  image: ghcr.io/cozystack/cozystack/kubevirt-csi-driver:0.0.0@sha256:72154a97054e16cdf3dea6129d962b8d7e86b55cf9386095e8ac2ce7c8b69172
diff --git a/packages/system/lineage-controller-webhook/Makefile b/packages/system/lineage-controller-webhook/Makefile
index d5bada31..cdc1cf01 100644
--- a/packages/system/lineage-controller-webhook/Makefile
+++ b/packages/system/lineage-controller-webhook/Makefile
@@ -4,8 +4,13 @@ NAMESPACE=cozy-system
 include ../../../hack/common-envs.mk
 include ../../../hack/package.mk
 
+.PHONY: test
+
 image: image-lineage-controller-webhook
 
+test:
+	helm unittest .
+
 image-lineage-controller-webhook:
 	docker buildx build -f images/lineage-controller-webhook/Dockerfile ../../.. \
 		--tag $(REGISTRY)/lineage-controller-webhook:$(call settag,$(TAG)) \
diff --git a/packages/system/lineage-controller-webhook/README.md b/packages/system/lineage-controller-webhook/README.md
new file mode 100644
index 00000000..1e43c9c8
--- /dev/null
+++ b/packages/system/lineage-controller-webhook/README.md
@@ -0,0 +1,67 @@
+# lineage-controller-webhook
+
+Cozystack system package for the **lineage controller webhook** — a mutating
+admission webhook that stamps "lineage" labels onto tenant workloads, linking
+each resource back to the Cozystack `Application` that ultimately owns it.
+
+The webhook intercepts CREATE and UPDATE on `pods`, `secrets`, `services`,
+`persistentvolumeclaims`, `ingresses.networking.k8s.io`, and
+`workloadmonitors.cozystack.io` outside system namespaces. For each request it
+walks the ownership graph upward (Kubernetes `ownerReferences`, then the
+`HelmRelease` Flux installed the resource with, then the Cozystack
+`Application`-derived CRD that produced the HelmRelease) and writes the
+discovered application's group, kind and name as labels
+(`apps.cozystack.io/application.{group,kind,name}`) on the incoming object.
+Those labels let the aggregated API server, the Cozystack dashboard, and other
+lineage-aware consumers reason about which application a resource belongs to.
+
+The webhook serves TLS on port 9443 with a cert-manager issued certificate, and
+runs with `failurePolicy: Fail` — every CREATE/UPDATE on the resources above
+is gated on the webhook being reachable.
+
+## Topology
+
+The chart ships a single shape, modelled on `cozystack-api`:
+
+- **Deployment** with two replicas (override via `replicas`).
+- **Soft `nodeAffinity`** preferring `node-role.kubernetes.io/control-plane`
+  (`Exists`, so it matches both Talos's empty value and k3s/kubeadm's `"true"`).
+  Soft means: the pod lands on a control-plane node when one is reachable, and
+  on any worker otherwise — no override needed for managed Kubernetes,
+  Cozy-in-Cozy tenants, or any other cluster where control-plane nodes aren't
+  visible.
+- **Permissive `tolerations`** (`[{operator: Exists}]`) so the pod can land on
+  tainted control-plane nodes when the soft affinity is satisfiable.
+- **Soft `podAntiAffinity`** on `kubernetes.io/hostname` so replicas spread
+  across nodes when possible (best-effort).
+- **`PodDisruptionBudget`** with `maxUnavailable: 1`, unconditional. At
+  `replicas: 1` it's a useful no-op (allows full drain); at `replicas >= 2`
+  it caps disruption to one pod.
+- **`Service` with `spec.trafficDistribution: PreferClose`** so the apiserver
+  prefers a webhook endpoint on its own node when one exists, and transparently
+  falls over to a remote endpoint otherwise. Requires Kubernetes ≥ 1.31; on
+  older clusters the field is silently ignored and traffic uses default
+  cluster-wide distribution.
+
+## Parameters
+
+All keys live under the top-level `lineageControllerWebhook:` map.
+
+| Name                          | Description                                               | Value                                                                            |
+| ----------------------------- | --------------------------------------------------------- | -------------------------------------------------------------------------------- |
+| `image`                       | Container image (digest-pinned)                           | `ghcr.io/cozystack/cozystack/lineage-controller-webhook:v1.3.0@sha256:e898…6fb0` |
+| `debug`                       | Enable `--zap-log-level=debug` instead of `info`          | `false`                                                                          |
+| `replicas`                    | Deployment replica count                                  | `2`                                                                              |
+| `localK8sAPIEndpoint.enabled` | **Deprecated.** See note below.                           | `false`                                                                          |
+
+### `localK8sAPIEndpoint.enabled` (deprecated)
+
+When enabled, this injects `KUBERNETES_SERVICE_HOST=status.hostIP` and
+`KUBERNETES_SERVICE_PORT=6443` so the webhook talks to the kube-apiserver on
+its own node. It was originally added to avoid latency on the
+webhook-to-apiserver path, but it is only valid when the pod is actually
+scheduled on an apiserver-bearing node — which the chart's soft control-plane
+affinity no longer guarantees. With this flag enabled and the pod scheduled
+off a control-plane node, the controller will crash-loop dialing a non-
+apiserver hostIP. Slated for removal once the latency motivation is addressed
+in the webhook itself; **leave disabled**.
diff --git a/packages/system/lineage-controller-webhook/templates/pdb.yaml b/packages/system/lineage-controller-webhook/templates/pdb.yaml
new file mode 100644
index 00000000..0786c8ff
--- /dev/null
+++ b/packages/system/lineage-controller-webhook/templates/pdb.yaml
@@ -0,0 +1,9 @@
+apiVersion: policy/v1
+kind: PodDisruptionBudget
+metadata:
+  name: lineage-controller-webhook
+spec:
+  maxUnavailable: 1
+  selector:
+    matchLabels:
+      app: lineage-controller-webhook
diff --git a/packages/system/lineage-controller-webhook/templates/service.yaml b/packages/system/lineage-controller-webhook/templates/service.yaml
index c5df90fb..fbb3ad54 100644
--- a/packages/system/lineage-controller-webhook/templates/service.yaml
+++ b/packages/system/lineage-controller-webhook/templates/service.yaml
@@ -5,7 +5,7 @@ metadata:
   labels:
     app: lineage-controller-webhook
 spec:
-  internalTrafficPolicy: Local
+  trafficDistribution: PreferClose
   type: ClusterIP
   ports:
     - port: 443
diff --git a/packages/system/lineage-controller-webhook/templates/daemonset.yaml b/packages/system/lineage-controller-webhook/templates/workload.yaml
similarity index 67%
rename from packages/system/lineage-controller-webhook/templates/daemonset.yaml
rename to packages/system/lineage-controller-webhook/templates/workload.yaml
index 860aee6e..097919c6 100644
--- a/packages/system/lineage-controller-webhook/templates/daemonset.yaml
+++ b/packages/system/lineage-controller-webhook/templates/workload.yaml
@@ -1,10 +1,11 @@
 apiVersion: apps/v1
-kind: DaemonSet
+kind: Deployment
 metadata:
   name: lineage-controller-webhook
   labels:
     app: lineage-controller-webhook
 spec:
+  replicas: {{ .Values.lineageControllerWebhook.replicas }}
   selector:
     matchLabels:
       app: lineage-controller-webhook
@@ -13,13 +14,25 @@ spec:
       labels:
         app: lineage-controller-webhook
     spec:
-      {{- with .Values.lineageControllerWebhook.nodeSelector }}
-      nodeSelector:
-        {{- toYaml . | nindent 8 }}
-      {{- end }}
+      serviceAccountName: lineage-controller-webhook
       tolerations:
       - operator: Exists
-      serviceAccountName: lineage-controller-webhook
+      affinity:
+        nodeAffinity:
+          preferredDuringSchedulingIgnoredDuringExecution:
+          - weight: 100
+            preference:
+              matchExpressions:
+              - key: node-role.kubernetes.io/control-plane
+                operator: Exists
+        podAntiAffinity:
+          preferredDuringSchedulingIgnoredDuringExecution:
+          - weight: 100
+            podAffinityTerm:
+              topologyKey: kubernetes.io/hostname
+              labelSelector:
+                matchLabels:
+                  app: lineage-controller-webhook
       containers:
       - name: lineage-controller-webhook
         image: "{{ .Values.lineageControllerWebhook.image }}"
diff --git a/packages/system/lineage-controller-webhook/tests/service_test.yaml b/packages/system/lineage-controller-webhook/tests/service_test.yaml
new file mode 100644
index 00000000..3cdbddd8
--- /dev/null
+++ b/packages/system/lineage-controller-webhook/tests/service_test.yaml
@@ -0,0 +1,30 @@
+suite: lineage-controller-webhook service
+templates:
+  - templates/service.yaml
+
+release:
+  name: lineage-controller-webhook
+  namespace: cozy-system
+
+tests:
+  - it: renders a ClusterIP service on 443 -> 9443 with PreferClose traffic distribution
+    asserts:
+      - isKind:
+          of: Service
+      - equal:
+          path: spec.type
+          value: ClusterIP
+      - equal:
+          path: spec.trafficDistribution
+          value: PreferClose
+      - notExists:
+          path: spec.internalTrafficPolicy
+      - equal:
+          path: spec.ports[0].port
+          value: 443
+      - equal:
+          path: spec.ports[0].targetPort
+          value: 9443
+      - equal:
+          path: spec.selector.app
+          value: lineage-controller-webhook
diff --git a/packages/system/lineage-controller-webhook/tests/workload_test.yaml b/packages/system/lineage-controller-webhook/tests/workload_test.yaml
new file mode 100644
index 00000000..bcb24014
--- /dev/null
+++ b/packages/system/lineage-controller-webhook/tests/workload_test.yaml
@@ -0,0 +1,99 @@
+suite: lineage-controller-webhook workload + PDB
+templates:
+  - templates/workload.yaml
+  - templates/pdb.yaml
+
+release:
+  name: lineage-controller-webhook
+  namespace: cozy-system
+
+tests:
+  # ---- Default Deployment shape ----------------------------------------------
+
+  - it: defaults render a 2-replica Deployment with soft control-plane nodeAffinity, soft podAntiAffinity, and Exists tolerations
+    template: templates/workload.yaml
+    asserts:
+      - isKind:
+          of: Deployment
+      - equal:
+          path: spec.replicas
+          value: 2
+      - equal:
+          path: spec.template.spec.tolerations
+          value:
+            - operator: Exists
+      - equal:
+          path: spec.template.spec.affinity.nodeAffinity.preferredDuringSchedulingIgnoredDuringExecution
+          value:
+            - weight: 100
+              preference:
+                matchExpressions:
+                  - key: node-role.kubernetes.io/control-plane
+                    operator: Exists
+      - notExists:
+          path: spec.template.spec.affinity.nodeAffinity.requiredDuringSchedulingIgnoredDuringExecution
+      - equal:
+          path: spec.template.spec.affinity.podAntiAffinity.preferredDuringSchedulingIgnoredDuringExecution[0].podAffinityTerm.topologyKey
+          value: kubernetes.io/hostname
+
+  - it: defaults inject no localK8sAPIEndpoint env vars (knob is opt-in)
+    template: templates/workload.yaml
+    asserts:
+      - notExists:
+          path: spec.template.spec.containers[0].env
+
+  # ---- PDB --------------------------------------------------------------------
+
+  - it: PDB renders unconditionally with maxUnavailable=1 at the default replica count
+    template: templates/pdb.yaml
+    asserts:
+      - isKind:
+          of: PodDisruptionBudget
+      - equal:
+          path: spec.maxUnavailable
+          value: 1
+      - notExists:
+          path: spec.minAvailable
+
+  - it: PDB still renders at replicas=1 (no-op, but consistent shape)
+    set:
+      lineageControllerWebhook.replicas: 1
+    template: templates/pdb.yaml
+    asserts:
+      - isKind:
+          of: PodDisruptionBudget
+      - equal:
+          path: spec.maxUnavailable
+          value: 1
+
+  # ---- Replica override -------------------------------------------------------
+
+  - it: replicas value drives spec.replicas
+    set:
+      lineageControllerWebhook.replicas: 5
+    template: templates/workload.yaml
+    asserts:
+      - equal:
+          path: spec.replicas
+          value: 5
+
+  # ---- Deprecated localK8sAPIEndpoint knob -----------------------------------
+
+  - it: localK8sAPIEndpoint.enabled=true injects KUBERNETES_SERVICE_HOST and PORT env vars
+    set:
+      lineageControllerWebhook.localK8sAPIEndpoint.enabled: true
+    template: templates/workload.yaml
+    asserts:
+      - contains:
+          path: spec.template.spec.containers[0].env
+          content:
+            name: KUBERNETES_SERVICE_HOST
+            valueFrom:
+              fieldRef:
+                apiVersion: v1
+                fieldPath: status.hostIP
+      - contains:
+          path: spec.template.spec.containers[0].env
+          content:
+            name: KUBERNETES_SERVICE_PORT
+            value: "6443"
diff --git a/packages/system/lineage-controller-webhook/values.yaml b/packages/system/lineage-controller-webhook/values.yaml
index 103dbf85..e56d332d 100644
--- a/packages/system/lineage-controller-webhook/values.yaml
+++ b/packages/system/lineage-controller-webhook/values.yaml
@@ -1,10 +1,13 @@
 lineageControllerWebhook:
-  image: ghcr.io/cozystack/cozystack/lineage-controller-webhook:v1.2.1@sha256:0fef727af2ea8d87503ba998d2ec03c12eb432d03b6642a555b81b95fbec4cb0
+  image: ghcr.io/cozystack/cozystack/lineage-controller-webhook:v1.3.0@sha256:e8984709686a5eaf19b89da378d7b8c688ea5607e0783a88d9c9e4ccfca96fb0
   debug: false
+  replicas: 2
+  # DEPRECATED. Injects KUBERNETES_SERVICE_HOST=status.hostIP and
+  # KUBERNETES_SERVICE_PORT=6443 so the webhook talks to the apiserver on its
+  # own node — only valid when the pod is actually scheduled on an apiserver-
+  # bearing node. Now that the chart's nodeAffinity to control-plane is soft
+  # (preferred, not required), the pod can land off-control-plane and crash-
+  # loop dialing a non-apiserver hostIP. Slated for removal once webhook
+  # performance work obviates the locality motivation. Leave disabled.
   localK8sAPIEndpoint:
-    enabled: true
-  # nodeSelector for the DaemonSet
-  # Talos uses empty value: "node-role.kubernetes.io/control-plane": ""
-  # Generic k8s (k3s, kubeadm) uses: "node-role.kubernetes.io/control-plane": "true"
-  nodeSelector:
-    node-role.kubernetes.io/control-plane: ""
+    enabled: false
diff --git a/packages/system/linstor-gui/Chart.yaml b/packages/system/linstor-gui/Chart.yaml
new file mode 100644
index 00000000..e361a819
--- /dev/null
+++ b/packages/system/linstor-gui/Chart.yaml
@@ -0,0 +1,3 @@
+apiVersion: v2
+name: cozy-linstor-gui
+version: 0.0.0 # Placeholder, the actual version will be automatically set during the build process
diff --git a/packages/system/linstor-gui/Makefile b/packages/system/linstor-gui/Makefile
new file mode 100644
index 00000000..b408491d
--- /dev/null
+++ b/packages/system/linstor-gui/Makefile
@@ -0,0 +1,29 @@
+export NAME=linstor-gui
+export NAMESPACE=cozy-linstor
+
+include ../../../hack/common-envs.mk
+include ../../../hack/package.mk
+
+LINSTOR_GUI_VERSION ?= 2.3.0
+
+.PHONY: image image-linstor-gui test
+
+image: image-linstor-gui
+
+image-linstor-gui:
+	docker buildx build images/linstor-gui \
+		--build-arg LINSTOR_GUI_VERSION=$(LINSTOR_GUI_VERSION) \
+		--tag $(REGISTRY)/linstor-gui:$(call settag,$(LINSTOR_GUI_VERSION)) \
+		--tag $(REGISTRY)/linstor-gui:$(call settag,$(LINSTOR_GUI_VERSION)-$(TAG)) \
+		--cache-from type=registry,ref=$(REGISTRY)/linstor-gui:latest \
+		--cache-to type=inline \
+		--metadata-file images/linstor-gui.json \
+		$(BUILDX_ARGS)
+	REPOSITORY="$(REGISTRY)/linstor-gui" \
+		yq -i '.image.repository = strenv(REPOSITORY)' values.yaml
+	TAG="$(call settag,$(LINSTOR_GUI_VERSION))@$$(yq e '."containerimage.digest"' images/linstor-gui.json -o json -r)" \
+		yq -i '.image.tag = strenv(TAG)' values.yaml
+	rm -f images/linstor-gui.json
+
+test:
+	helm unittest .
diff --git a/packages/system/linstor-gui/README.md b/packages/system/linstor-gui/README.md
new file mode 100644
index 00000000..8f8053ab
--- /dev/null
+++ b/packages/system/linstor-gui/README.md
@@ -0,0 +1,64 @@
+# linstor-gui
+
+Cozystack system package for [LINBIT/linstor-gui](https://github.com/LINBIT/linstor-gui)
+— a web UI for managing LINSTOR nodes, resources, volumes and snapshots.
+
+Installed alongside the `linstor` package in the `cozy-linstor` namespace. The UI
+proxies the LINSTOR controller REST API at `https://linstor-controller.cozy-linstor.svc:3371`
+using mTLS with the `linstor-client-tls` secret created by the `linstor` package.
+
+## Exposing the UI
+
+### Option 1 — Keycloak-protected Ingress (recommended)
+
+The chart ships an `oauth2-proxy` based gatekeeper plus a `KeycloakClient` CRD
+so the UI can be published on `linstor-gui.<root-host>` behind the cluster
+Keycloak realm. Access is restricted to members of the
+`cozystack-cluster-admin` Keycloak group — the same group that grants
+cluster-admin RBAC on the host cluster. Authenticating against the `cozy`
+realm alone is not sufficient; users outside that group receive a 403 from
+oauth2-proxy before any request reaches the UI or the LINSTOR controller.
+
+To turn it on, add `linstor-gui` to `publishing.exposedServices` in the core
+`cozystack` values (same list that controls `dashboard`). OIDC must be
+enabled (`authentication.oidc.enabled: true`) — if it is not, the Ingress and
+gatekeeper Deployment are deliberately **not** rendered, because the LINSTOR
+REST API surface must not be exposed unauthenticated.
+
+Once enabled, the UI is reachable at `https://linstor-gui.<root-host>` and
+authentication is delegated to Keycloak via the `linstor-gui` client
+(auto-provisioned through the `KeycloakClient` CRD; the client secret is
+persisted in the `linstor-gui-client` Secret in `cozy-linstor`).
+
+### Option 2 — Port-forward
+
+If you have not set up Keycloak or want ad-hoc access, use the `ClusterIP`
+Service:
+
+```bash
+kubectl -n cozy-linstor port-forward svc/linstor-gui 3373:80
+```
+
+then open <http://localhost:3373>.
+
+## Parameters
+
+### Image
+
+| Name               | Description                                                | Value                                     |
+| ------------------ | ---------------------------------------------------------- | ----------------------------------------- |
+| `image.repository` | LINSTOR GUI container image repository                     | `ghcr.io/cozystack/cozystack/linstor-gui` |
+| `image.tag`        | LINSTOR GUI container image tag (digest recommended)       | `2.3.0`                                   |
+
+### Deployment
+
+| Name       | Description                     | Value |
+| ---------- | ------------------------------- | ----- |
+| `replicas` | Number of linstor-gui replicas  | `1`   |
+
+### LINSTOR controller connection
+
+| Name                    | Description                                                                                                                          | Value                                                    |
+| ----------------------- | ------------------------------------------------------------------------------------------------------------------------------------ | -------------------------------------------------------- |
+| `linstor.endpoint`      | In-cluster URL of the LINSTOR controller REST API (HTTPS, mTLS)                                                                      | `https://linstor-controller.cozy-linstor.svc:3371`       |
+| `linstor.clientSecret`  | Kubernetes Secret with `tls.crt`, `tls.key`, `ca.crt` used as the mTLS client certificate against the LINSTOR controller. Created by the `linstor` package. | `linstor-client-tls`                                     |
diff --git a/packages/system/linstor-gui/images/linstor-gui/Dockerfile b/packages/system/linstor-gui/images/linstor-gui/Dockerfile
new file mode 100644
index 00000000..2082ef0d
--- /dev/null
+++ b/packages/system/linstor-gui/images/linstor-gui/Dockerfile
@@ -0,0 +1,19 @@
+# Upstream: https://github.com/LINBIT/linstor-gui (GPL-3.0)
+# Serves the pre-built LINSTOR GUI tarball published by LINBIT from a hardened
+# nginx-unprivileged image. nginx.conf is supplied by the chart via ConfigMap,
+# so the upstream docker-entrypoint.sh / nginx.conf.template is not used.
+FROM nginxinc/nginx-unprivileged:1.29-alpine
+
+ARG LINSTOR_GUI_VERSION
+USER root
+RUN apk add --no-cache curl tar && \
+    curl -fsSL "https://pkg.linbit.com/downloads/linstor/linstor-gui-${LINSTOR_GUI_VERSION}.tar.gz" -o /tmp/linstor-gui.tar.gz && \
+    mkdir -p /usr/share/nginx/html && \
+    tar -xzf /tmp/linstor-gui.tar.gz -C /usr/share/nginx/html --strip-components=2 "linstor-gui-${LINSTOR_GUI_VERSION}/dist" && \
+    rm -f /tmp/linstor-gui.tar.gz && \
+    apk del curl tar && \
+    chown -R 101:101 /usr/share/nginx/html
+
+USER 101
+EXPOSE 3373
+CMD ["nginx", "-g", "daemon off;"]
diff --git a/packages/system/linstor-gui/templates/_helpers.tpl b/packages/system/linstor-gui/templates/_helpers.tpl
new file mode 100644
index 00000000..6cad99ee
--- /dev/null
+++ b/packages/system/linstor-gui/templates/_helpers.tpl
@@ -0,0 +1,17 @@
+{{/*
+Common labels
+*/}}
+{{- define "linstor-gui.labels" -}}
+app.kubernetes.io/name: linstor-gui
+app.kubernetes.io/instance: {{ .Release.Name }}
+app.kubernetes.io/managed-by: {{ .Release.Service }}
+app.kubernetes.io/part-of: cozystack
+{{- end }}
+
+{{/*
+Selector labels
+*/}}
+{{- define "linstor-gui.selectorLabels" -}}
+app.kubernetes.io/name: linstor-gui
+app.kubernetes.io/instance: {{ .Release.Name }}
+{{- end }}
diff --git a/packages/system/linstor-gui/templates/configmap-nginx.yaml b/packages/system/linstor-gui/templates/configmap-nginx.yaml
new file mode 100644
index 00000000..86acd0f7
--- /dev/null
+++ b/packages/system/linstor-gui/templates/configmap-nginx.yaml
@@ -0,0 +1,113 @@
+{{/*
+Parse the LINSTOR endpoint so nginx can set `proxy_ssl_name` correctly.
+The endpoint must be an https:// URL — http:// would downgrade the
+mTLS-protected controller path.
+*/}}
+{{- if not (hasPrefix "https://" .Values.linstor.endpoint) -}}
+{{- fail "linstor.endpoint must start with https:// to enforce mTLS to the LINSTOR controller" -}}
+{{- end -}}
+{{- $endpoint := trimPrefix "https://" .Values.linstor.endpoint -}}
+{{- $endpointHost := (splitList ":" $endpoint) | first -}}
+apiVersion: v1
+kind: ConfigMap
+metadata:
+  name: linstor-gui-nginx
+  labels:
+    {{- include "linstor-gui.labels" . | nindent 4 }}
+data:
+  nginx.conf: |
+    worker_processes auto;
+    pid /tmp/nginx.pid;
+
+    events {
+      worker_connections 1024;
+    }
+
+    http {
+      include       /etc/nginx/mime.types;
+      default_type  application/octet-stream;
+      server_tokens off;
+
+      client_body_temp_path /tmp/client_body;
+      proxy_temp_path       /tmp/proxy;
+      fastcgi_temp_path     /tmp/fastcgi;
+      uwsgi_temp_path       /tmp/uwsgi;
+      scgi_temp_path        /tmp/scgi;
+
+      sendfile        on;
+      keepalive_timeout  65;
+
+      server {
+        listen 3373;
+        server_name _;
+
+        root /usr/share/nginx/html;
+        index index.html;
+
+        # Shared proxy + mTLS settings for the LINSTOR controller upstream
+        proxy_set_header Host $host;
+        proxy_set_header X-Real-IP $remote_addr;
+        proxy_ssl_certificate         /etc/linstor/client/tls.crt;
+        proxy_ssl_certificate_key     /etc/linstor/client/tls.key;
+        proxy_ssl_trusted_certificate /etc/linstor/client/ca.crt;
+        proxy_ssl_verify       on;
+        proxy_ssl_verify_depth 2;
+        proxy_ssl_server_name  on;
+        proxy_ssl_name         {{ $endpointHost }};
+
+        # Static UI assets
+        location / {
+          try_files $uri $uri/ /index.html;
+        }
+
+        # Lock the linstor-gui in-app authentication panel.
+        #
+        # Unlike most LINSTOR features, the GUI's "authenticationEnabled"
+        # toggle and user list are NOT controller-side properties — they
+        # are persisted as ordinary KeyValueStore instances named
+        # `__gui__settings` and `__gui__users`. The SPA reads them on
+        # every page load and, if `authenticationEnabled=true`, gates the
+        # whole UI behind its own login screen against a self-managed
+        # admin user.
+        #
+        # Cozystack already enforces authentication at the Ingress
+        # (oauth2-proxy + Keycloak), and the in-app login is a foot-gun:
+        # if a user enables it and forgets/misconfigures the password, the
+        # only recovery is shelling into the linstor-controller pod and
+        # running `linstor key-value-store modify __gui__settings
+        # authenticationEnabled false`.
+        #
+        # We allow GETs (so the SPA can read state and render normally
+        # with auth=off) but reject every mutating method on those two
+        # KeyValueStore instances. Any other key-value-store entry the
+        # GUI uses (e.g. __gui__mode) keeps working as expected.
+        location ~ ^/v1/key-value-store/__gui__(settings|users)(/|$) {
+          default_type application/json;
+          # `if ... return` is the one documented-safe use of `if` in
+          # nginx (see "If is Evil"), so it's fine here. We allow GET/HEAD
+          # for the SPA's read path and reject every mutating method.
+          if ($request_method !~ ^(GET|HEAD)$) {
+            return 403 '{"ret_code":-1,"message":"linstor-gui in-app authentication is disabled by cozystack: authentication is enforced at the Ingress via Keycloak."}';
+          }
+          proxy_pass {{ .Values.linstor.endpoint }};
+        }
+
+        # Proxy LINSTOR REST API over mTLS to the controller
+        location /v1 {
+          proxy_pass {{ .Values.linstor.endpoint }};
+        }
+
+        location /metrics {
+          proxy_pass {{ .Values.linstor.endpoint }};
+        }
+
+        location = /healthz {
+          access_log off;
+          # default_type must come BEFORE `return` — `add_header` after a
+          # `return` is silently dropped by nginx because the response is
+          # already finalized.
+          default_type text/plain;
+          return 200 'ok';
+        }
+      }
+    }
diff --git a/packages/system/linstor-gui/templates/deployment.yaml b/packages/system/linstor-gui/templates/deployment.yaml
new file mode 100644
index 00000000..92e2f7a6
--- /dev/null
+++ b/packages/system/linstor-gui/templates/deployment.yaml
@@ -0,0 +1,100 @@
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: linstor-gui
+  labels:
+    {{- include "linstor-gui.labels" . | nindent 4 }}
+  annotations:
+    reloader.stakater.com/auto: "true"
+spec:
+  replicas: {{ .Values.replicas }}
+  selector:
+    matchLabels:
+      {{- include "linstor-gui.selectorLabels" . | nindent 6 }}
+  strategy:
+    type: RollingUpdate
+    rollingUpdate:
+      maxUnavailable: 0
+      maxSurge: 1
+  template:
+    metadata:
+      labels:
+        {{- include "linstor-gui.labels" . | nindent 8 }}
+    spec:
+      serviceAccountName: linstor-gui
+      automountServiceAccountToken: false
+      securityContext:
+        runAsNonRoot: true
+        seccompProfile:
+          type: RuntimeDefault
+      affinity:
+        podAntiAffinity:
+          preferredDuringSchedulingIgnoredDuringExecution:
+            - weight: 100
+              podAffinityTerm:
+                labelSelector:
+                  matchLabels:
+                    {{- include "linstor-gui.selectorLabels" . | nindent 20 }}
+                topologyKey: kubernetes.io/hostname
+      containers:
+        - name: linstor-gui
+          image: {{ .Values.image.repository }}:{{ .Values.image.tag }}
+          imagePullPolicy: IfNotPresent
+          ports:
+            - name: http
+              containerPort: 3373
+              protocol: TCP
+          resources:
+            requests:
+              cpu: 50m
+              memory: 64Mi
+            limits:
+              cpu: 200m
+              memory: 128Mi
+          livenessProbe:
+            httpGet:
+              path: /healthz
+              port: http
+            initialDelaySeconds: 5
+            periodSeconds: 20
+          readinessProbe:
+            httpGet:
+              path: /healthz
+              port: http
+            initialDelaySeconds: 2
+            periodSeconds: 10
+          securityContext:
+            allowPrivilegeEscalation: false
+            readOnlyRootFilesystem: true
+            runAsNonRoot: true
+            runAsUser: 101
+            runAsGroup: 101
+            capabilities:
+              drop:
+                - ALL
+          volumeMounts:
+            - name: nginx-config
+              mountPath: /etc/nginx/nginx.conf
+              subPath: nginx.conf
+              readOnly: true
+            - name: linstor-client-tls
+              mountPath: /etc/linstor/client
+              readOnly: true
+            - name: tmp
+              mountPath: /tmp
+            - name: nginx-cache
+              mountPath: /var/cache/nginx
+      volumes:
+        - name: nginx-config
+          configMap:
+            name: linstor-gui-nginx
+        - name: linstor-client-tls
+          secret:
+            secretName: {{ .Values.linstor.clientSecret }}
+            # 0444 so nginx (UID 101) can read the mTLS client keypair;
+            # the secret volume is pod-local, not a broader disclosure.
+            defaultMode: 0444
+        - name: tmp
+          emptyDir: {}
+        - name: nginx-cache
+          emptyDir: {}
diff --git a/packages/system/linstor-gui/templates/gatekeeper-sa.yaml b/packages/system/linstor-gui/templates/gatekeeper-sa.yaml
new file mode 100644
index 00000000..d89f68dc
--- /dev/null
+++ b/packages/system/linstor-gui/templates/gatekeeper-sa.yaml
@@ -0,0 +1,13 @@
+{{- $oidcEnabled := index .Values._cluster "oidc-enabled" }}
+{{- if eq $oidcEnabled "true" }}
+apiVersion: v1
+kind: ServiceAccount
+metadata:
+  name: linstor-gui-gatekeeper
+  labels:
+    app.kubernetes.io/instance: linstor-gui
+    app.kubernetes.io/name: gatekeeper
+    app.kubernetes.io/managed-by: {{ .Release.Service }}
+    app.kubernetes.io/part-of: cozystack
+automountServiceAccountToken: false
+{{- end }}
diff --git a/packages/system/linstor-gui/templates/gatekeeper-svc.yaml b/packages/system/linstor-gui/templates/gatekeeper-svc.yaml
new file mode 100644
index 00000000..7d50219e
--- /dev/null
+++ b/packages/system/linstor-gui/templates/gatekeeper-svc.yaml
@@ -0,0 +1,22 @@
+{{- $oidcEnabled := index .Values._cluster "oidc-enabled" }}
+{{- if eq $oidcEnabled "true" }}
+apiVersion: v1
+kind: Service
+metadata:
+  name: linstor-gui-gatekeeper
+  labels:
+    app.kubernetes.io/instance: linstor-gui
+    app.kubernetes.io/name: gatekeeper
+    app.kubernetes.io/managed-by: {{ .Release.Service }}
+    app.kubernetes.io/part-of: cozystack
+spec:
+  type: ClusterIP
+  ports:
+    - name: ingress
+      port: 8000
+      protocol: TCP
+      targetPort: 8000
+  selector:
+    app.kubernetes.io/instance: linstor-gui
+    app.kubernetes.io/name: gatekeeper
+{{- end }}
diff --git a/packages/system/linstor-gui/templates/gatekeeper.yaml b/packages/system/linstor-gui/templates/gatekeeper.yaml
new file mode 100644
index 00000000..81451f41
--- /dev/null
+++ b/packages/system/linstor-gui/templates/gatekeeper.yaml
@@ -0,0 +1,148 @@
+{{- $host := index .Values._cluster "root-host" }}
+{{- $oidcEnabled := index .Values._cluster "oidc-enabled" }}
+{{- $oidcInsecureSkipVerify := index .Values._cluster "oidc-insecure-skip-verify" }}
+{{- $keycloakInternalUrl := index .Values._cluster "keycloak-internal-url" | default "" }}
+
+{{- /*
+Gatekeeper (oauth2-proxy) fronts the linstor-gui Service and handles the
+OIDC flow against the cluster Keycloak realm. The Ingress below points at
+this Service — not at linstor-gui directly — so every external request
+authenticates before hitting nginx/mTLS→controller.
+
+Only rendered when oidc-enabled=true. If the cluster has OIDC disabled we
+deliberately do not ship an unauthenticated token-proxy fallback (unlike
+dashboard): linstor-gui surfaces raw storage state and should not be
+reachable via Ingress without Keycloak login.
+
+Access is further restricted to the cozystack-cluster-admin group
+(--allowed-group below). The proxy authenticates with a static mTLS client
+cert and LINSTOR itself has no per-user RBAC, so group membership is the
+only boundary between a realm user and raw storage state.
+*/ -}}
+{{- if eq $oidcEnabled "true" }}
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: linstor-gui-gatekeeper
+  labels:
+    app.kubernetes.io/instance: linstor-gui
+    app.kubernetes.io/name: gatekeeper
+    app.kubernetes.io/managed-by: {{ .Release.Service }}
+    app.kubernetes.io/part-of: cozystack
+  annotations:
+    reloader.stakater.com/auto: "true"
+spec:
+  replicas: 1
+  selector:
+    matchLabels:
+      app.kubernetes.io/instance: linstor-gui
+      app.kubernetes.io/name: gatekeeper
+  strategy:
+    type: RollingUpdate
+    rollingUpdate:
+      maxSurge: 1
+      maxUnavailable: 0
+  template:
+    metadata:
+      labels:
+        app.kubernetes.io/instance: linstor-gui
+        app.kubernetes.io/name: gatekeeper
+    spec:
+      serviceAccountName: linstor-gui-gatekeeper
+      automountServiceAccountToken: false
+      affinity:
+        podAntiAffinity:
+          preferredDuringSchedulingIgnoredDuringExecution:
+            - weight: 100
+              podAffinityTerm:
+                labelSelector:
+                  matchLabels:
+                    app.kubernetes.io/instance: linstor-gui
+                    app.kubernetes.io/name: gatekeeper
+                topologyKey: kubernetes.io/hostname
+      containers:
+        - name: auth-proxy
+          image: quay.io/oauth2-proxy/oauth2-proxy:v7.12.0
+          imagePullPolicy: IfNotPresent
+          args:
+            - --provider=oidc
+            - --upstream=http://linstor-gui.{{ .Release.Namespace }}.svc:80
+            - --http-address=0.0.0.0:8000
+            - --redirect-url=https://linstor-gui.{{ $host }}/oauth2/callback
+            - --oidc-issuer-url=https://keycloak.{{ $host }}/realms/cozy
+            {{- if $keycloakInternalUrl }}
+            - --skip-oidc-discovery
+            - --login-url=https://keycloak.{{ $host }}/realms/cozy/protocol/openid-connect/auth
+            - --redeem-url={{ $keycloakInternalUrl }}/protocol/openid-connect/token
+            - --oidc-jwks-url={{ $keycloakInternalUrl }}/protocol/openid-connect/certs
+            - --validate-url={{ $keycloakInternalUrl }}/protocol/openid-connect/userinfo
+            - --backend-logout-url={{ $keycloakInternalUrl }}/protocol/openid-connect/logout?id_token_hint={id_token}
+            {{- else }}
+            - --backend-logout-url=https://keycloak.{{ $host }}/realms/cozy/protocol/openid-connect/logout?id_token_hint={id_token}
+            {{- end }}
+            - --whitelist-domain=keycloak.{{ $host }}
+            - --email-domain=*
+            - --pass-access-token=true
+            - --pass-authorization-header=true
+            - --cookie-refresh=3m
+            - --cookie-name=kc-access
+            - --cookie-secure=true
+            - --cookie-secret=$(OAUTH2_PROXY_COOKIE_SECRET)
+            - --skip-provider-button
+            - --scope=openid email profile groups offline_access
+            - --allowed-group=cozystack-cluster-admin
+            {{- if eq $oidcInsecureSkipVerify "true" }}
+            - --ssl-insecure-skip-verify=true
+            {{- end }}
+          env:
+            - name: OAUTH2_PROXY_CLIENT_ID
+              value: linstor-gui
+            - name: OAUTH2_PROXY_CLIENT_SECRET
+              valueFrom:
+                secretKeyRef:
+                  name: linstor-gui-client
+                  key: client-secret-key
+            - name: OAUTH2_PROXY_COOKIE_SECRET
+              valueFrom:
+                secretKeyRef:
+                  name: linstor-gui-auth-config
+                  key: cookieSecret
+          ports:
+            - name: proxy
+              containerPort: 8000
+              protocol: TCP
+          livenessProbe:
+            httpGet:
+              path: /ping
+              port: proxy
+            initialDelaySeconds: 10
+            periodSeconds: 15
+            timeoutSeconds: 2
+            failureThreshold: 3
+          readinessProbe:
+            httpGet:
+              path: /ping
+              port: proxy
+            initialDelaySeconds: 5
+            periodSeconds: 10
+            timeoutSeconds: 2
+            failureThreshold: 3
+          resources:
+            requests:
+              cpu: 50m
+              memory: 64Mi
+            limits:
+              cpu: 200m
+              memory: 256Mi
+          securityContext:
+            runAsNonRoot: true
+            runAsUser: 1001
+            runAsGroup: 1001
+            allowPrivilegeEscalation: false
+            readOnlyRootFilesystem: true
+            capabilities:
+              drop:
+                - ALL
+            seccompProfile:
+              type: RuntimeDefault
+{{- end }}
diff --git a/packages/system/linstor-gui/templates/ingress.yaml b/packages/system/linstor-gui/templates/ingress.yaml
new file mode 100644
index 00000000..c50b0f86
--- /dev/null
+++ b/packages/system/linstor-gui/templates/ingress.yaml
@@ -0,0 +1,54 @@
+{{- $solver := (index .Values._cluster "solver") | default "http01" }}
+{{- $clusterIssuer := (index .Values._cluster "issuer-name") | default "letsencrypt-prod" }}
+{{- $host := index .Values._cluster "root-host" }}
+{{- $oidcEnabled := index .Values._cluster "oidc-enabled" }}
+{{- $exposeServices := splitList "," ((index .Values._cluster "expose-services") | default "") }}
+{{- $exposeIngress := (index .Values._cluster "expose-ingress") | default "tenant-root" }}
+
+{{- /*
+Only publish an Ingress when the operator has opted this service in AND
+OIDC is enabled cluster-wide. If OIDC is off we deliberately do not
+expose linstor-gui: without Keycloak there is no authentication layer in
+front of the LINSTOR REST API proxy. Users can still reach the UI via
+`kubectl port-forward` to the ClusterIP service.
+*/ -}}
+{{- if and (has "linstor-gui" $exposeServices) (eq $oidcEnabled "true") }}
+apiVersion: networking.k8s.io/v1
+kind: Ingress
+metadata:
+  name: linstor-gui-ingress
+  labels:
+    {{- include "linstor-gui.labels" . | nindent 4 }}
+  annotations:
+    cert-manager.io/cluster-issuer: {{ $clusterIssuer }}
+    {{- if eq $solver "http01" }}
+    acme.cert-manager.io/http01-ingress-ingressclassname: {{ $exposeIngress }}
+    {{- end }}
+    nginx.ingress.kubernetes.io/proxy-body-size: 100m
+    # Keycloak access+refresh+id tokens make the oauth2-proxy session
+    # cookie ~8–10 KB (split across multiple Set-Cookie headers), which
+    # overflows the default proxy_buffer_size and causes "upstream sent
+    # too big header" -> HTTP 502 on /oauth2/callback. Same numbers as
+    # dashboard ingress.
+    nginx.ingress.kubernetes.io/proxy-buffer-size: 100m
+    nginx.ingress.kubernetes.io/proxy-buffers-number: "4"
+    nginx.ingress.kubernetes.io/proxy-read-timeout: "3600"
+    nginx.ingress.kubernetes.io/proxy-send-timeout: "3600"
+spec:
+  ingressClassName: {{ $exposeIngress }}
+  rules:
+    - host: linstor-gui.{{ $host }}
+      http:
+        paths:
+          - path: /
+            pathType: Prefix
+            backend:
+              service:
+                name: linstor-gui-gatekeeper
+                port:
+                  number: 8000
+  tls:
+    - hosts:
+        - linstor-gui.{{ $host }}
+      secretName: linstor-gui-ingress-tls
+{{- end }}
diff --git a/packages/system/linstor-gui/templates/keycloakclient.yaml b/packages/system/linstor-gui/templates/keycloakclient.yaml
new file mode 100644
index 00000000..e5dab87b
--- /dev/null
+++ b/packages/system/linstor-gui/templates/keycloakclient.yaml
@@ -0,0 +1,72 @@
+{{- $host := index .Values._cluster "root-host" }}
+
+{{- /*
+Persist client + cookie secrets across upgrades: if the Secrets already
+exist, reuse their values; otherwise generate new random ones. The
+KeycloakClient CRD references the same client secret so Keycloak stays
+in sync with what oauth2-proxy reads.
+*/ -}}
+{{- $existingClientSecret := lookup "v1" "Secret" .Release.Namespace "linstor-gui-client" }}
+{{- $clientSecret := "" }}
+{{- if $existingClientSecret }}
+  {{- $clientSecret = index $existingClientSecret.data "client-secret-key" | b64dec }}
+{{- else }}
+  {{- $clientSecret = randAlphaNum 32 }}
+{{- end }}
+
+{{- $existingAuthConfig := lookup "v1" "Secret" .Release.Namespace "linstor-gui-auth-config" }}
+{{- $cookieSecret := "" }}
+{{- if $existingAuthConfig }}
+  {{- $cookieSecret = index $existingAuthConfig.data "cookieSecret" | b64dec }}
+{{- else }}
+  {{- $cookieSecret = randAlphaNum 16 }}
+{{- end }}
+---
+apiVersion: v1
+kind: Secret
+metadata:
+  name: linstor-gui-client
+  labels:
+    {{- include "linstor-gui.labels" . | nindent 4 }}
+type: Opaque
+data:
+  client-secret-key: {{ $clientSecret | b64enc }}
+---
+apiVersion: v1
+kind: Secret
+metadata:
+  name: linstor-gui-auth-config
+  labels:
+    {{- include "linstor-gui.labels" . | nindent 4 }}
+type: Opaque
+data:
+  cookieSecret: {{ $cookieSecret | b64enc }}
+{{- if .Capabilities.APIVersions.Has "v1.edp.epam.com/v1" }}
+---
+apiVersion: v1.edp.epam.com/v1
+kind: KeycloakClient
+metadata:
+  name: linstor-gui-client
+  annotations:
+    # Re-reconcile the Keycloak client if the random secret regenerates.
+    secret-hash: {{ $clientSecret | sha256sum }}
+spec:
+  realmRef:
+    name: keycloakrealm-cozy
+    kind: ClusterKeycloakRealm
+  secret: $linstor-gui-client:client-secret-key
+  advancedProtocolMappers: true
+  name: linstor-gui
+  clientId: linstor-gui
+  directAccess: true
+  public: false
+  webUrl: "https://linstor-gui.{{ $host }}"
+  defaultClientScopes:
+    - groups
+  optionalClientScopes:
+    - offline_access
+  attributes:
+    post.logout.redirect.uris: "+"
+  redirectUris:
+    - "https://linstor-gui.{{ $host }}/oauth2/callback/*"
+{{- end }}
diff --git a/packages/system/linstor-gui/templates/service.yaml b/packages/system/linstor-gui/templates/service.yaml
new file mode 100644
index 00000000..66ae2b2d
--- /dev/null
+++ b/packages/system/linstor-gui/templates/service.yaml
@@ -0,0 +1,15 @@
+apiVersion: v1
+kind: Service
+metadata:
+  name: linstor-gui
+  labels:
+    {{- include "linstor-gui.labels" . | nindent 4 }}
+spec:
+  type: ClusterIP
+  ports:
+    - name: http
+      port: 80
+      targetPort: http
+      protocol: TCP
+  selector:
+    {{- include "linstor-gui.selectorLabels" . | nindent 4 }}
diff --git a/packages/system/linstor-gui/templates/serviceaccount.yaml b/packages/system/linstor-gui/templates/serviceaccount.yaml
new file mode 100644
index 00000000..8e472671
--- /dev/null
+++ b/packages/system/linstor-gui/templates/serviceaccount.yaml
@@ -0,0 +1,7 @@
+apiVersion: v1
+kind: ServiceAccount
+metadata:
+  name: linstor-gui
+  labels:
+    {{- include "linstor-gui.labels" . | nindent 4 }}
+automountServiceAccountToken: false
diff --git a/packages/system/linstor-gui/tests/deployment_test.yaml b/packages/system/linstor-gui/tests/deployment_test.yaml
new file mode 100644
index 00000000..50711eb3
--- /dev/null
+++ b/packages/system/linstor-gui/tests/deployment_test.yaml
@@ -0,0 +1,95 @@
+suite: linstor-gui deployment
+templates:
+  - templates/deployment.yaml
+  - templates/service.yaml
+  - templates/configmap-nginx.yaml
+
+tests:
+  - it: renders a ClusterIP service on port 80 -> http
+    template: templates/service.yaml
+    release:
+      name: linstor-gui
+      namespace: cozy-linstor
+    asserts:
+      - isKind:
+          of: Service
+      - equal:
+          path: spec.type
+          value: ClusterIP
+      - equal:
+          path: spec.ports[0].port
+          value: 80
+      - equal:
+          path: spec.ports[0].targetPort
+          value: http
+
+  - it: mounts the LINSTOR client TLS secret into the pod
+    template: templates/deployment.yaml
+    release:
+      name: linstor-gui
+      namespace: cozy-linstor
+    asserts:
+      - isKind:
+          of: Deployment
+      - contains:
+          path: spec.template.spec.volumes
+          content:
+            name: linstor-client-tls
+            secret:
+              secretName: linstor-client-tls
+              defaultMode: 0444
+      - contains:
+          path: spec.template.spec.containers[0].volumeMounts
+          content:
+            name: linstor-client-tls
+            mountPath: /etc/linstor/client
+            readOnly: true
+
+  - it: runs as a non-root user with a read-only root filesystem
+    template: templates/deployment.yaml
+    release:
+      name: linstor-gui
+      namespace: cozy-linstor
+    asserts:
+      - equal:
+          path: spec.template.spec.containers[0].securityContext.runAsNonRoot
+          value: true
+      - equal:
+          path: spec.template.spec.containers[0].securityContext.readOnlyRootFilesystem
+          value: true
+      - equal:
+          path: spec.template.spec.containers[0].securityContext.allowPrivilegeEscalation
+          value: false
+
+  - it: proxies /v1 to the configured LINSTOR controller with mTLS
+    template: templates/configmap-nginx.yaml
+    release:
+      name: linstor-gui
+      namespace: cozy-linstor
+    asserts:
+      - isKind:
+          of: ConfigMap
+      - matchRegex:
+          path: data["nginx.conf"]
+          pattern: "location /v1"
+      - matchRegex:
+          path: data["nginx.conf"]
+          pattern: "proxy_pass https://linstor-controller.cozy-linstor.svc:3371"
+      - matchRegex:
+          path: data["nginx.conf"]
+          pattern: "proxy_ssl_certificate\\s+/etc/linstor/client/tls.crt"
+
+  - it: overrides linstor endpoint and client secret
+    template: templates/configmap-nginx.yaml
+    release:
+      name: linstor-gui
+      namespace: cozy-linstor
+    set:
+      linstor.endpoint: https://my-controller.example.svc:3371
+    asserts:
+      - matchRegex:
+          path: data["nginx.conf"]
+          pattern: "proxy_pass https://my-controller.example.svc:3371"
+      - matchRegex:
+          path: data["nginx.conf"]
+          pattern: "proxy_ssl_name         my-controller.example.svc"
diff --git a/packages/system/linstor-gui/tests/ingress_auth_test.yaml b/packages/system/linstor-gui/tests/ingress_auth_test.yaml
new file mode 100644
index 00000000..2f0b1d73
--- /dev/null
+++ b/packages/system/linstor-gui/tests/ingress_auth_test.yaml
@@ -0,0 +1,177 @@
+suite: linstor-gui ingress + keycloak auth
+templates:
+  - templates/ingress.yaml
+  - templates/gatekeeper.yaml
+  - templates/gatekeeper-svc.yaml
+  - templates/gatekeeper-sa.yaml
+  - templates/keycloakclient.yaml
+
+release:
+  name: linstor-gui
+  namespace: cozy-linstor
+
+tests:
+  - it: does not render an Ingress when OIDC is disabled cluster-wide
+    template: templates/ingress.yaml
+    set:
+      _cluster:
+        root-host: example.org
+        oidc-enabled: "false"
+        expose-services: "linstor-gui"
+        expose-ingress: "tenant-root"
+        issuer-name: "letsencrypt-prod"
+        solver: "http01"
+    asserts:
+      - hasDocuments:
+          count: 0
+
+  - it: does not render an Ingress when linstor-gui is not in expose-services
+    template: templates/ingress.yaml
+    set:
+      _cluster:
+        root-host: example.org
+        oidc-enabled: "true"
+        expose-services: "dashboard,api"
+        expose-ingress: "tenant-root"
+        issuer-name: "letsencrypt-prod"
+        solver: "http01"
+    asserts:
+      - hasDocuments:
+          count: 0
+
+  - it: renders an Ingress to the gatekeeper service when exposed and OIDC enabled
+    template: templates/ingress.yaml
+    set:
+      _cluster:
+        root-host: dev10.example.org
+        oidc-enabled: "true"
+        expose-services: "dashboard,linstor-gui"
+        expose-ingress: "tenant-root"
+        issuer-name: "letsencrypt-prod"
+        solver: "http01"
+    asserts:
+      - isKind:
+          of: Ingress
+      - equal:
+          path: spec.ingressClassName
+          value: tenant-root
+      - equal:
+          path: spec.rules[0].host
+          value: linstor-gui.dev10.example.org
+      - equal:
+          path: spec.rules[0].http.paths[0].backend.service.name
+          value: linstor-gui-gatekeeper
+      - equal:
+          path: spec.rules[0].http.paths[0].backend.service.port.number
+          value: 8000
+      - equal:
+          path: metadata.annotations["cert-manager.io/cluster-issuer"]
+          value: letsencrypt-prod
+
+  - it: does not render the gatekeeper Deployment when OIDC is disabled
+    template: templates/gatekeeper.yaml
+    set:
+      _cluster:
+        root-host: example.org
+        oidc-enabled: "false"
+        expose-services: "linstor-gui"
+    asserts:
+      - hasDocuments:
+          count: 0
+
+  - it: renders an oauth2-proxy Deployment pointing at the cluster Keycloak realm
+    template: templates/gatekeeper.yaml
+    set:
+      _cluster:
+        root-host: dev10.example.org
+        oidc-enabled: "true"
+        oidc-insecure-skip-verify: "false"
+        keycloak-internal-url: ""
+        expose-services: "linstor-gui"
+    asserts:
+      - isKind:
+          of: Deployment
+      - equal:
+          path: metadata.name
+          value: linstor-gui-gatekeeper
+      - matchRegex:
+          path: spec.template.spec.containers[0].image
+          pattern: "^quay\\.io/oauth2-proxy/oauth2-proxy:"
+      - contains:
+          path: spec.template.spec.containers[0].args
+          content: --provider=oidc
+      - contains:
+          path: spec.template.spec.containers[0].args
+          content: --upstream=http://linstor-gui.cozy-linstor.svc:80
+      - contains:
+          path: spec.template.spec.containers[0].args
+          content: --redirect-url=https://linstor-gui.dev10.example.org/oauth2/callback
+      - contains:
+          path: spec.template.spec.containers[0].args
+          content: --oidc-issuer-url=https://keycloak.dev10.example.org/realms/cozy
+      - contains:
+          path: spec.template.spec.containers[0].args
+          content: --scope=openid email profile groups offline_access
+      - contains:
+          path: spec.template.spec.containers[0].args
+          content: --allowed-group=cozystack-cluster-admin
+      - contains:
+          path: spec.template.spec.containers[0].env
+          content:
+            name: OAUTH2_PROXY_CLIENT_ID
+            value: linstor-gui
+      - equal:
+          path: spec.template.spec.containers[0].securityContext.readOnlyRootFilesystem
+          value: true
+
+  - it: creates client + cookie Secrets and a KeycloakClient CRD
+    template: templates/keycloakclient.yaml
+    capabilities:
+      apiVersions:
+        - v1.edp.epam.com/v1
+    set:
+      _cluster:
+        root-host: dev10.example.org
+    asserts:
+      - hasDocuments:
+          count: 3
+      - documentIndex: 0
+        isKind:
+          of: Secret
+      - documentIndex: 0
+        equal:
+          path: metadata.name
+          value: linstor-gui-client
+      - documentIndex: 1
+        isKind:
+          of: Secret
+      - documentIndex: 1
+        equal:
+          path: metadata.name
+          value: linstor-gui-auth-config
+      - documentIndex: 2
+        isKind:
+          of: KeycloakClient
+      - documentIndex: 2
+        equal:
+          path: spec.clientId
+          value: linstor-gui
+      - documentIndex: 2
+        contains:
+          path: spec.redirectUris
+          content: "https://linstor-gui.dev10.example.org/oauth2/callback/*"
+
+  - it: skips the KeycloakClient CRD when the API is absent, still creates Secrets
+    template: templates/keycloakclient.yaml
+    set:
+      _cluster:
+        root-host: example.org
+    asserts:
+      - hasDocuments:
+          count: 2
+      - documentIndex: 0
+        isKind:
+          of: Secret
+      - documentIndex: 1
+        isKind:
+          of: Secret
diff --git a/packages/system/linstor-gui/values.yaml b/packages/system/linstor-gui/values.yaml
new file mode 100644
index 00000000..e3c49aea
--- /dev/null
+++ b/packages/system/linstor-gui/values.yaml
@@ -0,0 +1,17 @@
+## @section Image
+## @param image.repository LINSTOR GUI container image repository
+## @param image.tag LINSTOR GUI container image tag (digest recommended)
+image:
+  repository: ghcr.io/cozystack/cozystack/linstor-gui
+  tag: 2.3.0
+
+## @section Deployment
+## @param replicas Number of linstor-gui replicas
+replicas: 1
+
+## @section LINSTOR controller connection
+## @param linstor.endpoint In-cluster URL of the LINSTOR controller REST API (HTTPS, mTLS)
+## @param linstor.clientSecret Kubernetes Secret with `tls.crt`, `tls.key`, `ca.crt` used as the mTLS client certificate against the LINSTOR controller. Created by the `linstor` package.
+linstor:
+  endpoint: "https://linstor-controller.cozy-linstor.svc:3371"
+  clientSecret: "linstor-client-tls"
diff --git a/packages/system/linstor-scheduler/Makefile b/packages/system/linstor-scheduler/Makefile
index f9986c9a..d83a2929 100644
--- a/packages/system/linstor-scheduler/Makefile
+++ b/packages/system/linstor-scheduler/Makefile
@@ -3,9 +3,13 @@ export NAMESPACE=cozy-linstor
 
 include ../../../hack/package.mk
 
+test:
+	helm unittest .
+
 update:
 	rm -rf charts
 	helm repo add piraeus-charts https://piraeus.io/helm-charts/
 	helm repo update piraeus-charts
 	helm pull piraeus-charts/linstor-scheduler --untar --untardir charts
 	patch --no-backup-if-mismatch -p4 < patches/disable-ca-key-rotation.patch
+	patch --no-backup-if-mismatch -p4 < patches/add-scheduler-component-label.patch
diff --git a/packages/system/linstor-scheduler/charts/linstor-scheduler/templates/deployment.yaml b/packages/system/linstor-scheduler/charts/linstor-scheduler/templates/deployment.yaml
index ea43cb83..289021fd 100644
--- a/packages/system/linstor-scheduler/charts/linstor-scheduler/templates/deployment.yaml
+++ b/packages/system/linstor-scheduler/charts/linstor-scheduler/templates/deployment.yaml
@@ -20,6 +20,7 @@ spec:
       {{- end }}
       labels:
         {{- include "linstor-scheduler.selectorLabels" . | nindent 8 }}
+        app.kubernetes.io/component: scheduler
     spec:
       {{- with .Values.imagePullSecrets }}
       imagePullSecrets:
diff --git a/packages/system/linstor-scheduler/patches/add-scheduler-component-label.patch b/packages/system/linstor-scheduler/patches/add-scheduler-component-label.patch
new file mode 100644
index 00000000..01eed9a8
--- /dev/null
+++ b/packages/system/linstor-scheduler/patches/add-scheduler-component-label.patch
@@ -0,0 +1,9 @@
+diff --git a/packages/system/linstor-scheduler/charts/linstor-scheduler/templates/deployment.yaml b/packages/system/linstor-scheduler/charts/linstor-scheduler/templates/deployment.yaml
+--- a/packages/system/linstor-scheduler/charts/linstor-scheduler/templates/deployment.yaml
++++ b/packages/system/linstor-scheduler/charts/linstor-scheduler/templates/deployment.yaml
+@@ -21,4 +21,5 @@
+       labels:
+         {{- include "linstor-scheduler.selectorLabels" . | nindent 8 }}
++        app.kubernetes.io/component: scheduler
+     spec:
+       {{- with .Values.imagePullSecrets }}
diff --git a/packages/system/linstor-scheduler/templates/extender-service.yaml b/packages/system/linstor-scheduler/templates/extender-service.yaml
new file mode 100644
index 00000000..e8148b43
--- /dev/null
+++ b/packages/system/linstor-scheduler/templates/extender-service.yaml
@@ -0,0 +1,26 @@
+{{/*
+  Exposes the linstor-scheduler-extender sidecar (port 8099) as a
+  cluster-internal Service so that other schedulers (e.g. cozystack-scheduler)
+  can use it as a scheduler extender.
+
+  The selector labels match the subchart's deployment pods. They are
+  hardcoded here because the subchart helpers expect a subchart rendering
+  context that is unavailable in the wrapper chart.
+*/}}
+apiVersion: v1
+kind: Service
+metadata:
+  name: linstor-scheduler-extender
+  labels:
+    app.kubernetes.io/component: extender
+spec:
+  type: ClusterIP
+  ports:
+    - port: 8099
+      targetPort: 8099
+      protocol: TCP
+      name: http
+  selector:
+    app.kubernetes.io/name: linstor-scheduler
+    app.kubernetes.io/instance: linstor-scheduler
+    app.kubernetes.io/component: scheduler
diff --git a/packages/system/linstor-scheduler/tests/extender-service_test.yaml b/packages/system/linstor-scheduler/tests/extender-service_test.yaml
new file mode 100644
index 00000000..60e272c2
--- /dev/null
+++ b/packages/system/linstor-scheduler/tests/extender-service_test.yaml
@@ -0,0 +1,35 @@
+suite: extender service targets deployment pods
+
+templates:
+  - templates/extender-service.yaml
+  - charts/linstor-scheduler/templates/deployment.yaml
+
+release:
+  name: linstor-scheduler
+
+tests:
+  - it: service selector matches deployment pod labels
+    template: charts/linstor-scheduler/templates/deployment.yaml
+    asserts:
+      - equal:
+          path: spec.template.metadata.labels["app.kubernetes.io/name"]
+          value: linstor-scheduler
+      - equal:
+          path: spec.template.metadata.labels["app.kubernetes.io/instance"]
+          value: linstor-scheduler
+      - equal:
+          path: spec.template.metadata.labels["app.kubernetes.io/component"]
+          value: scheduler
+
+  - it: service selector uses the same values
+    template: templates/extender-service.yaml
+    asserts:
+      - equal:
+          path: spec.selector["app.kubernetes.io/name"]
+          value: linstor-scheduler
+      - equal:
+          path: spec.selector["app.kubernetes.io/instance"]
+          value: linstor-scheduler
+      - equal:
+          path: spec.selector["app.kubernetes.io/component"]
+          value: scheduler
diff --git a/packages/system/linstor/Makefile b/packages/system/linstor/Makefile
index 2ad4def9..8d566678 100644
--- a/packages/system/linstor/Makefile
+++ b/packages/system/linstor/Makefile
@@ -5,7 +5,7 @@ include ../../../hack/common-envs.mk
 include ../../../hack/package.mk
 
 LINSTOR_VERSION ?= 1.33.2
-LINSTOR_CSI_VERSION ?= v1.10.5
+LINSTOR_CSI_VERSION ?= v1.10.6
 
 image: image-piraeus-server image-linstor-csi
 
diff --git a/packages/system/linstor/images/linstor-csi/Dockerfile b/packages/system/linstor/images/linstor-csi/Dockerfile
index 6cfd44aa..df5d179d 100644
--- a/packages/system/linstor/images/linstor-csi/Dockerfile
+++ b/packages/system/linstor/images/linstor-csi/Dockerfile
@@ -1,6 +1,6 @@
 FROM golang:1.25 AS builder
 
-ARG VERSION=v1.10.5
+ARG VERSION=v1.10.6
 ARG LINSTOR_WAIT_UNTIL_VERSION=v0.3.1
 ARG TARGETARCH
 ARG TARGETOS
diff --git a/packages/system/linstor/images/linstor-csi/patches/001-relocate-after-clone-restore.diff b/packages/system/linstor/images/linstor-csi/patches/001-relocate-after-clone-restore.diff
index ff0aec0c..0e272670 100644
--- a/packages/system/linstor/images/linstor-csi/patches/001-relocate-after-clone-restore.diff
+++ b/packages/system/linstor/images/linstor-csi/patches/001-relocate-after-clone-restore.diff
@@ -1,25 +1,7 @@
 diff --git a/pkg/client/linstor.go b/pkg/client/linstor.go
-index f544493..98e7fde 100644
+index ac4651d..46d565d 100644
 --- a/pkg/client/linstor.go
 +++ b/pkg/client/linstor.go
-@@ -181,7 +181,7 @@ func LogLevel(s string) func(*Linstor) error {
- func (s *Linstor) ListAllWithStatus(ctx context.Context) ([]volume.VolumeStatus, error) {
- 	var vols []volume.VolumeStatus
- 
--	resourcesByName := make(map[string][]lapi.Resource)
-+	resourcesByName := make(map[string][]lapi.ResourceWithVolumes)
- 
- 	resDefs, err := s.client.ResourceDefinitions.GetAll(ctx, lapi.RDGetAllRequest{WithVolumeDefinitions: true})
- 	if err != nil {
-@@ -194,7 +194,7 @@ func (s *Linstor) ListAllWithStatus(ctx context.Context) ([]volume.VolumeStatus,
- 	}
- 
- 	for i := range allResources {
--		resourcesByName[allResources[i].Name] = append(resourcesByName[allResources[i].Name], allResources[i].Resource)
-+		resourcesByName[allResources[i].Name] = append(resourcesByName[allResources[i].Name], allResources[i])
- 	}
- 
- 	for _, rd := range resDefs {
 @@ -462,6 +462,14 @@ func (s *Linstor) Clone(ctx context.Context, vol, src *volume.Info, params *volu
  		return err
  	}
@@ -35,7 +17,7 @@ index f544493..98e7fde 100644
  	logger.Debug("reconcile extra properties")
  
  	err = s.client.ResourceDefinitions.Modify(ctx, vol.ID, lapi.GenericPropsModify{OverrideProps: vol.Properties})
-@@ -1280,6 +1288,14 @@ func (s *Linstor) VolFromSnap(ctx context.Context, snap *volume.Snapshot, vol *v
+@@ -1297,6 +1305,14 @@ func (s *Linstor) VolFromSnap(ctx context.Context, snap *volume.Snapshot, vol *v
  		return err
  	}
  
@@ -50,7 +32,7 @@ index f544493..98e7fde 100644
  	logger.Debug("reconcile extra properties")
  
  	err = s.client.ResourceDefinitions.Modify(ctx, vol.ID, lapi.GenericPropsModify{OverrideProps: vol.Properties})
-@@ -1470,9 +1486,8 @@ func (s *Linstor) reconcileSnapshotResources(ctx context.Context, snapshot *volu
+@@ -1487,9 +1503,8 @@ func (s *Linstor) reconcileSnapshotResources(ctx context.Context, snapshot *volu
  		return fmt.Errorf("snapshot '%s' not deployed on any node", snap.Name)
  	}
  
@@ -62,7 +44,7 @@ index f544493..98e7fde 100644
  	for _, snapNode := range snap.Nodes {
  		if err := s.NodeAvailable(ctx, snapNode); err != nil {
  			logger.WithField("selected node candidate", snapNode).WithError(err).Debug("node is not available")
-@@ -1480,13 +1495,23 @@ func (s *Linstor) reconcileSnapshotResources(ctx context.Context, snapshot *volu
+@@ -1497,13 +1512,23 @@ func (s *Linstor) reconcileSnapshotResources(ctx context.Context, snapshot *volu
  		}
  
  		if slices.Contains(preferredNodes, snapNode) {
@@ -92,7 +74,7 @@ index f544493..98e7fde 100644
  	}
  
  	if selectedNode == "" {
-@@ -1679,6 +1704,114 @@ func (s *Linstor) reconcileResourcePlacement(ctx context.Context, vol *volume.In
+@@ -1696,6 +1721,114 @@ func (s *Linstor) reconcileResourcePlacement(ctx context.Context, vol *volume.In
  	return nil
  }
  
@@ -207,40 +189,8 @@ index f544493..98e7fde 100644
  // FindSnapsByID searches the snapshot in the backend
  func (s *Linstor) FindSnapsByID(ctx context.Context, id string) ([]*volume.Snapshot, error) {
  	snapshotId, err := volume.ParseSnapshotId(id)
-@@ -2173,7 +2306,7 @@ func (s *Linstor) Status(ctx context.Context, volId string) ([]string, *csi.Volu
- 		"volume": volId,
- 	}).Debug("getting assignments")
- 
--	ress, err := s.client.Resources.GetAll(ctx, volId)
-+	ress, err := s.client.Resources.GetResourceView(ctx, &lapi.ListOpts{Resource: []string{volId}})
- 	if err != nil {
- 		return nil, nil, fmt.Errorf("failed to list resources for '%s': %w", volId, err)
- 	}
-@@ -2261,13 +2394,20 @@ func GetSnapshotRemoteAndReadiness(snap *lapi.Snapshot) (string, bool, error) {
- 	return "", slices.Contains(snap.Flags, lapiconsts.FlagSuccessful), nil
- }
- 
--func NodesAndConditionFromResources(ress []lapi.Resource) ([]string, *csi.VolumeCondition) {
-+func NodesAndConditionFromResources(ress []lapi.ResourceWithVolumes) ([]string, *csi.VolumeCondition) {
- 	var allNodes, abnormalNodes []string
- 
- 	for i := range ress {
- 		res := &ress[i]
- 
--		allNodes = append(allNodes, res.NodeName)
-+		// A resource is a CSI publish target if any of its volumes were created
-+		// by ControllerPublishVolume, identified by the temporary-diskless-attach property.
-+		if slices.ContainsFunc(res.Volumes, func(v lapi.Volume) bool {
-+			createdFor, ok := v.Props[linstor.PropertyCreatedFor]
-+			return ok && createdFor == linstor.CreatedForTemporaryDisklessAttach
-+		}) {
-+			allNodes = append(allNodes, res.NodeName)
-+		}
- 
- 		if res.State == nil {
- 			abnormalNodes = append(abnormalNodes, res.NodeName)
 diff --git a/pkg/volume/parameter.go b/pkg/volume/parameter.go
-index 39acd95..aed18ab 100644
+index 39acd95..54d1dfb 100644
 --- a/pkg/volume/parameter.go
 +++ b/pkg/volume/parameter.go
 @@ -50,6 +50,7 @@ const (
diff --git a/packages/system/linstor/images/linstor-csi/patches/002-protocol-c-override-for-dual-attach.diff b/packages/system/linstor/images/linstor-csi/patches/002-protocol-c-override-for-dual-attach.diff
new file mode 100644
index 00000000..279c636b
--- /dev/null
+++ b/packages/system/linstor/images/linstor-csi/patches/002-protocol-c-override-for-dual-attach.diff
@@ -0,0 +1,132 @@
+diff --git a/pkg/client/linstor.go b/pkg/client/linstor.go
+index ac4651d..9d75c79 100644
+--- a/pkg/client/linstor.go
++++ b/pkg/client/linstor.go
+@@ -653,11 +653,35 @@ func (s *Linstor) Attach(ctx context.Context, volId, node string, rwxBlock bool)
+ 	}
+ 
+ 	if otherResInUse > 0 && rwxBlock {
+-		rdPropsModify := lapi.GenericPropsModify{OverrideProps: map[string]string{
++		rdProps := map[string]string{
+ 			linstor.PropertyAllowTwoPrimaries: "yes",
+-		}}
++		}
+ 
+-		err = s.client.ResourceDefinitions.Modify(ctx, volId, rdPropsModify)
++		// DRBD requires Protocol C whenever allow-two-primaries is enabled.
++		// If the resource is configured with Protocol A or B (typically
++		// through its resource-group), drbdadm adjust on the satellites
++		// fails with "Protocol C required" (errno 139) and live migration
++		// cannot proceed. Override Protocol to C on the resource-definition
++		// itself so it applies to every connection (including diskless
++		// peers, e.g. a TieBreaker). The marker lets Detach revert exactly
++		// the override we installed without disturbing operator-set props.
++		proto, protoErr := s.getEffectiveDrbdProtocol(ctx, volId)
++		if protoErr != nil {
++			s.log.WithError(protoErr).WithField("volume", volId).
++				Warn("failed to determine effective DRBD protocol; skipping Protocol=C override for dual-attach")
++		} else if proto == "A" || proto == "B" {
++			s.log.WithFields(logrus.Fields{
++				"volume":   volId,
++				"protocol": proto,
++			}).Info("installing Protocol=C override on resource-definition for dual-attach")
++
++			rdProps[linstor.PropertyDrbdNetProtocol] = "C"
++			rdProps[linstor.PropertyCsiProtocolOverride] = "yes"
++		}
++
++		err = s.client.ResourceDefinitions.Modify(ctx, volId, lapi.GenericPropsModify{
++			OverrideProps: rdProps,
++		})
+ 		if err != nil {
+ 			return "", err
+ 		}
+@@ -774,11 +798,26 @@ func (s *Linstor) Detach(ctx context.Context, volId, node string) error {
+ 	}
+ 
+ 	if resInUse == 1 {
+-		rdPropsModify := lapi.GenericPropsModify{DeleteProps: []string{
+-			linstor.PropertyAllowTwoPrimaries,
+-		}}
++		deleteProps := []string{linstor.PropertyAllowTwoPrimaries}
++
++		// Drop the Protocol=C override only if Attach installed it (gated by
++		// our marker), so an operator-set Protocol property on the
++		// resource-definition is preserved.
++		rd, getErr := s.client.ResourceDefinitions.Get(ctx, volId)
++		if getErr != nil {
++			log.WithError(getErr).Warn("failed to fetch resource-definition props; skipping Protocol=C override removal")
++		} else if rd.Props[linstor.PropertyCsiProtocolOverride] != "" {
++			log.Info("removing Protocol=C override from resource-definition")
+ 
+-		err = s.client.ResourceDefinitions.Modify(ctx, volId, rdPropsModify)
++			deleteProps = append(deleteProps,
++				linstor.PropertyDrbdNetProtocol,
++				linstor.PropertyCsiProtocolOverride,
++			)
++		}
++
++		err = s.client.ResourceDefinitions.Modify(ctx, volId, lapi.GenericPropsModify{
++			DeleteProps: deleteProps,
++		})
+ 		if err != nil {
+ 			return nil404(err)
+ 		}
+@@ -805,6 +844,33 @@ func (s *Linstor) Detach(ctx context.Context, volId, node string) error {
+ 	return nil404(s.client.Resources.Delete(ctx, volId, node))
+ }
+ 
++// getEffectiveDrbdProtocol returns the DRBD network protocol ("A", "B" or "C")
++// that LINSTOR will hand to drbdadm for the given resource. Resource-definition
++// properties take precedence; otherwise the value is inherited from the
++// resource-group. An empty string is returned when neither level sets the
++// property (LINSTOR's compiled-in default is "C").
++func (s *Linstor) getEffectiveDrbdProtocol(ctx context.Context, volId string) (string, error) {
++	rd, err := s.client.ResourceDefinitions.Get(ctx, volId)
++	if err != nil {
++		return "", err
++	}
++
++	if v, ok := rd.Props[linstor.PropertyDrbdNetProtocol]; ok {
++		return v, nil
++	}
++
++	if rd.ResourceGroupName == "" {
++		return "", nil
++	}
++
++	rg, err := s.client.ResourceGroups.Get(ctx, rd.ResourceGroupName)
++	if err != nil {
++		return "", err
++	}
++
++	return rg.Props[linstor.PropertyDrbdNetProtocol], nil
++}
++
+ // CapacityBytes returns the amount of free space in the storage pool specified by the params and topology.
+ func (s *Linstor) CapacityBytes(ctx context.Context, storagePools []string, overProvision *float64, segments map[string]string) (int64, error) {
+ 	log := s.log.WithField("storage-pools", storagePools).WithField("segments", segments)
+diff --git a/pkg/linstor/const.go b/pkg/linstor/const.go
+index 9a5f79c..c8bc9c3 100644
+--- a/pkg/linstor/const.go
++++ b/pkg/linstor/const.go
+@@ -44,6 +44,19 @@ const (
+ 	// PropertyAllowTwoPrimaries is DRBD option to allow second primary. Mainly used for live-migration.
+ 	PropertyAllowTwoPrimaries = lc.NamespcDrbdNetOptions + "/allow-two-primaries"
+ 
++	// PropertyDrbdNetProtocol is the DRBD network replication protocol option
++	// (A = async, B = semi-sync, C = sync). DRBD requires Protocol C whenever
++	// allow-two-primaries is enabled, otherwise drbdadm adjust fails with
++	// "Protocol C required" (errno 139).
++	PropertyDrbdNetProtocol = lc.NamespcDrbdNetOptions + "/protocol"
++
++	// PropertyCsiProtocolOverride marks a resource-connection where this driver
++	// has installed a temporary Protocol=C override to make a Protocol-A/B
++	// resource compatible with allow-two-primaries during live migration. The
++	// marker lets us safely remove only the overrides we set, without touching
++	// connection-level overrides installed by the operator.
++	PropertyCsiProtocolOverride = lc.NamespcAuxiliary + "/csi-protocol-override"
++
+ 	// CreatedForTemporaryDisklessAttach marks a resource as temporary, i.e. it should be removed after it is no longer
+ 	// needed.
+ 	CreatedForTemporaryDisklessAttach = "temporary-diskless-attach"
diff --git a/packages/system/linstor/templates/satellites-cozy.yaml b/packages/system/linstor/templates/satellites-cozy.yaml
index c621126d..e6f877a0 100644
--- a/packages/system/linstor/templates/satellites-cozy.yaml
+++ b/packages/system/linstor/templates/satellites-cozy.yaml
@@ -14,6 +14,11 @@ spec:
       containers:
       - name: linstor-satellite
         image: {{ .Values.piraeusServer.image.repository }}:{{ .Values.piraeusServer.image.tag }}
+        startupProbe:
+          tcpSocket:
+            port: linstor
+          periodSeconds: 10
+          failureThreshold: 30
         securityContext:
           # real-world installations need some debugging from time to time
           readOnlyRootFilesystem: false
diff --git a/packages/system/linstor/values.yaml b/packages/system/linstor/values.yaml
index b3f13443..62dd2cac 100644
--- a/packages/system/linstor/values.yaml
+++ b/packages/system/linstor/values.yaml
@@ -1,7 +1,7 @@
 piraeusServer:
   image:
     repository: ghcr.io/cozystack/cozystack/piraeus-server
-    tag: 1.32.3@sha256:d78071fdc33220168e3f2ab112a86208be95898b75268a0c62763b8a04e145ad
+    tag: 1.33.2@sha256:553f313ab35dc2e345ef3683156d29e75c23177e2750e9af3a83aa9e23941cbb
 # Talos-specific workarounds (disable for generic Linux like Ubuntu/Debian)
 talos:
   enabled: true
@@ -13,4 +13,4 @@ linstor:
 linstorCSI:
   image:
     repository: ghcr.io/cozystack/cozystack/linstor-csi
-    tag: v1.10.5@sha256:e80ae94d8acd8774cb35715b1f103071ae6cd3160f1ebea915e5e6ee49234448
+    tag: v1.10.5@sha256:b8f59b5659fb1791cb764d3f37df4cf29920aadcc10637231ba7d857233f377d
diff --git a/packages/system/monitoring-agents/alerts/gpu-recording.rules.yaml b/packages/system/monitoring-agents/alerts/gpu-recording.rules.yaml
new file mode 100644
index 00000000..5e0f9ea0
--- /dev/null
+++ b/packages/system/monitoring-agents/alerts/gpu-recording.rules.yaml
@@ -0,0 +1,172 @@
+apiVersion: operator.victoriametrics.com/v1beta1
+kind: VMRule
+metadata:
+  name: alerts-gpu-recording.rules
+spec:
+  groups:
+    - name: gpu.recording.cluster.1m
+      interval: 1m
+      params: {}
+      rules:
+        - record: cluster:gpu_count:total
+          expr: count(group by (UUID) (DCGM_FI_DEV_GPU_UTIL))
+        # Kube-allocated GPU count: GPUs requested by *active* pods — Pending
+        # and Running only. Source of truth for "what tenants asked for" — used
+        # for capacity planning and billing. Includes system pods so it stays
+        # consistent with cluster:gpu_count:total when computing :free.
+        #
+        # Phase filter via group_left() on kube_pod_status_phase matches the
+        # canonical pattern from k8s.rules.container_resource.yaml. Without it
+        # kube-state-metrics keeps series for Failed/Succeeded pods until the
+        # apiserver garbage-collects them, which inflates billing hours and
+        # can push :free negative.
+        - record: cluster:gpu_count:allocated
+          expr: |
+            sum(
+              kube_pod_container_resource_requests{resource="nvidia_com_gpu"}
+              * on (namespace, pod) group_left() max by (namespace, pod) (
+                kube_pod_status_phase{phase=~"Pending|Running"} == 1
+              )
+            )
+        # clamp_min guards against transients at DCGM/kube-state-metrics
+        # restart where cluster:gpu_count:total dips before :allocated catches
+        # up, or rare label drift between the two sources.
+        - record: cluster:gpu_count:free
+          expr: clamp_min(cluster:gpu_count:total - (cluster:gpu_count:allocated or vector(0)), 0)
+        # Cluster-layer filter asymmetry — intentional, not a bug:
+        #
+        #   * cluster:gpu_count:total and cluster:gpu_power_watts:sum do NOT
+        #     filter cozy-*/kube-* namespaces because they describe the
+        #     physical hardware fleet. A GPU draws power and occupies a
+        #     slot regardless of which namespace's pod happens to hold it;
+        #     excluding infra pods would under-report raw capacity and
+        #     break capacity-planning math.
+        #
+        # No namespace filter — DCGM exporter metrics carry the exporter's
+        # own namespace, not the workload namespace. Filtering by namespace
+        # here would silently drop all series.
+        - record: cluster:gpu_util:avg
+          expr: avg(DCGM_FI_DEV_GPU_UTIL)
+        - record: cluster:gpu_power_watts:sum
+          expr: sum(DCGM_FI_DEV_POWER_USAGE)
+
+    - name: gpu.recording.node.1m
+      interval: 1m
+      params: {}
+      rules:
+        # Kube-requested GPU count per namespace — billable/allocation view.
+        # This is the only namespace-level rule that works correctly because
+        # kube_pod_container_resource_requests carries the real workload
+        # namespace, unlike DCGM exporter metrics which carry the exporter's
+        # own namespace (cozy-gpu-operator).
+        - record: namespace:gpu_count:allocated
+          expr: |
+            sum by (namespace) (
+              kube_pod_container_resource_requests{resource="nvidia_com_gpu", namespace!="", namespace!~"cozy-.*|kube-.*"}
+              * on (namespace, pod) group_left() max by (namespace, pod) (
+                kube_pod_status_phase{phase=~"Pending|Running"} == 1
+              )
+            )
+        # Hardware metrics aggregated per node (Hostname). DCGM exporter
+        # metrics carry a capital-H "Hostname" label identifying the node
+        # but no usable workload namespace — so hardware telemetry is
+        # aggregated at node/GPU granularity, not namespace.
+        - record: node:gpu_util:avg
+          expr: avg by (Hostname) (DCGM_FI_DEV_GPU_UTIL)
+        - record: node:tensor_active:avg
+          expr: avg by (Hostname) (DCGM_FI_PROF_PIPE_TENSOR_ACTIVE)
+        - record: node:fb_used_bytes:sum
+          expr: sum by (Hostname) (DCGM_FI_DEV_FB_USED) * 1048576
+        - record: node:power_watts:sum
+          expr: sum by (Hostname) (DCGM_FI_DEV_POWER_USAGE)
+
+    # Known gap: there is no lower-bound sanity alert for under-reporting.
+    # If DCGM ever switches POWER/THERMAL_VIOLATION counter units the other
+    # direction — e.g. from the current nanoseconds back to microseconds,
+    # making rate()/1e9 produce values around 0.001 instead of real
+    # fractions — the throttle signal would silently collapse to near-zero
+    # instead of plateauing at 1.0. GPUThrottleFractionOverOne below catches
+    # the >1 drift; the <<1 drift would require correlation with independent
+    # signals (power draw vs TDP, clock dips) that we do not yet wire up.
+    # Documented here so future maintainers know this is a known gap, not
+    # an oversight.
+    - name: gpu.recording.efficiency.1m
+      interval: 1m
+      params: {}
+      rules:
+        # Tensor hardware saturation per GPU — the honest "am I using the GPU"
+        # metric for AI/LLM workloads. Aggregated at GPU level (Hostname +
+        # gpu + UUID) because DCGM metrics don't carry workload namespace.
+        # Stored as a 0..1 ratio; consumers multiply by 100 at display time.
+        - record: gpu:tensor_saturation:avg5m
+          expr: |
+            avg by (Hostname, gpu, UUID) (
+              avg_over_time(DCGM_FI_PROF_PIPE_TENSOR_ACTIVE[5m])
+            )
+
+        # Power efficiency per GPU — utilization per watt, reveals
+        # unoptimized workloads. Explicit on(...) pins the matching set
+        # to GPU-identifying labels.
+        - record: gpu:util_per_watt:avg5m
+          expr: |
+            max by (Hostname, gpu, UUID) (
+              avg_over_time(DCGM_FI_DEV_GPU_UTIL[5m])
+            )
+            / on (Hostname, gpu, UUID)
+            clamp_min(
+              max by (Hostname, gpu, UUID) (
+                avg_over_time(DCGM_FI_DEV_POWER_USAGE[5m])
+              ),
+              1
+            )
+
+        # Fraction of time power-throttled (TDP cap) — 1.0 = fully throttled.
+        # DCGM_FI_DEV_*_VIOLATION is documented as µs but on A10/DCGM 3.x the
+        # counter grows in nanoseconds in practice — divide by 1e9 to get a
+        # 0..1 fraction (verified empirically when /1e6 yielded >100× reality).
+        # clamp_max protects against rate() artefacts at counter resets.
+        #
+        # max by (Hostname, gpu, UUID) collapses duplicate series from
+        # dcgm-exporter's pod-mapping labels (one physical GPU may appear
+        # under multiple pod/container labels during restarts or under
+        # MIG/MPS). Throttling is a physical-GPU property; taking max
+        # keeps consumer avg(...) honest — a naive avg() would dilute the
+        # signal by the number of pods sharing the GPU.
+        - record: gpu:power_throttle_fraction:rate5m
+          expr: clamp_max(max by (Hostname, gpu, UUID) (rate(DCGM_FI_DEV_POWER_VIOLATION[5m]) / 1e9), 1)
+
+        # Fraction of time thermal-throttled.
+        - record: gpu:thermal_throttle_fraction:rate5m
+          expr: clamp_max(max by (Hostname, gpu, UUID) (rate(DCGM_FI_DEV_THERMAL_VIOLATION[5m]) / 1e9), 1)
+
+    # Regression watch — the /1e9 divisor on *_VIOLATION was derived
+    # empirically against DCGM 3.x on A10. If a future DCGM version
+    # restores the documented µs units, pre-clamp values would exceed 1.0
+    # while clamp_max(..., 1) silently masks the drift — recorded
+    # throttle fractions would plateau at 1.0 and consumers would think
+    # every GPU is always throttled. This alert fires on that condition
+    # so we can rescale to /1e6 before dashboards start lying.
+    - name: gpu.recording.throttle.validation.5m
+      interval: 5m
+      params: {}
+      rules:
+        - alert: GPUThrottleFractionOverOne
+          expr: |
+            max by (Hostname, gpu, UUID) (rate(DCGM_FI_DEV_POWER_VIOLATION[5m]) / 1e9) > 1
+            or
+            max by (Hostname, gpu, UUID) (rate(DCGM_FI_DEV_THERMAL_VIOLATION[5m]) / 1e9) > 1
+          for: 15m
+          labels:
+            severity: warning
+            component: gpu-monitoring
+          annotations:
+            summary: "DCGM throttle fraction exceeds 1.0 pre-clamp on {{ $labels.Hostname }}/{{ $labels.gpu }}"
+            description: |
+              gpu:{power,thermal}_throttle_fraction:rate5m clamps to 1.0, but the raw
+              rate(DCGM_FI_DEV_*_VIOLATION)/1e9 value is already >1 on this GPU — the
+              empirical nanosecond assumption is broken. Likely the DCGM exporter
+              version changed and the counter now grows in its documented µs units.
+              Re-verify on the exporter Pod and adjust the divisor (/1e9 → /1e6) in
+              gpu.recording.efficiency.1m before dashboards plateau at 100% throttle.
+            verified_on: "NVIDIA A10, DCGM 3.x — other GPU families (H100, L40S) may require a different divisor"
+            runbook: "If this alert fires, DCGM may have changed POWER_VIOLATION/THERMAL_VIOLATION counter units. Recalibrate by comparing raw counter rate against known throttle events; see gpu-recording.rules.yaml comments."
diff --git a/packages/system/monitoring/dashboards-infra.list b/packages/system/monitoring/dashboards-infra.list
index b5b9f52a..4c7e494d 100644
--- a/packages/system/monitoring/dashboards-infra.list
+++ b/packages/system/monitoring/dashboards-infra.list
@@ -32,3 +32,8 @@ hubble/overview
 hubble/dns-namespace
 hubble/l7-http-metrics
 hubble/network-overview
+gpu/gpu-performance
+gpu/gpu-efficiency
+gpu/gpu-quotas
+gpu/gpu-fleet
+gpu/gpu-tenants
diff --git a/packages/system/monitoring/templates/alerta/alerta-db.yaml b/packages/system/monitoring/templates/alerta/alerta-db.yaml
index 71df5428..af9dc72c 100644
--- a/packages/system/monitoring/templates/alerta/alerta-db.yaml
+++ b/packages/system/monitoring/templates/alerta/alerta-db.yaml
@@ -5,7 +5,9 @@ metadata:
   name: alerta-db
 spec:
   instances: 2
-  imageName: ghcr.io/cloudnative-pg/postgresql:17.7
+  {{- $existingCluster := lookup "postgresql.cnpg.io/v1" "Cluster" .Release.Namespace "alerta-db" }}
+  {{- $image := dig "spec" "imageName" "ghcr.io/cloudnative-pg/postgresql:17.7-standard-trixie" $existingCluster }}
+  imageName: {{ if regexMatch ":17\\." $image }}ghcr.io/cloudnative-pg/postgresql:17.7-standard-trixie{{ else if regexMatch ":[0-9]+\\.[0-9]+$" $image }}{{ printf "%s-standard-trixie" $image }}{{ else }}{{ $image }}{{ end }}
   {{- if .Values._cluster.scheduling }}
   {{- $rawConstraints := get .Values._cluster.scheduling "globalAppTopologySpreadConstraints" }}
   {{- if $rawConstraints }}
diff --git a/packages/system/monitoring/templates/alerta/alerta.yaml b/packages/system/monitoring/templates/alerta/alerta.yaml
index 44046d1f..d30e698d 100644
--- a/packages/system/monitoring/templates/alerta/alerta.yaml
+++ b/packages/system/monitoring/templates/alerta/alerta.yaml
@@ -181,7 +181,7 @@ metadata:
     app: alerta
   annotations:
     {{- if eq $solver "http01" }}
-    acme.cert-manager.io/http01-ingress-class: {{ $ingress }}
+    acme.cert-manager.io/http01-ingress-ingressclassname: {{ $ingress }}
     {{- end }}
     cert-manager.io/cluster-issuer: {{ $clusterIssuer }}
 spec:
diff --git a/packages/system/monitoring/templates/dashboards.yaml b/packages/system/monitoring/templates/dashboards.yaml
index 01f1c009..0427bfa0 100644
--- a/packages/system/monitoring/templates/dashboards.yaml
+++ b/packages/system/monitoring/templates/dashboards.yaml
@@ -14,7 +14,7 @@ spec:
   url: http://grafana-dashboards.cozy-grafana-operator.svc/{{ . }}.json
 {{- end }}
 {{- end }}
-{{- if eq .Release.Namespace "tenant-root" }}
+{{- if or (eq .Release.Namespace "tenant-root") (eq .Release.Namespace "cozy-monitoring") }}
 {{- range (split "\n" (.Files.Get "dashboards-infra.list")) }}
 {{- $parts := split "/" . }}
 {{- if eq (len $parts) 2 }}
diff --git a/packages/system/monitoring/templates/grafana/db.yaml b/packages/system/monitoring/templates/grafana/db.yaml
index 73d6502e..afc2d0d7 100644
--- a/packages/system/monitoring/templates/grafana/db.yaml
+++ b/packages/system/monitoring/templates/grafana/db.yaml
@@ -4,7 +4,9 @@ metadata:
   name: grafana-db
 spec:
   instances: 2
-  imageName: ghcr.io/cloudnative-pg/postgresql:17.7
+  {{- $existingCluster := lookup "postgresql.cnpg.io/v1" "Cluster" .Release.Namespace "grafana-db" }}
+  {{- $image := dig "spec" "imageName" "ghcr.io/cloudnative-pg/postgresql:17.7-standard-trixie" $existingCluster }}
+  imageName: {{ if regexMatch ":17\\." $image }}ghcr.io/cloudnative-pg/postgresql:17.7-standard-trixie{{ else if regexMatch ":[0-9]+\\.[0-9]+$" $image }}{{ printf "%s-standard-trixie" $image }}{{ else }}{{ $image }}{{ end }}
   storage:
     size: {{ .Values.grafana.db.size }}
   {{- if .Values._cluster.scheduling }}
diff --git a/packages/system/monitoring/templates/grafana/grafana.yaml b/packages/system/monitoring/templates/grafana/grafana.yaml
index d7f51e31..d65a7dc4 100644
--- a/packages/system/monitoring/templates/grafana/grafana.yaml
+++ b/packages/system/monitoring/templates/grafana/grafana.yaml
@@ -74,7 +74,7 @@ spec:
     metadata:
       annotations:
         {{- if eq $solver "http01" }}
-        acme.cert-manager.io/http01-ingress-class: "{{ $ingress }}"
+        acme.cert-manager.io/http01-ingress-ingressclassname: "{{ $ingress }}"
         {{- end }}
         cert-manager.io/cluster-issuer: {{ $clusterIssuer }}
     spec:
diff --git a/packages/system/multus/templates/multus-daemonset-thick.yml b/packages/system/multus/templates/multus-daemonset-thick.yml
index bedffee3..cad330ea 100644
--- a/packages/system/multus/templates/multus-daemonset-thick.yml
+++ b/packages/system/multus/templates/multus-daemonset-thick.yml
@@ -155,7 +155,7 @@ spec:
       serviceAccountName: multus
       containers:
         - name: kube-multus
-          image: ghcr.io/cozystack/cozystack/multus-cni:v1.2.1@sha256:aaf2ed6a5db1ee5acc0cd4f4683ea33570ff922c211bef37d407d6a01427b566
+          image: ghcr.io/cozystack/cozystack/multus-cni:v1.3.0@sha256:6735ffc12e5e660951f5a42943b38c308f33774a55836e94c191001405b58ec0
           command: [ "/usr/src/multus-cni/bin/multus-daemon" ]
           resources:
             requests:
@@ -201,7 +201,7 @@ spec:
                   fieldPath: spec.nodeName
       initContainers:
         - name: install-multus-binary
-          image: ghcr.io/cozystack/cozystack/multus-cni:v1.2.1@sha256:aaf2ed6a5db1ee5acc0cd4f4683ea33570ff922c211bef37d407d6a01427b566
+          image: ghcr.io/cozystack/cozystack/multus-cni:v1.3.0@sha256:6735ffc12e5e660951f5a42943b38c308f33774a55836e94c191001405b58ec0
           command:
             - "/usr/src/multus-cni/bin/install_multus"
             - "-d"
diff --git a/packages/system/objectstorage-controller/values.yaml b/packages/system/objectstorage-controller/values.yaml
index 1439e583..a3859194 100644
--- a/packages/system/objectstorage-controller/values.yaml
+++ b/packages/system/objectstorage-controller/values.yaml
@@ -1,3 +1,3 @@
 objectstorage:
   controller:
-    image: "ghcr.io/cozystack/cozystack/objectstorage-controller:v1.2.1@sha256:036be10f26c4871c63fb63c6a294476289aa063610e7fbbc0dbd21d3cd7bfe5e"
+    image: "ghcr.io/cozystack/cozystack/objectstorage-controller:v1.3.0@sha256:7a9e4bf9c3f95f364756396815bb51b6c0f58f85db653460574d94b96cd3c7d5"
diff --git a/packages/system/postgres-operator/Makefile b/packages/system/postgres-operator/Makefile
index f6d7ddaa..5279f8a7 100644
--- a/packages/system/postgres-operator/Makefile
+++ b/packages/system/postgres-operator/Makefile
@@ -3,6 +3,9 @@ export NAMESPACE=cozy-$(NAME)
 
 include ../../../hack/package.mk
 
+test:
+	helm unittest .
+
 update:
 	rm -rf charts
 	helm repo add cnpg https://cloudnative-pg.github.io/charts
diff --git a/packages/system/postgres-operator/templates/webhook-ready-hook.yaml b/packages/system/postgres-operator/templates/webhook-ready-hook.yaml
new file mode 100644
index 00000000..51a8ba81
--- /dev/null
+++ b/packages/system/postgres-operator/templates/webhook-ready-hook.yaml
@@ -0,0 +1,148 @@
+{{- /*
+  Post-install gate: block the HelmRelease from reporting Ready until the
+  cnpg admission webhook actually serves through the cluster Service. Helm
+  --wait on the controller pod passes once its readinessProbe passes, but
+  EndpointSlice propagation and kube-proxy/cilium data-plane programming
+  can lag by a second or two — long enough for any HelmRelease that
+  depends on postgres-operator (e.g. cozy-keycloak, tenant Postgres apps)
+  to fire its own install and have kube-apiserver hit the mcluster.cnpg.io
+  mutating webhook with "dial tcp <svc-ip>:443: connect: connection refused".
+
+  The Job uses the apiserver service proxy, which exercises the same
+  endpoint-resolution and apiserver-initiated pod dial that the admission
+  webhook path uses. Once /readyz answers through the proxy the data-plane
+  race is resolved. It does not verify the webhook's TLS CA bundle, so
+  this gate is scoped to reachability regressions, not cert rotation.
+
+  The service name and port name are hardcoded literals. Upstream cnpg
+  pins the service name in charts/cloudnative-pg/values.yaml with a
+  comment "DO NOT CHANGE THE SERVICE NAME as it is currently used to
+  generate the certificate and can not be configured". The port name is
+  fixed in charts/cloudnative-pg/templates/service.yaml (ports[0].name:
+  webhook-server). If a future `make update` ever changes either literal
+  upstream, the sync-check helm-unittest test
+  (tests/webhook-ready-hook_test.yaml) renders the subchart Service and
+  fails if the literal drifts — forcing this template to be updated in
+  the same change.
+*/}}
+{{- $_ := required "webhookReady.image.repository must be set to the container image providing kubectl for the post-install readiness Job" .Values.webhookReady.image.repository -}}
+{{- $_ := required "webhookReady.image.tag must be set for the post-install readiness Job" .Values.webhookReady.image.tag -}}
+{{- /* $svcName and $portName are hardcoded literals; see header comment. */ -}}
+{{- $svcName := "cnpg-webhook-service" -}}
+{{- /* $portName is the service port NAME, not number — matches ports[0].name in the vendored subchart's Service. */ -}}
+{{- $portName := "webhook-server" -}}
+{{- $resourceName := printf "https:%s:%s" $svcName $portName -}}
+{{- $maxAttempts := .Values.webhookReady.maxAttempts | default 60 -}}
+{{- $sleepSeconds := .Values.webhookReady.sleepSeconds | default 2 -}}
+{{- /* Derive activeDeadlineSeconds from retries + 60s slack so a values override that raises maxAttempts doesn't get silently cut. */ -}}
+{{- $deadline := add (mul (int $maxAttempts) (int $sleepSeconds)) 60 -}}
+{{- $image := printf "%s:%s" .Values.webhookReady.image.repository .Values.webhookReady.image.tag -}}
+{{- if .Values.webhookReady.image.digest }}
+{{- $image = printf "%s@%s" $image .Values.webhookReady.image.digest -}}
+{{- end }}
+---
+apiVersion: v1
+kind: ServiceAccount
+metadata:
+  name: {{ .Release.Name }}-webhook-ready
+  namespace: {{ .Release.Namespace }}
+  annotations:
+    helm.sh/hook: post-install,post-upgrade
+    helm.sh/hook-weight: "0"
+    helm.sh/hook-delete-policy: before-hook-creation,hook-succeeded
+---
+apiVersion: rbac.authorization.k8s.io/v1
+kind: Role
+metadata:
+  name: {{ .Release.Name }}-webhook-ready
+  namespace: {{ .Release.Namespace }}
+  annotations:
+    helm.sh/hook: post-install,post-upgrade
+    helm.sh/hook-weight: "0"
+    helm.sh/hook-delete-policy: before-hook-creation,hook-succeeded
+rules:
+  - apiGroups: [""]
+    resources: ["services/proxy"]
+    resourceNames: [{{ $resourceName | quote }}]
+    verbs: ["get"]
+---
+apiVersion: rbac.authorization.k8s.io/v1
+kind: RoleBinding
+metadata:
+  name: {{ .Release.Name }}-webhook-ready
+  namespace: {{ .Release.Namespace }}
+  annotations:
+    helm.sh/hook: post-install,post-upgrade
+    helm.sh/hook-weight: "0"
+    helm.sh/hook-delete-policy: before-hook-creation,hook-succeeded
+roleRef:
+  apiGroup: rbac.authorization.k8s.io
+  kind: Role
+  name: {{ .Release.Name }}-webhook-ready
+subjects:
+  - kind: ServiceAccount
+    name: {{ .Release.Name }}-webhook-ready
+    namespace: {{ .Release.Namespace }}
+---
+apiVersion: batch/v1
+kind: Job
+metadata:
+  name: {{ .Release.Name }}-webhook-ready
+  namespace: {{ .Release.Namespace }}
+  annotations:
+    helm.sh/hook: post-install,post-upgrade
+    helm.sh/hook-weight: "10"
+    helm.sh/hook-delete-policy: before-hook-creation,hook-succeeded
+spec:
+  backoffLimit: {{ .Values.webhookReady.backoffLimit | default 2 }}
+  activeDeadlineSeconds: {{ $deadline }}
+  template:
+    spec:
+      restartPolicy: Never
+      serviceAccountName: {{ .Release.Name }}-webhook-ready
+      securityContext:
+        runAsNonRoot: true
+        runAsUser: 65532
+        runAsGroup: 65532
+        seccompProfile:
+          type: RuntimeDefault
+      containers:
+        - name: wait
+          image: {{ $image }}
+          imagePullPolicy: {{ if .Values.webhookReady.image.digest }}IfNotPresent{{ else }}Always{{ end }}
+          resources:
+            requests:
+              cpu: 10m
+              memory: 32Mi
+            limits:
+              cpu: 100m
+              memory: 64Mi
+          securityContext:
+            allowPrivilegeEscalation: false
+            readOnlyRootFilesystem: true
+            capabilities:
+              drop: ["ALL"]
+          command:
+            - sh
+            - -c
+            - |
+              set -e
+              ns={{ .Release.Namespace }}
+              proxy="/api/v1/namespaces/${ns}/services/{{ $resourceName }}/proxy/readyz"
+              max_attempts={{ $maxAttempts }}
+              sleep_seconds={{ $sleepSeconds }}
+              i=0
+              last_err=""
+              until last_err=$(kubectl get --raw "$proxy" 2>&1 >/dev/null); do
+                i=$((i + 1))
+                if [ $i -gt $max_attempts ]; then
+                  echo "timeout: cnpg webhook did not respond through the apiserver proxy after ${max_attempts} attempts (${sleep_seconds}s each)"
+                  echo "last error: ${last_err}"
+                  exit 1
+                fi
+                if [ $((i % 10)) -eq 1 ]; then
+                  echo "attempt $i/${max_attempts}: ${last_err}"
+                fi
+                sleep "$sleep_seconds"
+              done
+              echo "cnpg webhook is ready"
diff --git a/packages/system/postgres-operator/tests/webhook-ready-hook_test.yaml b/packages/system/postgres-operator/tests/webhook-ready-hook_test.yaml
new file mode 100644
index 00000000..3253c663
--- /dev/null
+++ b/packages/system/postgres-operator/tests/webhook-ready-hook_test.yaml
@@ -0,0 +1,328 @@
+suite: cnpg webhook post-install readiness gate
+
+templates:
+  - templates/webhook-ready-hook.yaml
+  - charts/cloudnative-pg/templates/service.yaml
+
+release:
+  name: postgres-operator
+  namespace: cozy-postgres-operator
+
+tests:
+  - it: renders four hook objects (SA + Role + RoleBinding + Job)
+    template: templates/webhook-ready-hook.yaml
+    asserts:
+      - hasDocuments:
+          count: 4
+
+  - it: every rendered object carries post-install and post-upgrade hook annotations
+    template: templates/webhook-ready-hook.yaml
+    asserts:
+      - documentIndex: 0
+        equal:
+          path: metadata.annotations["helm.sh/hook"]
+          value: post-install,post-upgrade
+      - documentIndex: 1
+        equal:
+          path: metadata.annotations["helm.sh/hook"]
+          value: post-install,post-upgrade
+      - documentIndex: 2
+        equal:
+          path: metadata.annotations["helm.sh/hook"]
+          value: post-install,post-upgrade
+      - documentIndex: 3
+        equal:
+          path: metadata.annotations["helm.sh/hook"]
+          value: post-install,post-upgrade
+
+  - it: RBAC is created before the Job (hook-weight ordering)
+    template: templates/webhook-ready-hook.yaml
+    asserts:
+      - documentIndex: 0
+        equal:
+          path: kind
+          value: ServiceAccount
+      - documentIndex: 0
+        equal:
+          path: metadata.annotations["helm.sh/hook-weight"]
+          value: "0"
+      - documentIndex: 1
+        equal:
+          path: kind
+          value: Role
+      - documentIndex: 1
+        equal:
+          path: metadata.annotations["helm.sh/hook-weight"]
+          value: "0"
+      - documentIndex: 2
+        equal:
+          path: kind
+          value: RoleBinding
+      - documentIndex: 2
+        equal:
+          path: metadata.annotations["helm.sh/hook-weight"]
+          value: "0"
+      - documentIndex: 3
+        equal:
+          path: kind
+          value: Job
+      - documentIndex: 3
+        equal:
+          path: metadata.annotations["helm.sh/hook-weight"]
+          value: "10"
+
+  - it: RBAC resourceName matches the exact proxy URL segment the Job probes
+    template: templates/webhook-ready-hook.yaml
+    asserts:
+      - documentIndex: 1
+        equal:
+          path: rules[0].resources[0]
+          value: services/proxy
+      - documentIndex: 1
+        equal:
+          path: rules[0].resourceNames[0]
+          value: https:cnpg-webhook-service:webhook-server
+      - documentIndex: 3
+        matchRegex:
+          path: spec.template.spec.containers[0].command[2]
+          pattern: "/api/v1/namespaces/\\$\\{ns\\}/services/https:cnpg-webhook-service:webhook-server/proxy/readyz"
+
+  - it: hardcoded service name in the hook matches the vendored cnpg subchart Service (drift guard for make update)
+    template: charts/cloudnative-pg/templates/service.yaml
+    asserts:
+      - equal:
+          path: metadata.name
+          value: cnpg-webhook-service
+      - equal:
+          path: spec.ports[0].name
+          value: webhook-server
+
+  - it: Job calls kubectl get --raw on the proxy path
+    template: templates/webhook-ready-hook.yaml
+    asserts:
+      - documentIndex: 3
+        matchRegex:
+          path: spec.template.spec.containers[0].command[2]
+          pattern: "kubectl get --raw"
+
+  - it: Job image is digest-pinned when webhookReady.image.digest is set
+    template: templates/webhook-ready-hook.yaml
+    set:
+      webhookReady:
+        image:
+          repository: docker.io/clastix/kubectl
+          tag: v1.32
+          digest: sha256:b9ef7d8dbe65bcc81a46c09b8dc7543103055021c4f43287bf59e92a8f4fe05c
+    asserts:
+      - documentIndex: 3
+        equal:
+          path: spec.template.spec.containers[0].image
+          value: docker.io/clastix/kubectl:v1.32@sha256:b9ef7d8dbe65bcc81a46c09b8dc7543103055021c4f43287bf59e92a8f4fe05c
+
+  - it: Job image falls back to tag-only when digest is not configured
+    template: templates/webhook-ready-hook.yaml
+    set:
+      webhookReady:
+        image:
+          repository: docker.io/clastix/kubectl
+          tag: v1.32
+          digest: ""
+    asserts:
+      - documentIndex: 3
+        equal:
+          path: spec.template.spec.containers[0].image
+          value: docker.io/clastix/kubectl:v1.32
+
+  - it: backoffLimit defaults to 2 so transient pod-level failures retry instead of killing the HelmRelease
+    template: templates/webhook-ready-hook.yaml
+    asserts:
+      - documentIndex: 3
+        equal:
+          path: spec.backoffLimit
+          value: 2
+
+  - it: backoffLimit is overridable
+    template: templates/webhook-ready-hook.yaml
+    set:
+      webhookReady:
+        backoffLimit: 5
+    asserts:
+      - documentIndex: 3
+        equal:
+          path: spec.backoffLimit
+          value: 5
+
+  - it: chart render fails when webhookReady.image.repository is empty
+    template: templates/webhook-ready-hook.yaml
+    set:
+      webhookReady:
+        image:
+          repository: ""
+          tag: v1.32
+    asserts:
+      - failedTemplate:
+          errorMessage: "webhookReady.image.repository must be set to the container image providing kubectl for the post-install readiness Job"
+
+  - it: chart render fails when webhookReady.image.tag is empty
+    template: templates/webhook-ready-hook.yaml
+    set:
+      webhookReady:
+        image:
+          repository: docker.io/clastix/kubectl
+          tag: ""
+    asserts:
+      - failedTemplate:
+          errorMessage: "webhookReady.image.tag must be set for the post-install readiness Job"
+
+  - it: retry loop bounds default when blanked so a wiped override still produces a working Job
+    template: templates/webhook-ready-hook.yaml
+    set:
+      webhookReady:
+        image:
+          repository: docker.io/clastix/kubectl
+          tag: v1.32
+        maxAttempts: null
+        sleepSeconds: null
+    asserts:
+      - documentIndex: 3
+        matchRegex:
+          path: spec.template.spec.containers[0].command[2]
+          pattern: "max_attempts=60"
+      - documentIndex: 3
+        matchRegex:
+          path: spec.template.spec.containers[0].command[2]
+          pattern: "sleep_seconds=2"
+
+  - it: Job pod runs non-root with seccomp RuntimeDefault for restricted-PSA clusters
+    template: templates/webhook-ready-hook.yaml
+    asserts:
+      - documentIndex: 3
+        equal:
+          path: spec.template.spec.securityContext.runAsNonRoot
+          value: true
+      - documentIndex: 3
+        equal:
+          path: spec.template.spec.securityContext.seccompProfile.type
+          value: RuntimeDefault
+
+  - it: wait container declares resource requests and limits for predictable scheduling
+    template: templates/webhook-ready-hook.yaml
+    asserts:
+      - documentIndex: 3
+        equal:
+          path: spec.template.spec.containers[0].resources.requests.cpu
+          value: 10m
+      - documentIndex: 3
+        equal:
+          path: spec.template.spec.containers[0].resources.requests.memory
+          value: 32Mi
+      - documentIndex: 3
+        equal:
+          path: spec.template.spec.containers[0].resources.limits.cpu
+          value: 100m
+      - documentIndex: 3
+        equal:
+          path: spec.template.spec.containers[0].resources.limits.memory
+          value: 64Mi
+
+  - it: Job container drops all capabilities and runs read-only rootfs
+    template: templates/webhook-ready-hook.yaml
+    asserts:
+      - documentIndex: 3
+        equal:
+          path: spec.template.spec.containers[0].securityContext.allowPrivilegeEscalation
+          value: false
+      - documentIndex: 3
+        equal:
+          path: spec.template.spec.containers[0].securityContext.readOnlyRootFilesystem
+          value: true
+      - documentIndex: 3
+        equal:
+          path: spec.template.spec.containers[0].securityContext.capabilities.drop[0]
+          value: ALL
+
+  - it: imagePullPolicy is IfNotPresent when digest-pinned, Always when tag-only
+    template: templates/webhook-ready-hook.yaml
+    set:
+      webhookReady:
+        image:
+          repository: docker.io/clastix/kubectl
+          tag: v1.32
+          digest: sha256:b9ef7d8dbe65bcc81a46c09b8dc7543103055021c4f43287bf59e92a8f4fe05c
+    asserts:
+      - documentIndex: 3
+        equal:
+          path: spec.template.spec.containers[0].imagePullPolicy
+          value: IfNotPresent
+
+  - it: imagePullPolicy is Always when no digest is configured
+    template: templates/webhook-ready-hook.yaml
+    set:
+      webhookReady:
+        image:
+          repository: docker.io/clastix/kubectl
+          tag: v1.32
+          digest: ""
+    asserts:
+      - documentIndex: 3
+        equal:
+          path: spec.template.spec.containers[0].imagePullPolicy
+          value: Always
+
+  - it: retry loop captures and surfaces the last kubectl error message on timeout
+    template: templates/webhook-ready-hook.yaml
+    asserts:
+      - documentIndex: 3
+        matchRegex:
+          path: spec.template.spec.containers[0].command[2]
+          pattern: 'last_err=\$\(kubectl get --raw'
+      - documentIndex: 3
+        matchRegex:
+          path: spec.template.spec.containers[0].command[2]
+          pattern: 'last error: \$\{last_err\}'
+
+  - it: retry loop error message stays in sync when maxAttempts is bumped
+    template: templates/webhook-ready-hook.yaml
+    set:
+      webhookReady:
+        image:
+          repository: docker.io/clastix/kubectl
+          tag: v1.32
+        maxAttempts: 90
+        sleepSeconds: 3
+    asserts:
+      - documentIndex: 3
+        matchRegex:
+          path: spec.template.spec.containers[0].command[2]
+          pattern: "max_attempts=90"
+      - documentIndex: 3
+        matchRegex:
+          path: spec.template.spec.containers[0].command[2]
+          pattern: "sleep_seconds=3"
+      - documentIndex: 3
+        matchRegex:
+          path: spec.template.spec.containers[0].command[2]
+          pattern: 'after \$\{max_attempts\} attempts'
+
+  - it: activeDeadlineSeconds scales with retry bounds so an override raise does not silently cut
+    template: templates/webhook-ready-hook.yaml
+    set:
+      webhookReady:
+        image:
+          repository: docker.io/clastix/kubectl
+          tag: v1.32
+        maxAttempts: 180
+        sleepSeconds: 3
+    asserts:
+      - documentIndex: 3
+        equal:
+          path: spec.activeDeadlineSeconds
+          value: 600
+
+  - it: activeDeadlineSeconds defaults include 60s slack over the default retry window
+    template: templates/webhook-ready-hook.yaml
+    asserts:
+      - documentIndex: 3
+        equal:
+          path: spec.activeDeadlineSeconds
+          value: 180
diff --git a/packages/system/postgres-operator/values.yaml b/packages/system/postgres-operator/values.yaml
index cae3dc53..c408220e 100644
--- a/packages/system/postgres-operator/values.yaml
+++ b/packages/system/postgres-operator/values.yaml
@@ -3,3 +3,30 @@ cloudnative-pg:
     create: true
   image:
     tag: "1.27.3"
+# Image used by the post-install webhook-readiness Job (see templates/webhook-ready-hook.yaml).
+# Any image with kubectl on PATH works; the Job calls the apiserver service proxy with the
+# hook ServiceAccount's token to confirm the mcluster.cnpg.io webhook is reachable end-to-end
+# before the HelmRelease reports Ready, closing the "connection refused" bootstrap race.
+#
+# The tag is digest-pinned so an upstream retag does not change what runs on every install
+# and upgrade across the fleet. Refresh by resolving the current manifest-list digest
+# (`docker manifest inspect docker.io/clastix/kubectl:v1.32`) and updating `digest` below.
+# renovate: datasource=docker depName=docker.io/clastix/kubectl
+webhookReady:
+  image:
+    repository: docker.io/clastix/kubectl
+    tag: v1.32
+    digest: sha256:b9ef7d8dbe65bcc81a46c09b8dc7543103055021c4f43287bf59e92a8f4fe05c
+  # Retry loop bounds for the readiness probe. Defaults total ~120s wall clock.
+  # Both are `default`-coerced in the template so an override that blanks them still
+  # produces a working Job.
+  maxAttempts: 60
+  sleepSeconds: 2
+  # Pod-level retry budget for transient node/registry failures (image pull rate limit,
+  # OOM, CNI hiccup). The shell loop inside the container only covers reachability retries.
+  # Pod retries and activeDeadlineSeconds (wall-clock bound on the whole Job across all
+  # pod retries) are ANDed, so activeDeadlineSeconds is the shorter of the two gates with
+  # the defaults above: the 60*2+60 = 180s deadline cuts the Job before backoffLimit=2
+  # ever matters if pod-level failures eat more than ~60s of the budget. Raise
+  # maxAttempts/sleepSeconds alongside backoffLimit when tuning for slow image pulls.
+  backoffLimit: 2
diff --git a/packages/system/postgres-rd/cozyrds/postgres.yaml b/packages/system/postgres-rd/cozyrds/postgres.yaml
index c74c5783..72d013f1 100644
--- a/packages/system/postgres-rd/cozyrds/postgres.yaml
+++ b/packages/system/postgres-rd/cozyrds/postgres.yaml
@@ -8,7 +8,7 @@ spec:
     singular: postgres
     plural: postgreses
     openAPISchema: |-
-      {"title":"Chart Values","type":"object","properties":{"replicas":{"description":"Number of Postgres replicas.","type":"integer","default":2},"resources":{"description":"Explicit CPU and memory configuration for each PostgreSQL replica. When omitted, the preset defined in `resourcesPreset` is applied.","type":"object","default":{},"properties":{"cpu":{"description":"CPU available to each replica.","pattern":"^(\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))(([KMGTPE]i)|[numkMGTPE]|([eE](\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))))?$","anyOf":[{"type":"integer"},{"type":"string"}],"x-kubernetes-int-or-string":true},"memory":{"description":"Memory (RAM) available to each replica.","pattern":"^(\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))(([KMGTPE]i)|[numkMGTPE]|([eE](\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))))?$","anyOf":[{"type":"integer"},{"type":"string"}],"x-kubernetes-int-or-string":true}}},"resourcesPreset":{"description":"Default sizing preset used when `resources` is omitted.","type":"string","default":"micro","enum":["nano","micro","small","medium","large","xlarge","2xlarge"]},"size":{"description":"Persistent Volume Claim size available for application data.","default":"10Gi","pattern":"^(\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))(([KMGTPE]i)|[numkMGTPE]|([eE](\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))))?$","anyOf":[{"type":"integer"},{"type":"string"}],"x-kubernetes-int-or-string":true},"storageClass":{"description":"StorageClass used to store the data.","type":"string","default":""},"external":{"description":"Enable external access from outside the cluster.","type":"boolean","default":false},"version":{"description":"PostgreSQL major version to deploy","type":"string","default":"v18","enum":["v18","v17","v16","v15","v14","v13"]},"postgresql":{"description":"PostgreSQL server configuration.","type":"object","default":{},"properties":{"parameters":{"description":"PostgreSQL server parameters.","type":"object","default":{},"properties":{"max_connections":{"description":"Maximum number of concurrent connections to the database server.","type":"integer","default":100}}}}},"quorum":{"description":"Quorum configuration for synchronous replication.","type":"object","default":{},"required":["maxSyncReplicas","minSyncReplicas"],"properties":{"maxSyncReplicas":{"description":"Maximum number of synchronous replicas allowed (must be less than total replicas).","type":"integer","default":0},"minSyncReplicas":{"description":"Minimum number of synchronous replicas required for commit.","type":"integer","default":0}}},"users":{"description":"Users configuration map.","type":"object","default":{},"additionalProperties":{"type":"object","properties":{"password":{"description":"Password for the user.","type":"string"},"replication":{"description":"Whether the user has replication privileges.","type":"boolean"}}}},"databases":{"description":"Databases configuration map.","type":"object","default":{},"additionalProperties":{"type":"object","properties":{"extensions":{"description":"List of enabled PostgreSQL extensions.","type":"array","items":{"type":"string"}},"roles":{"description":"Roles assigned to users.","type":"object","properties":{"admin":{"description":"List of users with admin privileges.","type":"array","items":{"type":"string"}},"readonly":{"description":"List of users with read-only privileges.","type":"array","items":{"type":"string"}}}}}}},"backup":{"description":"Backup configuration.","type":"object","default":{},"required":["enabled"],"properties":{"destinationPath":{"description":"Destination path for backups (e.g. s3://bucket/path/).","type":"string","default":"s3://bucket/path/to/folder/"},"enabled":{"description":"Enable regular backups.","type":"boolean","default":false},"endpointURL":{"description":"S3 endpoint URL for uploads.","type":"string","default":"http://minio-gateway-service:9000"},"retentionPolicy":{"description":"Retention policy (e.g. \"30d\").","type":"string","default":"30d"},"s3AccessKey":{"description":"Access key for S3 authentication.","type":"string","default":"<your-access-key>"},"s3SecretKey":{"description":"Secret key for S3 authentication.","type":"string","default":"<your-secret-key>"},"schedule":{"description":"Cron schedule for automated backups.","type":"string","default":"0 2 * * * *"}}},"bootstrap":{"description":"Bootstrap configuration.","type":"object","default":{},"required":["enabled","oldName"],"properties":{"enabled":{"description":"Whether to restore from a backup.","type":"boolean","default":false},"oldName":{"description":"Previous cluster name before deletion.","type":"string","default":""},"recoveryTime":{"description":"Timestamp (RFC3339) for point-in-time recovery; empty means latest.","type":"string","default":""}}}}}
+      {"title":"Chart Values","type":"object","properties":{"replicas":{"description":"Number of Postgres replicas.","type":"integer","default":2},"resources":{"description":"Explicit CPU and memory configuration for each PostgreSQL replica. When omitted, the preset defined in `resourcesPreset` is applied.","type":"object","default":{},"properties":{"cpu":{"description":"CPU available to each replica.","pattern":"^(\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))(([KMGTPE]i)|[numkMGTPE]|([eE](\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))))?$","anyOf":[{"type":"integer"},{"type":"string"}],"x-kubernetes-int-or-string":true},"memory":{"description":"Memory (RAM) available to each replica.","pattern":"^(\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))(([KMGTPE]i)|[numkMGTPE]|([eE](\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))))?$","anyOf":[{"type":"integer"},{"type":"string"}],"x-kubernetes-int-or-string":true}}},"resourcesPreset":{"description":"Default sizing preset used when `resources` is omitted.","type":"string","default":"micro","enum":["nano","micro","small","medium","large","xlarge","2xlarge"]},"size":{"description":"Persistent Volume Claim size available for application data.","default":"10Gi","pattern":"^(\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))(([KMGTPE]i)|[numkMGTPE]|([eE](\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))))?$","anyOf":[{"type":"integer"},{"type":"string"}],"x-kubernetes-int-or-string":true},"storageClass":{"description":"StorageClass used to store the data.","type":"string","default":""},"external":{"description":"Enable external access from outside the cluster.","type":"boolean","default":false},"version":{"description":"PostgreSQL major version to deploy","type":"string","default":"v18","enum":["v18","v17","v16","v15","v14","v13"]},"postgresql":{"description":"PostgreSQL server configuration.","type":"object","default":{},"properties":{"parameters":{"description":"PostgreSQL server parameters.","type":"object","default":{},"properties":{"max_connections":{"description":"Maximum number of concurrent connections to the database server.","type":"integer","default":100}}}}},"quorum":{"description":"Quorum configuration for synchronous replication.","type":"object","default":{},"required":["maxSyncReplicas","minSyncReplicas"],"properties":{"maxSyncReplicas":{"description":"Maximum number of synchronous replicas allowed (must be less than total replicas).","type":"integer","default":0},"minSyncReplicas":{"description":"Minimum number of synchronous replicas required for commit.","type":"integer","default":0}}},"users":{"description":"Users configuration map.","type":"object","default":{},"additionalProperties":{"type":"object","properties":{"password":{"description":"Password for the user.","type":"string"},"replication":{"description":"Whether the user has replication privileges.","type":"boolean"}}}},"databases":{"description":"Databases configuration map.","type":"object","default":{},"additionalProperties":{"type":"object","properties":{"extensions":{"description":"List of enabled PostgreSQL extensions.","type":"array","items":{"type":"string"}},"roles":{"description":"Roles assigned to users.","type":"object","properties":{"admin":{"description":"List of users with admin privileges.","type":"array","items":{"type":"string"}},"readonly":{"description":"List of users with read-only privileges.","type":"array","items":{"type":"string"}}}}}}},"backup":{"description":"Backup configuration.","type":"object","default":{},"required":["enabled"],"properties":{"destinationPath":{"description":"Destination path for backups (e.g. s3://bucket/path/).","type":"string","default":"s3://bucket/path/to/folder/"},"enabled":{"description":"Enable regular backups.","type":"boolean","default":false},"endpointURL":{"description":"S3 endpoint URL for uploads.","type":"string","default":"http://minio-gateway-service:9000"},"retentionPolicy":{"description":"Retention policy (e.g. \"30d\").","type":"string","default":"30d"},"s3AccessKey":{"description":"Access key for S3 authentication.","type":"string","default":"<your-access-key>"},"s3SecretKey":{"description":"Secret key for S3 authentication.","type":"string","default":"<your-secret-key>"},"schedule":{"description":"Cron schedule for automated backups.","type":"string","default":"0 2 * * * *"}}},"bootstrap":{"description":"Bootstrap configuration.","type":"object","default":{},"required":["enabled","oldName"],"properties":{"enabled":{"description":"Whether to restore from a backup.","type":"boolean","default":false},"oldName":{"description":"Previous cluster name before deletion.","type":"string","default":""},"recoveryTime":{"description":"Timestamp (RFC3339) for point-in-time recovery; empty means latest.","type":"string","default":""},"serverName":{"description":"Barman server name (S3 path prefix) used by the original cluster when writing backups. Set this only when the original cluster had an explicit barmanObjectStore.serverName that differed from its Kubernetes resource name.","type":"string","default":""}}}}}
   release:
     prefix: postgres-
     labels:
@@ -33,7 +33,7 @@ spec:
     #    labelSelector:
     #      helm.toolkit.fluxcd.io/name: "{reqs[0]['metadata','name']}"
 
-    keysOrder: [["apiVersion"], ["appVersion"], ["kind"], ["metadata"], ["metadata", "name"], ["spec", "replicas"], ["spec", "resources"], ["spec", "resourcesPreset"], ["spec", "size"], ["spec", "storageClass"], ["spec", "external"], ["spec", "version"], ["spec", "postgresql"], ["spec", "postgresql", "parameters"], ["spec", "postgresql", "parameters", "max_connections"], ["spec", "quorum"], ["spec", "quorum", "minSyncReplicas"], ["spec", "quorum", "maxSyncReplicas"], ["spec", "users"], ["spec", "databases"], ["spec", "backup"], ["spec", "backup", "enabled"], ["spec", "backup", "retentionPolicy"], ["spec", "backup", "destinationPath"], ["spec", "backup", "endpointURL"], ["spec", "backup", "schedule"], ["spec", "backup", "s3AccessKey"], ["spec", "backup", "s3SecretKey"], ["spec", "bootstrap"], ["spec", "bootstrap", "enabled"], ["spec", "bootstrap", "recoveryTime"], ["spec", "bootstrap", "oldName"]]
+    keysOrder: [["apiVersion"], ["appVersion"], ["kind"], ["metadata"], ["metadata", "name"], ["spec", "replicas"], ["spec", "resources"], ["spec", "resourcesPreset"], ["spec", "size"], ["spec", "storageClass"], ["spec", "external"], ["spec", "version"], ["spec", "postgresql"], ["spec", "postgresql", "parameters"], ["spec", "postgresql", "parameters", "max_connections"], ["spec", "quorum"], ["spec", "quorum", "minSyncReplicas"], ["spec", "quorum", "maxSyncReplicas"], ["spec", "users"], ["spec", "databases"], ["spec", "backup"], ["spec", "backup", "enabled"], ["spec", "backup", "retentionPolicy"], ["spec", "backup", "destinationPath"], ["spec", "backup", "endpointURL"], ["spec", "backup", "schedule"], ["spec", "backup", "s3AccessKey"], ["spec", "backup", "s3SecretKey"], ["spec", "bootstrap"], ["spec", "bootstrap", "enabled"], ["spec", "bootstrap", "recoveryTime"], ["spec", "bootstrap", "oldName"], ["spec", "bootstrap", "serverName"]]
   secrets:
     exclude: []
     include:
diff --git a/packages/system/seaweedfs/templates/database.yaml b/packages/system/seaweedfs/templates/database.yaml
index 1c116dd0..b9b50c49 100644
--- a/packages/system/seaweedfs/templates/database.yaml
+++ b/packages/system/seaweedfs/templates/database.yaml
@@ -5,7 +5,9 @@ metadata:
   name: seaweedfs-db
 spec:
   instances: {{ .Values.db.replicas }}
-  imageName: ghcr.io/cloudnative-pg/postgresql:17.7
+  {{- $existingCluster := lookup "postgresql.cnpg.io/v1" "Cluster" .Release.Namespace "seaweedfs-db" }}
+  {{- $image := dig "spec" "imageName" "ghcr.io/cloudnative-pg/postgresql:17.7-standard-trixie" $existingCluster }}
+  imageName: {{ if regexMatch ":17\\." $image }}ghcr.io/cloudnative-pg/postgresql:17.7-standard-trixie{{ else if regexMatch ":[0-9]+\\.[0-9]+$" $image }}{{ printf "%s-standard-trixie" $image }}{{ else }}{{ $image }}{{ end }}
   storage:
     size: {{ .Values.db.size }}
     {{- with .Values.db.storageClass }}
diff --git a/packages/system/seaweedfs/values.yaml b/packages/system/seaweedfs/values.yaml
index 7d598649..3542a77f 100644
--- a/packages/system/seaweedfs/values.yaml
+++ b/packages/system/seaweedfs/values.yaml
@@ -111,7 +111,7 @@ seaweedfs:
         nginx.ingress.kubernetes.io/client-body-timeout: "3600"
         nginx.ingress.kubernetes.io/client-header-timeout: "120"
         nginx.ingress.kubernetes.io/service-upstream: "true"
-        acme.cert-manager.io/http01-ingress-class: tenant-root
+        acme.cert-manager.io/http01-ingress-ingressclassname: tenant-root
         cert-manager.io/cluster-issuer: letsencrypt-prod
       tls:
         - hosts:
@@ -177,7 +177,7 @@ seaweedfs:
     bucketClassName: "seaweedfs"
     region: ""
     sidecar:
-      image: "ghcr.io/cozystack/cozystack/objectstorage-sidecar:v1.2.1@sha256:1ad07fb9e96477e5c322f240f6023ceedd8762261d8379ec784dde92797ee206"
+      image: "ghcr.io/cozystack/cozystack/objectstorage-sidecar:v1.3.0@sha256:5bcccbdb13979a16cee535eb5fbcdf0d87973689010a10b7b25e55c5be3edaa6"
   certificates:
     commonName: "SeaweedFS CA"
     ipAddresses: []
diff --git a/packages/system/velero/templates/backupstrategy-controller-rbac.yaml b/packages/system/velero/templates/backupstrategy-controller-rbac.yaml
new file mode 100644
index 00000000..43c51cd7
--- /dev/null
+++ b/packages/system/velero/templates/backupstrategy-controller-rbac.yaml
@@ -0,0 +1,26 @@
+# Grants the backupstrategy-controller permission to manage ResourceModifier
+# ConfigMaps in the Velero install namespace. Lives here (not in the
+# backupstrategy-controller chart) so that the Role is only created when
+# velero is actually installed — otherwise the chart would try to create it
+# in a namespace that does not exist.
+apiVersion: rbac.authorization.k8s.io/v1
+kind: Role
+metadata:
+  name: backups.cozystack.io:strategy-controller:velero-configmaps
+rules:
+- apiGroups: [""]
+  resources: ["configmaps"]
+  verbs: ["create", "get", "list", "delete", "deletecollection", "patch", "update"]
+---
+apiVersion: rbac.authorization.k8s.io/v1
+kind: RoleBinding
+metadata:
+  name: backups.cozystack.io:strategy-controller:velero-configmaps
+roleRef:
+  apiGroup: rbac.authorization.k8s.io
+  kind: Role
+  name: backups.cozystack.io:strategy-controller:velero-configmaps
+subjects:
+- kind: ServiceAccount
+  name: backupstrategy-controller
+  namespace: cozy-backup-controller
diff --git a/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/Chart.lock b/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/Chart.lock
index 1b0e2a46..2868e66f 100644
--- a/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/Chart.lock
+++ b/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/Chart.lock
@@ -1,9 +1,9 @@
 dependencies:
 - name: victoria-metrics-common
-  repository: https://victoriametrics.github.io/helm-charts
-  version: 0.0.46
+  repository: oci://ghcr.io/victoriametrics/helm-charts
+  version: 0.3.0
 - name: crds
   repository: ""
   version: 0.0.*
-digest: sha256:43d3d210a9d1a2234e6c56518f8c477125a5ad5e8ed08d46209528f19acd9c89
-generated: "2025-12-23T12:58:01.428960334Z"
+digest: sha256:adec954bf2814f744f5b69fcf67c8b246bf21f416adb0ab37d91c9b42f72d353
+generated: "2026-04-16T10:14:10.795552631Z"
diff --git a/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/Chart.yaml b/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/Chart.yaml
index bb9253b1..030af8f0 100644
--- a/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/Chart.yaml
+++ b/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/Chart.yaml
@@ -1,7 +1,7 @@
 annotations:
   artifacthub.io/category: monitoring-logging
   artifacthub.io/changes: |
-    - updates operator to [v0.68.1](https://github.com/VictoriaMetrics/operator/releases/tag/v0.68.1) version
+    - revert change in Deployment's matchLabels, that was introduced in release 0.60.0
   artifacthub.io/license: Apache-2.0
   artifacthub.io/links: |
     - name: Sources
@@ -16,13 +16,14 @@ annotations:
   artifacthub.io/readme: |
     # VictoriaMetrics Operator Helm chart
 
-    Chart documentation is available [here](https://docs.victoriametrics.com/helm/victoria-metrics-operator/)
+    Chart documentation is available [here](https://docs.victoriametrics.com/helm/victoria-metrics-operator/).
+    Changelog is [here](https://docs.victoriametrics.com/helm/victoria-metrics-operator/changelog/).
 apiVersion: v2
-appVersion: v0.68.1
+appVersion: v0.68.4
 dependencies:
 - name: victoria-metrics-common
-  repository: https://victoriametrics.github.io/helm-charts
-  version: 0.0.*
+  repository: oci://ghcr.io/victoriametrics/helm-charts
+  version: 0.3.*
 - condition: crds.plain
   name: crds
   repository: ""
@@ -46,4 +47,4 @@ sources:
 - https://github.com/VictoriaMetrics/helm-charts
 - https://github.com/VictoriaMetrics/operator
 type: application
-version: 0.59.1
+version: 0.61.0
diff --git a/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/README.md b/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/README.md
index 49a06de2..6a080424 100644
--- a/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/README.md
+++ b/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/README.md
@@ -1,3 +1,4 @@
 # VictoriaMetrics Operator Helm chart
 
-Chart documentation is available [here](https://docs.victoriametrics.com/helm/victoria-metrics-operator/)
+Chart documentation is available [here](https://docs.victoriametrics.com/helm/victoria-metrics-operator/).
+Changelog is [here](https://docs.victoriametrics.com/helm/victoria-metrics-operator/changelog/).
diff --git a/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/RELEASE_NOTES b/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/RELEASE_NOTES
index bf22fd3f..3529783a 100644
--- a/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/RELEASE_NOTES
+++ b/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/RELEASE_NOTES
@@ -1,7 +1,7 @@
-# Release notes for version 0.59.1
+# Release notes for version 0.61.0
 
-**Release date:** 03 Mar 2026
+**Release date:** 16 Apr 2026
 
-![Helm: v3](https://img.shields.io/badge/Helm-v3.14%2B-informational?color=informational&logo=helm&link=https%3A%2F%2Fgithub.com%2Fhelm%2Fhelm%2Freleases%2Ftag%2Fv3.14.0) ![AppVersion: v0.68.1](https://img.shields.io/badge/v0.68.1-success?logo=VictoriaMetrics&labelColor=gray&link=https%3A%2F%2Fdocs.victoriametrics.com%2Foperator%2Fchangelog%2F%23v0681)
+![Helm: v3](https://img.shields.io/badge/Helm-v3.14%2B-informational?color=informational&logo=helm&link=https%3A%2F%2Fgithub.com%2Fhelm%2Fhelm%2Freleases%2Ftag%2Fv3.14.0) ![AppVersion: v0.68.4](https://img.shields.io/badge/v0.68.4-success?logo=VictoriaMetrics&labelColor=gray&link=https%3A%2F%2Fdocs.victoriametrics.com%2Foperator%2Fchangelog%2F%23v0684)
 
-- updates operator to [v0.68.1](https://github.com/VictoriaMetrics/operator/releases/tag/v0.68.1) version
+- revert change in Deployment's matchLabels, that was introduced in release 0.60.0
diff --git a/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/charts/crds/crds/crd.yaml b/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/charts/crds/crds/crd.yaml
index c4138be8..bdf7e8d0 100644
--- a/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/charts/crds/crds/crd.yaml
+++ b/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/charts/crds/crds/crd.yaml
@@ -6079,6 +6079,7 @@ spec:
                 - spec
                 type: object
               shardCount:
+                format: int32
                 type: integer
               startupProbe:
                 type: object
@@ -6087,6 +6088,14 @@ spec:
                 type: boolean
               statefulRollingUpdateStrategy:
                 type: string
+              statefulRollingUpdateStrategyBehavior:
+                properties:
+                  maxUnavailable:
+                    anyOf:
+                    - type: integer
+                    - type: string
+                    x-kubernetes-int-or-string: true
+                type: object
               statefulStorage:
                 properties:
                   emptyDir:
@@ -12326,6 +12335,7 @@ spec:
                 type: object
                 x-kubernetes-preserve-unknown-fields: true
               shardCount:
+                format: int32
                 type: integer
               startupProbe:
                 type: object
@@ -16808,6 +16818,9 @@ spec:
                 type: boolean
               vmauth:
                 properties:
+                  enabled:
+                    default: true
+                    type: boolean
                   name:
                     type: string
                   spec:
@@ -24727,6 +24740,14 @@ spec:
                             type: boolean
                           statefulRollingUpdateStrategy:
                             type: string
+                          statefulRollingUpdateStrategyBehavior:
+                            properties:
+                              maxUnavailable:
+                                anyOf:
+                                - type: integer
+                                - type: string
+                                x-kubernetes-int-or-string: true
+                            type: object
                           statefulStorage:
                             properties:
                               emptyDir:
@@ -28628,6 +28649,14 @@ spec:
                               type: boolean
                             statefulRollingUpdateStrategy:
                               type: string
+                            statefulRollingUpdateStrategyBehavior:
+                              properties:
+                                maxUnavailable:
+                                  anyOf:
+                                  - type: integer
+                                  - type: string
+                                  x-kubernetes-int-or-string: true
+                              type: object
                             statefulStorage:
                               properties:
                                 emptyDir:
diff --git a/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/charts/crds/files/crd.yaml.bz2 b/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/charts/crds/files/crd.yaml.bz2
index 7f518a91..ab7c4d95 100644
Binary files a/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/charts/crds/files/crd.yaml.bz2 and b/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/charts/crds/files/crd.yaml.bz2 differ
diff --git a/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/charts/victoria-metrics-common/Chart.yaml b/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/charts/victoria-metrics-common/Chart.yaml
index 04aba91f..e8f67b95 100644
--- a/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/charts/victoria-metrics-common/Chart.yaml
+++ b/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/charts/victoria-metrics-common/Chart.yaml
@@ -1,7 +1,7 @@
 annotations:
   artifacthub.io/category: monitoring-logging
   artifacthub.io/changes: |
-    - allow setting global labels for each resource using `.Values.global.extraLabels`. See [#2607](https://github.com/VictoriaMetrics/helm-charts/issues/2607).
+    - reverted usage of `app` label in `vm.selectorLabels`
   artifacthub.io/license: Apache-2.0
   artifacthub.io/links: |
     - name: Sources
@@ -11,7 +11,8 @@ annotations:
   artifacthub.io/readme: |
     # VictoriaMetrics Common Helm chart
 
-    Chart documentation is available [here](https://docs.victoriametrics.com/helm/victoria-metrics-common/)
+    Chart documentation is available [here](https://docs.victoriametrics.com/helm/victoria-metrics-common/).
+    Changelog is [here](https://docs.victoriametrics.com/helm/victoria-metrics-common/changelog/).
 apiVersion: v2
 description: VictoriaMetrics Common - contains shared templates for all Victoria Metrics
   helm charts
@@ -29,4 +30,4 @@ name: victoria-metrics-common
 sources:
 - https://github.com/VictoriaMetrics/helm-charts
 type: library
-version: 0.0.46
+version: 0.3.0
diff --git a/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/charts/victoria-metrics-common/README.md b/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/charts/victoria-metrics-common/README.md
index 9ba0b4f5..c8ee7051 100644
--- a/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/charts/victoria-metrics-common/README.md
+++ b/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/charts/victoria-metrics-common/README.md
@@ -1,3 +1,4 @@
 # VictoriaMetrics Common Helm chart
 
-Chart documentation is available [here](https://docs.victoriametrics.com/helm/victoria-metrics-common/)
+Chart documentation is available [here](https://docs.victoriametrics.com/helm/victoria-metrics-common/).
+Changelog is [here](https://docs.victoriametrics.com/helm/victoria-metrics-common/changelog/).
diff --git a/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/charts/victoria-metrics-common/RELEASE_NOTES b/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/charts/victoria-metrics-common/RELEASE_NOTES
index 055cd4c5..ad0cdfa9 100644
--- a/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/charts/victoria-metrics-common/RELEASE_NOTES
+++ b/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/charts/victoria-metrics-common/RELEASE_NOTES
@@ -1,7 +1,7 @@
-# Release notes for version 0.0.46
+# Release notes for version 0.3.0
 
-**Release date:** 23 Dec 2025
+**Release date:** 16 Apr 2026
 
 ![Helm: v3](https://img.shields.io/badge/Helm-v3.14%2B-informational?color=informational&logo=helm&link=https%3A%2F%2Fgithub.com%2Fhelm%2Fhelm%2Freleases%2Ftag%2Fv3.14.0)
 
-- allow setting global labels for each resource using `.Values.global.extraLabels`. See [#2607](https://github.com/VictoriaMetrics/helm-charts/issues/2607).
+- reverted usage of `app` label in `vm.selectorLabels`
diff --git a/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/charts/victoria-metrics-common/templates/_helpers.tpl b/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/charts/victoria-metrics-common/templates/_helpers.tpl
index 99ac30d5..bd945678 100644
--- a/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/charts/victoria-metrics-common/templates/_helpers.tpl
+++ b/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/charts/victoria-metrics-common/templates/_helpers.tpl
@@ -125,8 +125,8 @@ If release name contains chart name it will be used as a full name.
     {{- $ctx := . -}}
     {{- $values := $Values -}}
     {{- range $ak := $appKey }}
-      {{- $values = ternary (default dict) (index $values $ak | default dict) (empty $values) -}}
-      {{- $ctx = ternary (default dict) (index $ctx $ak | default dict) (empty $ctx) -}}
+      {{- $values = ternary (dict) (index $values $ak | default dict) (empty $values) -}}
+      {{- $ctx = ternary (dict) (index $ctx $ak | default dict) (empty $ctx) -}}
       {{- if and (empty $values) (empty $ctx) -}}
         {{- fail (printf "No data for appKey %s" (join "->" $appKey)) -}}
       {{- end -}}
@@ -187,6 +187,9 @@ If release name contains chart name it will be used as a full name.
   {{- include "vm.validate.args" . -}}
   {{- $Release := (.helm).Release | default .Release -}}
   {{- $labels := fromYaml (include "vm.selectorLabels" .) -}}
+  {{- with $labels.app -}}
+    {{- $_ := set $labels "app.kubernetes.io/component" . -}}
+  {{- end -}}
   {{- $labels = mergeOverwrite $labels (.extraLabels | default dict) -}}
   {{- $_ := set $labels "app.kubernetes.io/managed-by" $Release.Service -}}
   {{- toYaml $labels -}}
@@ -197,7 +200,7 @@ If release name contains chart name it will be used as a full name.
   {{- include "vm.validate.args" . -}}
   {{- $Values := (.helm).Values | default .Values -}}
   {{- $globalLabels := deepCopy (($Values.global).extraLabels | default dict) -}}
-  {{- $labels := fromYaml (include "vm.selectorLabels" .) -}}
+  {{- $labels := fromYaml (include "vm.commonLabels" .) -}}
   {{- $labels = mergeOverwrite $globalLabels $labels (fromYaml (include "vm.metaLabels" .)) -}}
   {{- with (include "vm.image.tag" .) }}
     {{- $_ := set $labels "app.kubernetes.io/version" (regexReplaceAll "(.*)(@sha.*)" . "${1}") -}}
@@ -231,11 +234,16 @@ If release name contains chart name it will be used as a full name.
   {{- $_ := set $labels "app.kubernetes.io/name" (include "vm.name" .) -}}
   {{- $_ := set $labels "app.kubernetes.io/instance" (include "vm.release" .) -}}
   {{- with (include "vm.app.name" .) -}}
-    {{- if eq $.style "managed" -}}
-      {{- $_ := set $labels "app.kubernetes.io/component" (printf "%s-%s" (include "vm.name" $) .) -}}
-    {{- else -}}
-      {{- $_ := set $labels "app" . -}}
-    {{- end -}}
+    {{- $_ := set $labels "app" . -}}
   {{- end -}}
   {{- toYaml $labels -}}
 {{- end }}
+
+{{- define "vm.commonLabels" -}}
+  {{- $labels := fromYaml (include "vm.selectorLabels" . ) -}}
+  {{- with $labels.app -}}
+    {{- $_ := set $labels "app.kubernetes.io/component" . -}}
+    {{- $_ := unset $labels "app" -}}
+  {{- end -}}
+  {{- toYaml $labels -}}
+{{- end -}}
diff --git a/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/charts/victoria-metrics-common/templates/_image.tpl b/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/charts/victoria-metrics-common/templates/_image.tpl
index 6d30bd03..66d019e3 100644
--- a/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/charts/victoria-metrics-common/templates/_image.tpl
+++ b/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/charts/victoria-metrics-common/templates/_image.tpl
@@ -42,8 +42,8 @@ Victoria Metrics Image
   {{- with .appKey -}}
     {{- $appKey := ternary (list .) . (kindIs "string" .) -}}
     {{- range $ak := $appKey -}}
-      {{- $values = ternary (default dict) (index $values $ak | default dict) (empty $values) -}}
-      {{- $ctx = ternary (default dict) (index $ctx $ak | default dict) (empty $ctx) -}}
+      {{- $values = ternary (dict) (index $values $ak | default dict) (empty $values) -}}
+      {{- $ctx = ternary (dict) (index $ctx $ak | default dict) (empty $ctx) -}}
       {{- if and (empty $values) (empty $ctx) -}}
         {{- fail (printf "No data for appKey %s" (join "->" $appKey)) -}}
       {{- end -}}
diff --git a/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/charts/victoria-metrics-common/templates/_pod.tpl b/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/charts/victoria-metrics-common/templates/_pod.tpl
index 7534ae2d..62981dfb 100644
--- a/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/charts/victoria-metrics-common/templates/_pod.tpl
+++ b/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/charts/victoria-metrics-common/templates/_pod.tpl
@@ -39,9 +39,9 @@ Render probe
 {{- define "vm.probe" -}}
   {{- /* undefined value */ -}}
   {{- $null := (fromYaml "value: null").value -}}
-  {{- $probe := dig .type (default dict) .app.probe -}}
+  {{- $probe := dig .type (dict) .app.probe -}}
   {{- $probeType := "" -}}
-  {{- $defaultProbe := default dict -}}
+  {{- $defaultProbe := dict -}}
   {{- if ne (dig "httpGet" $null $probe) $null -}}
     {{- /* httpGet probe */ -}}
     {{- $defaultProbe = dict "path" (include "vm.probe.http.path" .) "scheme" (include "vm.probe.http.scheme" .) "port" (include "vm.probe.port" .) -}}
@@ -51,7 +51,7 @@ Render probe
     {{- $defaultProbe = dict "port" (include "vm.probe.port" .) -}}
     {{- $probeType = "tcpSocket" -}}
   {{- end -}}
-  {{- $defaultProbe = ternary (default dict) (dict $probeType $defaultProbe) (empty $probeType) -}}
+  {{- $defaultProbe = ternary (dict) (dict $probeType $defaultProbe) (empty $probeType) -}}
   {{- $probe = mergeOverwrite $defaultProbe $probe -}}
   {{- range $key, $value := $probe -}}
     {{- if and (has (kindOf $value) (list "object" "map")) (ne $key $probeType) -}}
@@ -100,7 +100,7 @@ Net probe port
 command line arguments
 */ -}}
 {{- define "vm.args" -}}
-  {{- $args := default list -}}
+  {{- $args := list -}}
   {{- range $key, $value := . -}}
     {{- if not $key -}}
       {{- fail "Empty key in command line args is not allowed" -}}
diff --git a/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/charts/victoria-metrics-common/templates/_service.tpl b/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/charts/victoria-metrics-common/templates/_service.tpl
index 77a1365f..77cdeee7 100644
--- a/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/charts/victoria-metrics-common/templates/_service.tpl
+++ b/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/charts/victoria-metrics-common/templates/_service.tpl
@@ -37,10 +37,10 @@
     {{- $values := $Values -}}
     {{- $ctx := . -}}
     {{- range $ak := $appKey -}}
-      {{- $values = ternary (default dict) (index $values $ak | default dict) (empty $values) -}}
-      {{- $ctx = ternary (default dict) (index $ctx $ak | default dict) (empty $ctx) -}}
+      {{- $values = ternary (dict) (index $values $ak | default dict) (empty $values) -}}
+      {{- $ctx = ternary (dict) (index $ctx $ak | default dict) (empty $ctx) -}}
     {{- end -}}
-    {{- $spec := default dict -}}
+    {{- $spec := dict -}}
     {{- if $ctx -}}
       {{- $spec = $ctx -}}
     {{- else if $values -}}
@@ -69,10 +69,10 @@
     {{- $values := $Values -}}
     {{- $ctx := . -}}
     {{- range $ak := $appKey -}}
-      {{- $values = ternary (default dict) (index $values $ak | default dict) (empty $values) -}}
-      {{- $ctx = ternary (default dict) (index $ctx $ak | default dict) (empty $ctx) -}}
+      {{- $values = ternary (dict) (index $values $ak | default dict) (empty $values) -}}
+      {{- $ctx = ternary (dict) (index $ctx $ak | default dict) (empty $ctx) -}}
     {{- end -}}
-    {{- $spec := default dict -}}
+    {{- $spec := dict -}}
     {{- if $values -}}
       {{- $spec = $values -}}
     {{- else if $ctx -}}
diff --git a/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/templates/NOTES.txt b/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/templates/NOTES.txt
index 7fb3fbd9..41f39937 100644
--- a/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/templates/NOTES.txt
+++ b/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/templates/NOTES.txt
@@ -1,5 +1,5 @@
 {{ include "vm.name" . }} has been installed. Check its status by running:
-  kubectl --namespace {{ include "vm.namespace" . }} get pods -l "app.kubernetes.io/instance={{ $.Release.Name }}"
+  kubectl --namespace {{ include "vm.namespace" . }} get pods -l "app.kubernetes.io/name={{ include "vm.name" . }}" -l "app.kubernetes.io/instance={{ include "vm.release" . }}"
 
 Get more information on https://github.com/VictoriaMetrics/helm-charts/tree/master/charts/victoria-metrics-operator.
 See "Getting started guide for VM Operator" on https://docs.victoriametrics.com/guides/getting-started-with-vm-operator
diff --git a/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/templates/pdb.yaml b/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/templates/pdb.yaml
index 8abd9dea..9ec3c2c2 100644
--- a/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/templates/pdb.yaml
+++ b/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/templates/pdb.yaml
@@ -18,6 +18,9 @@ spec:
   {{- with $pdb.maxUnavailable }}
   maxUnavailable: {{ . }}
   {{- end }}
+  {{- with $pdb.unhealthyPodEvictionPolicy }}
+  unhealthyPodEvictionPolicy: {{ . }}
+  {{- end }}
   selector:
     matchLabels: {{ include "vm.selectorLabels" $ctx | nindent 6 }}
 {{- end }}
diff --git a/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/templates/server.yaml b/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/templates/server.yaml
index e224c074..9d7fd7ed 100644
--- a/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/templates/server.yaml
+++ b/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/templates/server.yaml
@@ -18,9 +18,12 @@ metadata:
   annotations: {{ toYaml . | nindent 4 }}
   {{- end }}
 spec:
-  replicas: {{.Values.replicaCount }}
+  replicas: {{ .Values.replicaCount }}
   selector:
     matchLabels: {{ include "vm.selectorLabels" $ctx | nindent 6 }}
+  {{- with .Values.strategy }}
+  strategy: {{ toYaml . | nindent 4 }}
+  {{- end }}
   template:
     metadata:
       {{- with .Values.annotations }}
diff --git a/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/templates/webhook.yaml b/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/templates/webhook.yaml
index 51039cae..e38f7448 100644
--- a/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/templates/webhook.yaml
+++ b/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/templates/webhook.yaml
@@ -1,22 +1,31 @@
-{{- if .Values.admissionWebhooks.enabled }}
+{{- $webhooks := .Values.admissionWebhooks }}
+{{- if $webhooks.enabled }}
 {{- $ctx := dict "helm" . "extraLabels" .Values.extraLabels }}
 {{- $tls := fromYaml (include "vm-operator.certs" $ctx) }}
 {{- $fullname := include "vm.plain.fullname" $ctx }}
 {{- $domain := ((.Values.global).cluster).dnsDomain }}
 {{- $ns := include "vm.namespace" $ctx }}
-{{- $certManager := .Values.admissionWebhooks.certManager }}
+{{- $certManager := $webhooks.certManager }}
 {{- $files := .Files }}
 {{- $crds := $files.Get "crd.yaml" | splitList "---" }}
-{{- $disabledFor := .Values.admissionWebhooks.disabledFor | default list }}
+{{- $disabledFor := $webhooks.disabledFor | default list }}
 ---
 apiVersion: admissionregistration.k8s.io/v1
 kind: ValidatingWebhookConfiguration
 metadata:
   name:  {{ $fullname }}-admission
-  {{- if $certManager.enabled }}
+  {{- if or $certManager.enabled $webhooks.annotations }}
   annotations:
+    {{- if $certManager.enabled }}
     certmanager.k8s.io/inject-ca-from: {{ printf "%s/%s-validation" $ns $fullname | quote }}
     cert-manager.io/inject-ca-from: {{ printf "%s/%s-validation" $ns $fullname | quote }}
+    {{- end }}
+    {{- with $certManager.annotations }}
+    {{- toYaml . | nindent 4 }}
+    {{- end }}
+    {{- with $webhooks.annotations }}
+    {{- toYaml . | nindent 4 }}
+    {{- end }}
   {{- end }}
   labels: {{ include "vm.labels" $ctx | nindent 4 }}
 webhooks:
@@ -34,7 +43,7 @@ webhooks:
       {{- if not $certManager.enabled }}
       caBundle: {{ $tls.caCert }}
       {{- end }}
-    failurePolicy: {{ $.Values.admissionWebhooks.policy }}
+    failurePolicy: {{ $webhooks.policy }}
     name: '{{ $crd.metadata.name }}'
     admissionReviewVersions:
       - v1
diff --git a/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/values.yaml b/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/values.yaml
index 279ceee9..3a1027f2 100644
--- a/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/values.yaml
+++ b/packages/system/victoria-metrics-operator/charts/victoria-metrics-operator/values.yaml
@@ -125,7 +125,7 @@ crds:
     #   whenUnsatisfiable: DoNotSchedule
     #   labelSelector:
     #     matchLabels:
-    #       app: alertmanager
+    #       app.kubernetes.io/component: alertmanager
 
     # -- Labels to add to the upgrade job
     labels: {}
@@ -276,8 +276,12 @@ service:
 # -- See `kubectl explain poddisruptionbudget.spec` for more or check [these docs](https://kubernetes.io/docs/tasks/run-application/configure-pdb/)
 podDisruptionBudget:
   enabled: false
-  # minAvailable: 1
-  # maxUnavailable: 1
+  # -- min number or percentage of pods that can be unavailable
+  minAvailable: 0
+  # -- max number or percentage of pods that can be unavailable
+  maxUnavailable: 0
+  # -- Defines criteria when unhealthy pods should be considered for eviction
+  unhealthyPodEvictionPolicy:
   labels: {}
 
 # -- Graceful pod termination timeout. See [this article](https://kubernetes.io/docs/concepts/containers/container-lifecycle-hooks/#hook-handler-execution) for details.
@@ -299,6 +303,13 @@ resources:
 # -- Pod's node selector. Details are [here](https://kubernetes.io/docs/concepts/scheduling-eviction/assign-pod-node/#nodeselector)
 nodeSelector: {}
 
+# -- Deployment strategy, set to standard k8s default
+strategy:
+  type: RollingUpdate
+  rollingUpdate:
+    maxSurge: 25%
+    maxUnavailable: 25%
+
 # -- Name of Priority Class
 priorityClassName: ""
 
@@ -364,6 +375,8 @@ shareProcessNamespace: false
 admissionWebhooks:
   # -- Enables validation webhook.
   enabled: true
+  # -- Annotations for webhook. Can be used to define Helm or ArgoCD annotations.
+  annotations: {}
   # -- List of CRD names to disable validation for
   disabledFor: []
   # - vmagent
diff --git a/packages/system/vm-default-images/Chart.yaml b/packages/system/vm-default-images/Chart.yaml
new file mode 100644
index 00000000..aac0ccca
--- /dev/null
+++ b/packages/system/vm-default-images/Chart.yaml
@@ -0,0 +1,5 @@
+apiVersion: v2
+name: vm-default-images
+description: Global Golden Image collection for virtual machines
+type: application
+version: 0.0.0 # Placeholder, the actual version will be automatically set during the build process
diff --git a/packages/system/vm-default-images/Makefile b/packages/system/vm-default-images/Makefile
new file mode 100644
index 00000000..a3f9fac7
--- /dev/null
+++ b/packages/system/vm-default-images/Makefile
@@ -0,0 +1,4 @@
+export NAME=vm-default-images
+export NAMESPACE=cozy-system
+
+include ../../../hack/package.mk
diff --git a/packages/system/vm-default-images/templates/dv.yaml b/packages/system/vm-default-images/templates/dv.yaml
new file mode 100644
index 00000000..84b9a758
--- /dev/null
+++ b/packages/system/vm-default-images/templates/dv.yaml
@@ -0,0 +1,45 @@
+{{- range .Values.images }}
+---
+apiVersion: cdi.kubevirt.io/v1beta1
+kind: DataVolume
+metadata:
+  annotations:
+    vm-default-images.cozystack.io/name: {{ .name | quote }}
+    {{- with .os }}
+    {{- with .family }}
+    vm-default-images.cozystack.io/os-family: {{ . | quote }}
+    {{- end }}
+    {{- with .name }}
+    vm-default-images.cozystack.io/os-name: {{ . | quote }}
+    {{- end }}
+    {{- with .version }}
+    vm-default-images.cozystack.io/os-version: {{ . | quote }}
+    {{- end }}
+    {{- end }}
+    {{- with .architecture }}
+    vm-default-images.cozystack.io/architecture: {{ . | quote }}
+    {{- end }}
+    {{- with .description }}
+    vm-default-images.cozystack.io/description: {{ . | quote }}
+    {{- end }}
+    cdi.kubevirt.io/storage.bind.immediate.requested: "true"
+    cdi.kubevirt.io/storage.usePopulator: "true"
+  labels:
+    app.kubernetes.io/managed-by: cozystack
+  name: vm-default-images-{{ required "A valid .name entry required for each image!" .name }}
+  namespace: cozy-public
+spec:
+  contentType: kubevirt
+  source:
+    http:
+      url: {{ required (printf "A valid .url entry required for image '%s'!" .name) .url | quote }}
+  storage:
+    resources:
+      requests:
+        storage: {{ default "5Gi" .storage }}
+    {{- if .storageClass }}
+    storageClassName: {{ .storageClass }}
+    {{- else if $.Values.storageClass }}
+    storageClassName: {{ $.Values.storageClass }}
+    {{- end }}
+{{- end }}
diff --git a/packages/system/vm-default-images/values.yaml b/packages/system/vm-default-images/values.yaml
new file mode 100644
index 00000000..9d777f0b
--- /dev/null
+++ b/packages/system/vm-default-images/values.yaml
@@ -0,0 +1,169 @@
+##
+## @section Global Golden Image Collection
+##
+
+## @param {string} storageClass - Default StorageClass for all images. If empty, uses the cluster default StorageClass.
+## NOTE: The default set of images requires approximately 320Gi of storage (16 images × 20Gi each).
+## Adjust the image list or per-image storage sizes to match your cluster capacity.
+storageClass: "replicated"
+
+## @typedef {struct} ImageOS - Operating system metadata for a Golden Image.
+## @field {string} [family] - OS family (e.g. "Linux", "Windows").
+## @field {string} [name] - OS distribution name (e.g. "Ubuntu", "Fedora").
+## @field {string} [version] - OS version (e.g. "24.04", "40").
+
+## @typedef {struct} GlobalImage - A global Golden Image entry.
+## @field {string} name - Unique image name used to reference this image in vm-disk (e.g. "ubuntu").
+## @field {string} url - HTTP(S) URL to download the image from.
+## @field {quantity} storage - Storage size to allocate for this image.
+## @field {string} storageClass - StorageClass used to store the image (overrides global default).
+## @field {ImageOS} [os] - Operating system metadata.
+## @field {string} [architecture] - CPU architecture (e.g. "amd64", "arm64").
+## @field {string} [description] - Human-readable description.
+
+## @param {[]GlobalImage} images - List of global Golden Images to provision in the cozy-public namespace.
+images:
+  - name: ubuntu-22.04
+    url: https://cloud-images.ubuntu.com/jammy/current/jammy-server-cloudimg-amd64.img
+    storage: 20Gi
+    os:
+      family: Linux
+      name: Ubuntu
+      version: "22.04"
+    architecture: amd64
+    description: "Ubuntu 22.04 LTS (Jammy Jellyfish) cloud image"
+  - name: ubuntu-24.04
+    url: https://cloud-images.ubuntu.com/noble/current/noble-server-cloudimg-amd64.img
+    storage: 20Gi
+    os:
+      family: Linux
+      name: Ubuntu
+      version: "24.04"
+    architecture: amd64
+    description: "Ubuntu 24.04 LTS (Noble Numbat) cloud image"
+  - name: rocky-8
+    url: https://dl.rockylinux.org/pub/rocky/8/images/x86_64/Rocky-8-GenericCloud-Base.latest.x86_64.qcow2
+    storage: 20Gi
+    os:
+      family: Linux
+      name: Rocky Linux
+      version: "8"
+    architecture: amd64
+    description: "Rocky Linux 8 GenericCloud image"
+  - name: rocky-9
+    url: https://dl.rockylinux.org/pub/rocky/9/images/x86_64/Rocky-9-GenericCloud-Base.latest.x86_64.qcow2
+    storage: 20Gi
+    os:
+      family: Linux
+      name: Rocky Linux
+      version: "9"
+    architecture: amd64
+    description: "Rocky Linux 9 GenericCloud image"
+  - name: rocky-10
+    url: https://dl.rockylinux.org/pub/rocky/10/images/x86_64/Rocky-10-GenericCloud-Base.latest.x86_64.qcow2
+    storage: 20Gi
+    os:
+      family: Linux
+      name: Rocky Linux
+      version: "10"
+    architecture: amd64
+    description: "Rocky Linux 10 GenericCloud image"
+  - name: almalinux-8
+    url: https://repo.almalinux.org/almalinux/8/cloud/x86_64/images/AlmaLinux-8-GenericCloud-latest.x86_64.qcow2
+    storage: 20Gi
+    os:
+      family: Linux
+      name: AlmaLinux
+      version: "8"
+    architecture: amd64
+    description: "AlmaLinux 8 GenericCloud image"
+  - name: almalinux-9
+    url: https://repo.almalinux.org/almalinux/9/cloud/x86_64/images/AlmaLinux-9-GenericCloud-latest.x86_64.qcow2
+    storage: 20Gi
+    os:
+      family: Linux
+      name: AlmaLinux
+      version: "9"
+    architecture: amd64
+    description: "AlmaLinux 9 GenericCloud image"
+  - name: almalinux-10
+    url: https://repo.almalinux.org/almalinux/10/cloud/x86_64/images/AlmaLinux-10-GenericCloud-latest.x86_64.qcow2
+    storage: 20Gi
+    os:
+      family: Linux
+      name: AlmaLinux
+      version: "10"
+    architecture: amd64
+    description: "AlmaLinux 10 GenericCloud image"
+  - name: debian-12
+    url: https://cloud.debian.org/images/cloud/bookworm/latest/debian-12-generic-amd64.qcow2
+    storage: 20Gi
+    os:
+      family: Linux
+      name: Debian
+      version: "12"
+    architecture: amd64
+    description: "Debian 12 (Bookworm) generic cloud image"
+  - name: debian-13
+    url: https://cloud.debian.org/images/cloud/trixie/latest/debian-13-generic-amd64.qcow2
+    storage: 20Gi
+    os:
+      family: Linux
+      name: Debian
+      version: "13"
+    architecture: amd64
+    description: "Debian 13 (Trixie) generic cloud image"
+  - name: centos-stream-9
+    url: https://cloud.centos.org/centos/9-stream/x86_64/images/CentOS-Stream-GenericCloud-9-latest.x86_64.qcow2
+    storage: 20Gi
+    os:
+      family: Linux
+      name: CentOS
+      version: "9"
+    architecture: amd64
+    description: "CentOS Stream 9 GenericCloud image"
+  - name: centos-stream-10
+    url: https://cloud.centos.org/centos/10-stream/x86_64/images/CentOS-Stream-GenericCloud-10-latest.x86_64.qcow2
+    storage: 20Gi
+    os:
+      family: Linux
+      name: CentOS
+      version: "10"
+    architecture: amd64
+    description: "CentOS Stream 10 GenericCloud image"
+  - name: opensuse-leap-15.6
+    url: https://download.opensuse.org/repositories/Cloud:/Images:/Leap_15.6/images/openSUSE-Leap-15.6.x86_64-NoCloud.qcow2
+    storage: 20Gi
+    os:
+      family: Linux
+      name: openSUSE
+      version: "15.6"
+    architecture: amd64
+    description: "openSUSE Leap 15.6 NoCloud image"
+  - name: opensuse-leap-16.0
+    url: https://download.opensuse.org/repositories/openSUSE:/Leap:/16.0:/Images/images/Leap-16.0-Minimal-VM.x86_64-Cloud.qcow2
+    storage: 20Gi
+    os:
+      family: Linux
+      name: openSUSE
+      version: "16.0"
+    architecture: amd64
+    description: "openSUSE Leap 16.0 Minimal VM cloud image"
+  - name: ubuntu-20.04
+    url: https://cloud-images.ubuntu.com/focal/current/focal-server-cloudimg-amd64.img
+    storage: 20Gi
+    os:
+      family: Linux
+      name: Ubuntu
+      version: "20.04"
+    architecture: amd64
+    description: "Ubuntu 20.04 LTS (Focal Fossa) cloud image"
+  - name: alpine-3.21
+    url: https://dl-cdn.alpinelinux.org/alpine/v3.21/releases/cloud/nocloud_alpine-3.21.6-x86_64-bios-cloudinit-r0.qcow2
+    storage: 20Gi
+    os:
+      family: Linux
+      name: Alpine
+      version: "3.21"
+    architecture: amd64
+    description: "Alpine Linux 3.21 cloud image (cloud-init)"
diff --git a/packages/system/vm-disk-rd/cozyrds/vm-disk.yaml b/packages/system/vm-disk-rd/cozyrds/vm-disk.yaml
index 4d0d7e37..3ce046ed 100644
--- a/packages/system/vm-disk-rd/cozyrds/vm-disk.yaml
+++ b/packages/system/vm-disk-rd/cozyrds/vm-disk.yaml
@@ -8,7 +8,7 @@ spec:
     singular: vmdisk
     plural: vmdisks
     openAPISchema: |-
-      {"title":"Chart Values","type":"object","properties":{"source":{"description":"The source image location used to create a disk.","type":"object","default":{},"properties":{"http":{"description":"Download image from an HTTP source.","type":"object","required":["url"],"properties":{"url":{"description":"URL to download the image.","type":"string"}}},"image":{"description":"Use image by name.","type":"object","required":["name"],"properties":{"name":{"description":"Name of the image to use (uploaded as \"golden image\" or from the list: `ubuntu`, `fedora`, `cirros`, `alpine`, `talos`).","type":"string"}}},"upload":{"description":"Upload local image.","type":"object"}}},"optical":{"description":"Defines if disk should be considered optical.","type":"boolean","default":false},"storage":{"description":"The size of the disk allocated for the virtual machine.","default":"5Gi","pattern":"^(\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))(([KMGTPE]i)|[numkMGTPE]|([eE](\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))))?$","anyOf":[{"type":"integer"},{"type":"string"}],"x-kubernetes-int-or-string":true},"storageClass":{"description":"StorageClass used to store the data.","type":"string","default":"replicated"}}}
+      {"title":"Chart Values","type":"object","properties":{"source":{"description":"The source image location used to create a disk.","type":"object","default":{},"properties":{"disk":{"description":"Clone an existing vm-disk.","type":"object","required":["name"],"properties":{"name":{"description":"Name of the vm-disk to clone.","type":"string"}}},"http":{"description":"Download image from an HTTP source.","type":"object","required":["url"],"properties":{"url":{"description":"URL to download the image.","type":"string"}}},"image":{"description":"Use image by name from default collection.","type":"object","required":["name"],"properties":{"name":{"description":"Name of the image to use.","type":"string"}}},"upload":{"description":"Upload local image.","type":"object"}}},"optical":{"description":"Defines if disk should be considered optical.","type":"boolean","default":false},"storage":{"description":"The size of the disk allocated for the virtual machine.","default":"5Gi","pattern":"^(\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))(([KMGTPE]i)|[numkMGTPE]|([eE](\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))))?$","anyOf":[{"type":"integer"},{"type":"string"}],"x-kubernetes-int-or-string":true},"storageClass":{"description":"StorageClass used to store the data.","type":"string","default":"replicated"}}}
   release:
     prefix: vm-disk-
     labels:
diff --git a/packages/system/vm-instance-rd/cozyrds/vm-instance.yaml b/packages/system/vm-instance-rd/cozyrds/vm-instance.yaml
index e93fa86b..e56bd9dd 100644
--- a/packages/system/vm-instance-rd/cozyrds/vm-instance.yaml
+++ b/packages/system/vm-instance-rd/cozyrds/vm-instance.yaml
@@ -8,7 +8,7 @@ spec:
     singular: vminstance
     plural: vminstances
     openAPISchema: |-
-      {"title":"Chart Values","type":"object","properties":{"external":{"description":"Enable external access from outside the cluster.","type":"boolean","default":false},"externalMethod":{"description":"Method to pass through traffic to the VM.","type":"string","default":"PortList","enum":["PortList","WholeIP"]},"externalPorts":{"description":"Ports to forward from outside the cluster.","type":"array","default":[22],"items":{"type":"integer"}},"runStrategy":{"description":"Requested running state of the VirtualMachineInstance","type":"string","default":"Always","enum":["Always","Halted","Manual","RerunOnFailure","Once"]},"instanceType":{"description":"Virtual Machine instance type.","type":"string","default":"u1.medium"},"instanceProfile":{"description":"Virtual Machine preferences profile.","type":"string","default":"ubuntu","enum":["alpine","centos.7","centos.7.desktop","centos.stream10","centos.stream10.desktop","centos.stream8","centos.stream8.desktop","centos.stream8.dpdk","centos.stream9","centos.stream9.desktop","centos.stream9.dpdk","cirros","fedora","fedora.arm64","opensuse.leap","opensuse.tumbleweed","rhel.10","rhel.10.arm64","rhel.7","rhel.7.desktop","rhel.8","rhel.8.desktop","rhel.8.dpdk","rhel.9","rhel.9.arm64","rhel.9.desktop","rhel.9.dpdk","rhel.9.realtime","sles","ubuntu","windows.10","windows.10.virtio","windows.11","windows.11.virtio","windows.2k16","windows.2k16.virtio","windows.2k19","windows.2k19.virtio","windows.2k22","windows.2k22.virtio","windows.2k25","windows.2k25.virtio",""]},"disks":{"description":"List of disks to attach.","type":"array","default":[],"items":{"type":"object","required":["name"],"properties":{"bus":{"description":"Disk bus type (e.g. \"sata\").","type":"string"},"name":{"description":"Disk name.","type":"string"}}}},"networks":{"description":"Networks to attach the VM to.","type":"array","default":[],"items":{"type":"object","properties":{"name":{"description":"Network attachment name.","type":"string"}}}},"subnets":{"description":"Deprecated: use networks instead.","type":"array","default":[],"items":{"type":"object","properties":{"name":{"description":"Network attachment name.","type":"string"}}}},"gpus":{"description":"List of GPUs to attach (NVIDIA driver requires at least 4 GiB RAM).","type":"array","default":[],"items":{"type":"object","required":["name"],"properties":{"name":{"description":"The name of the GPU resource to attach.","type":"string"}}}},"cpuModel":{"description":"Model specifies the CPU model inside the VMI. List of available models https://github.com/libvirt/libvirt/tree/master/src/cpu_map","type":"string","default":""},"resources":{"description":"Resource configuration for the virtual machine.","type":"object","default":{},"properties":{"cpu":{"description":"Number of CPU cores allocated.","pattern":"^(\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))(([KMGTPE]i)|[numkMGTPE]|([eE](\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))))?$","anyOf":[{"type":"integer"},{"type":"string"}],"x-kubernetes-int-or-string":true},"memory":{"description":"Amount of memory allocated.","pattern":"^(\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))(([KMGTPE]i)|[numkMGTPE]|([eE](\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))))?$","anyOf":[{"type":"integer"},{"type":"string"}],"x-kubernetes-int-or-string":true},"sockets":{"description":"Number of CPU sockets (vCPU topology).","pattern":"^(\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))(([KMGTPE]i)|[numkMGTPE]|([eE](\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))))?$","anyOf":[{"type":"integer"},{"type":"string"}],"x-kubernetes-int-or-string":true}}},"sshKeys":{"description":"List of SSH public keys for authentication.","type":"array","default":[],"items":{"type":"string"}},"cloudInit":{"description":"Cloud-init user data.","type":"string","default":""},"cloudInitSeed":{"description":"Seed string to generate SMBIOS UUID for the VM.","type":"string","default":""}}}
+      {"title":"Chart Values","type":"object","properties":{"external":{"description":"Enable external access from outside the cluster.","type":"boolean","default":false},"externalMethod":{"description":"Method to pass through traffic to the VM.","type":"string","default":"PortList","enum":["PortList","WholeIP"]},"externalPorts":{"description":"Ports to forward from outside the cluster.","type":"array","default":[22],"items":{"type":"integer"}},"externalAllowICMP":{"description":"Whether to accept ICMP traffic to the VM in PortList mode (preserves ping and PMTU discovery). No effect in WholeIP mode. Default true so ping behaves as users expect even when port filtering is in effect.","type":"boolean","default":true},"runStrategy":{"description":"Requested running state of the VirtualMachineInstance","type":"string","default":"Always","enum":["Always","Halted","Manual","RerunOnFailure","Once"]},"instanceType":{"description":"Virtual Machine instance type.","type":"string","default":"u1.medium"},"instanceProfile":{"description":"Virtual Machine preferences profile.","type":"string","default":"ubuntu","enum":["alpine","centos.7","centos.7.desktop","centos.stream10","centos.stream10.desktop","centos.stream8","centos.stream8.desktop","centos.stream8.dpdk","centos.stream9","centos.stream9.desktop","centos.stream9.dpdk","cirros","fedora","fedora.arm64","opensuse.leap","opensuse.tumbleweed","rhel.10","rhel.10.arm64","rhel.7","rhel.7.desktop","rhel.8","rhel.8.desktop","rhel.8.dpdk","rhel.9","rhel.9.arm64","rhel.9.desktop","rhel.9.dpdk","rhel.9.realtime","sles","ubuntu","windows.10","windows.10.virtio","windows.11","windows.11.virtio","windows.2k16","windows.2k16.virtio","windows.2k19","windows.2k19.virtio","windows.2k22","windows.2k22.virtio","windows.2k25","windows.2k25.virtio",""]},"disks":{"description":"List of disks to attach.","type":"array","default":[],"items":{"type":"object","required":["name"],"properties":{"bus":{"description":"Disk bus type (e.g. \"sata\").","type":"string"},"name":{"description":"Disk name.","type":"string"}}}},"networks":{"description":"Networks to attach the VM to.","type":"array","default":[],"items":{"type":"object","properties":{"name":{"description":"Network attachment name.","type":"string"}}}},"subnets":{"description":"Deprecated: use networks instead.","type":"array","default":[],"items":{"type":"object","properties":{"name":{"description":"Network attachment name.","type":"string"}}}},"gpus":{"description":"List of GPUs to attach (NVIDIA driver requires at least 4 GiB RAM).","type":"array","default":[],"items":{"type":"object","required":["name"],"properties":{"name":{"description":"The name of the GPU resource to attach.","type":"string"}}}},"cpuModel":{"description":"Model specifies the CPU model inside the VMI. List of available models https://github.com/libvirt/libvirt/tree/master/src/cpu_map","type":"string","default":""},"resources":{"description":"Resource configuration for the virtual machine.","type":"object","default":{},"properties":{"cpu":{"description":"Number of CPU cores allocated.","pattern":"^(\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))(([KMGTPE]i)|[numkMGTPE]|([eE](\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))))?$","anyOf":[{"type":"integer"},{"type":"string"}],"x-kubernetes-int-or-string":true},"memory":{"description":"Amount of memory allocated.","pattern":"^(\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))(([KMGTPE]i)|[numkMGTPE]|([eE](\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))))?$","anyOf":[{"type":"integer"},{"type":"string"}],"x-kubernetes-int-or-string":true},"sockets":{"description":"Number of CPU sockets (vCPU topology).","pattern":"^(\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))(([KMGTPE]i)|[numkMGTPE]|([eE](\\+|-)?(([0-9]+(\\.[0-9]*)?)|(\\.[0-9]+))))?$","anyOf":[{"type":"integer"},{"type":"string"}],"x-kubernetes-int-or-string":true}}},"sshKeys":{"description":"List of SSH public keys for authentication.","type":"array","default":[],"items":{"type":"string"}},"cloudInit":{"description":"Cloud-init user data.","type":"string","default":""},"cloudInitSeed":{"description":"Seed string to generate SMBIOS UUID for the VM.","type":"string","default":""}}}
   release:
     prefix: vm-instance-
     labels:
@@ -26,7 +26,7 @@ spec:
     tags:
       - compute
     icon: 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
-    keysOrder: [["apiVersion"], ["appVersion"], ["kind"], ["metadata"], ["metadata", "name"], ["spec", "external"], ["spec", "externalMethod"], ["spec", "externalPorts"], ["spec", "runStrategy"], ["spec", "instanceType"], ["spec", "instanceProfile"], ["spec", "disks"], ["spec", "networks"], ["spec", "subnets"], ["spec", "gpus"], ["spec", "cpuModel"], ["spec", "resources"], ["spec", "sshKeys"], ["spec", "cloudInit"], ["spec", "cloudInitSeed"]]
+    keysOrder: [["apiVersion"], ["appVersion"], ["kind"], ["metadata"], ["metadata", "name"], ["spec", "external"], ["spec", "externalMethod"], ["spec", "externalPorts"], ["spec", "externalAllowICMP"], ["spec", "runStrategy"], ["spec", "instanceType"], ["spec", "instanceProfile"], ["spec", "disks"], ["spec", "networks"], ["spec", "subnets"], ["spec", "gpus"], ["spec", "cpuModel"], ["spec", "resources"], ["spec", "sshKeys"], ["spec", "cloudInit"], ["spec", "cloudInitSeed"]]
   secrets:
     exclude: []
     include: []
diff --git a/pkg/apis/apps/validation/validation.go b/pkg/apis/apps/validation/validation.go
index 273873a4..d62f2f22 100644
--- a/pkg/apis/apps/validation/validation.go
+++ b/pkg/apis/apps/validation/validation.go
@@ -17,16 +17,39 @@ limitations under the License.
 package validation
 
 import (
+	"regexp"
+
 	k8svalidation "k8s.io/apimachinery/pkg/util/validation"
 	"k8s.io/apimachinery/pkg/util/validation/field"
 )
 
-// ValidateApplicationName validates that an Application name conforms to DNS-1035.
-// This is required because Application names are used to create Kubernetes resources
-// (Services, Namespaces, etc.) that must have DNS-1035 compliant names.
-// Note: length validation is handled separately by validateNameLength in the REST
-// handler, which computes dynamic limits based on Helm release prefix.
-func ValidateApplicationName(name string, fldPath *field.Path) field.ErrorList {
+// TenantKind is the Application.Kind string that gates the tenant-specific
+// name rules below. It must stay in sync with the `kind` field of the tenant
+// ApplicationDefinition (packages/system/tenant-rd/cozyrds/tenant.yaml) which
+// is the upstream source the aggregated API reads at startup via
+// config.Application.Kind.
+const TenantKind = "Tenant"
+
+// tenantNameRegex enforces alphanumeric-only tenant names that begin with a
+// lowercase letter. This is stricter than DNS-1035 because the tenant Helm
+// chart's tenant.name helper (packages/apps/tenant/templates/_helpers.tpl)
+// splits Release.Name on "-" and fails unless the result is exactly
+// ["tenant", "<name>"]. Any dash inside <name> would break that invariant at
+// Helm template time, so the aggregated API must reject such names up-front
+// with a specific error. Requiring a leading letter (rather than letting
+// leading-digit names fall through to DNS-1035) keeps the error message
+// tenant-specific for all invalid inputs.
+var tenantNameRegex = regexp.MustCompile(`^[a-z][a-z0-9]*$`)
+
+// ValidateApplicationName validates that an Application name is acceptable for
+// the given kind. All applications must conform to DNS-1035 because their
+// names are used to create Kubernetes resources (Services, Namespaces, etc.)
+// that require DNS-1035 compliance. Tenant applications additionally must be
+// alphanumeric and begin with a lowercase letter because of the Helm chart
+// constraint described on tenantNameRegex.
+// Note: length validation is handled separately by validateNameLength in the
+// REST handler, which computes dynamic limits based on Helm release prefix.
+func ValidateApplicationName(name, kindName string, fldPath *field.Path) field.ErrorList {
 	allErrs := field.ErrorList{}
 
 	if len(name) == 0 {
@@ -34,6 +57,16 @@ func ValidateApplicationName(name string, fldPath *field.Path) field.ErrorList {
 		return allErrs
 	}
 
+	// Tenant names must be alphanumeric starting with a letter — see
+	// tenantNameRegex comment for the reason. Check before DNS-1035 so the
+	// error message is specific to the tenant contract, not the generic DNS
+	// label rules.
+	if kindName == TenantKind && !tenantNameRegex.MatchString(name) {
+		allErrs = append(allErrs, field.Invalid(fldPath, name,
+			"tenant names must start with a lowercase letter and contain only lowercase letters and digits; dashes are not allowed"))
+		return allErrs
+	}
+
 	for _, msg := range k8svalidation.IsDNS1035Label(name) {
 		allErrs = append(allErrs, field.Invalid(fldPath, name, msg))
 	}
diff --git a/pkg/apis/apps/validation/validation_test.go b/pkg/apis/apps/validation/validation_test.go
index 7bf194d5..7930f8ae 100644
--- a/pkg/apis/apps/validation/validation_test.go
+++ b/pkg/apis/apps/validation/validation_test.go
@@ -27,55 +27,148 @@ func TestValidateApplicationName(t *testing.T) {
 	tests := []struct {
 		name      string
 		appName   string
+		kindName  string
 		wantError bool
 	}{
-		// Valid names
-		{"valid simple name", "tenant-one", false},
-		{"valid single letter", "a", false},
-		{"valid with numbers", "abc-123", false},
-		{"valid lowercase", "my-tenant", false},
-		{"valid long name", "my-very-long-tenant-name", false},
-		{"valid double hyphen", "my--tenant", false},
-		{"valid at DNS-1035 max (63 chars)", strings.Repeat("a", 63), false},
+		// Valid names (non-tenant kinds permit DNS-1035 including hyphens)
+		{"valid simple name", "tenant-one", "MySQL", false},
+		{"valid single letter", "a", "MySQL", false},
+		{"valid with numbers", "abc-123", "MySQL", false},
+		{"valid lowercase", "my-tenant", "MySQL", false},
+		{"valid long name", "my-very-long-tenant-name", "MySQL", false},
+		{"valid double hyphen", "my--tenant", "MySQL", false},
+		{"valid at DNS-1035 max (63 chars)", strings.Repeat("a", 63), "MySQL", false},
+		{"valid with empty kind", "my-db", "", false},
 
 		// Invalid: starts with wrong character
-		{"starts with digit", "1john", true},
-		{"only digits", "123", true},
-		{"starts with hyphen", "-tenant", true},
+		{"starts with digit", "1john", "MySQL", true},
+		{"only digits", "123", "MySQL", true},
+		{"starts with hyphen", "-tenant", "MySQL", true},
 
 		// Invalid: ends with wrong character
-		{"ends with hyphen", "tenant-", true},
+		{"ends with hyphen", "tenant-", "MySQL", true},
 
 		// Invalid: wrong characters
-		{"uppercase letters", "Tenant", true},
-		{"mixed case", "myTenant", true},
-		{"underscore", "my_tenant", true},
-		{"dot", "my.tenant", true},
-		{"space", "my tenant", true},
-		{"unicode cyrillic", "тенант", true},
-		{"unicode emoji", "tenant🚀", true},
-		{"special chars", "tenant@home", true},
-		{"colon", "tenant:one", true},
-		{"slash", "tenant/one", true},
+		{"uppercase letters", "Tenant", "MySQL", true},
+		{"mixed case", "myTenant", "MySQL", true},
+		{"underscore", "my_tenant", "MySQL", true},
+		{"dot", "my.tenant", "MySQL", true},
+		{"space", "my tenant", "MySQL", true},
+		{"unicode cyrillic", "тенант", "MySQL", true},
+		{"unicode emoji", "tenant🚀", "MySQL", true},
+		{"special chars", "tenant@home", "MySQL", true},
+		{"colon", "tenant:one", "MySQL", true},
+		{"slash", "tenant/one", "MySQL", true},
 
 		// Invalid: empty or whitespace
-		{"empty string", "", true},
-		{"only spaces", "   ", true},
-		{"leading space", " tenant", true},
-		{"trailing space", "tenant ", true},
+		{"empty string", "", "MySQL", true},
+		{"only spaces", "   ", "MySQL", true},
+		{"leading space", " tenant", "MySQL", true},
+		{"trailing space", "tenant ", "MySQL", true},
 
 		// Invalid: exceeds DNS-1035 max length (63)
-		{"too long (64 chars)", strings.Repeat("a", 64), true},
-		{"way too long (100 chars)", strings.Repeat("a", 100), true},
+		{"too long (64 chars)", strings.Repeat("a", 64), "MySQL", true},
+		{"way too long (100 chars)", strings.Repeat("a", 100), "MySQL", true},
+
+		// Tenant kind: stricter alphanumeric-only rule.
+		// The tenant Helm chart's tenant.name helper (packages/apps/tenant/templates/_helpers.tpl)
+		// splits Release.Name on "-" and fails unless the result is exactly
+		// ["tenant", "<name>"]. Any dash inside <name> breaks that invariant, so
+		// the aggregated API must reject tenant names containing dashes up-front
+		// with a specific error — instead of letting Flux reconciliation fail later.
+		{"tenant alphanumeric simple", "foo", "Tenant", false},
+		{"tenant alphanumeric with digits", "foo123", "Tenant", false},
+		{"tenant single char", "a", "Tenant", false},
+		{"tenant single hyphen", "foo-bar", "Tenant", true},
+		{"tenant leading hyphen", "-foo", "Tenant", true},
+		{"tenant trailing hyphen", "foo-", "Tenant", true},
+		{"tenant double hyphen", "foo--bar", "Tenant", true},
+		{"tenant uppercase", "Foo", "Tenant", true},
+		{"tenant underscore", "foo_bar", "Tenant", true},
+		{"tenant empty", "", "Tenant", true},
+		// Leading digit must be caught by the tenant-specific regex (not by
+		// falling through to DNS-1035) so the error message reflects the
+		// tenant contract — see TestValidateApplicationName_TenantErrorMessage.
+		{"tenant leading digit", "123foo", "Tenant", true},
 	}
 
 	for _, tt := range tests {
 		t.Run(tt.name, func(t *testing.T) {
-			errs := ValidateApplicationName(tt.appName, field.NewPath("metadata").Child("name"))
+			errs := ValidateApplicationName(tt.appName, tt.kindName, field.NewPath("metadata").Child("name"))
 			if (len(errs) > 0) != tt.wantError {
-				t.Errorf("ValidateApplicationName(%q) returned %d errors, wantError = %v, errors = %v",
-					tt.appName, len(errs), tt.wantError, errs)
+				t.Errorf("ValidateApplicationName(%q, kind=%q) returned %d errors, wantError = %v, errors = %v",
+					tt.appName, tt.kindName, len(errs), tt.wantError, errs)
 			}
 		})
 	}
 }
+
+// TestValidateApplicationName_TenantErrorMessage pins the contract that when
+// a tenant name is invalid, the returned error message is specific to the
+// tenant naming rule — not the generic DNS-1035 message. Otherwise users get
+// back "must start with an alphabetic character" or similar and have no way
+// to know the constraint is tied to the tenant Helm chart.
+func TestValidateApplicationName_TenantErrorMessage(t *testing.T) {
+	// Every tenant-invalid name below must surface a tenant-specific error
+	// message. In particular, "123foo" starts with a digit — the original
+	// implementation let that fall through to DNS-1035 with a generic error;
+	// the regex is tightened specifically so this case fails up-front.
+	invalidTenantNames := []string{
+		"foo-bar",  // dash
+		"-foo",     // leading dash
+		"foo-",     // trailing dash
+		"foo--bar", // double dash
+		"Foo",      // uppercase
+		"foo_bar",  // underscore
+		"foo.bar",  // dot
+		"foo bar",  // space
+		"123foo",   // leading digit — must not fall through to DNS-1035
+	}
+
+	const wantSubstring = "tenant names must"
+
+	for _, name := range invalidTenantNames {
+		t.Run(name, func(t *testing.T) {
+			errs := ValidateApplicationName(name, "Tenant", field.NewPath("metadata").Child("name"))
+			if len(errs) == 0 {
+				t.Fatalf("expected error for tenant name %q, got none", name)
+			}
+			if !strings.Contains(errs[0].Detail, wantSubstring) {
+				t.Errorf("tenant name %q: error detail = %q, want substring %q (generic DNS-1035 message is not tenant-specific)",
+					name, errs[0].Detail, wantSubstring)
+			}
+		})
+	}
+}
+
+// TestValidateApplicationName_TenantLengthFallthrough documents the one
+// invalid-tenant case where the error message is intentionally NOT tenant-
+// specific: when a name contains only valid tenant characters but exceeds
+// the DNS-1035 63-char label limit, the length error comes from DNS-1035
+// because length is not a tenant-specific constraint (every application
+// kind is subject to the same Kubernetes label limit). REST.validateNameLength
+// further tightens the limit using the Helm release prefix, so tenants cannot
+// actually reach 64 characters end-to-end — this test only pins the package-
+// level fallthrough so a future refactor does not accidentally promote the
+// length error into tenant-specific wording.
+//
+// NOTE: this is an architectural decision, not a user-facing requirement.
+// If tenant length is ever promoted into a tenant-specific rule (e.g. to
+// include the Helm release prefix budget in this package's error message),
+// this test should be updated or deleted — it is not a backwards-compat
+// guarantee, just a checkpoint on the current layering.
+func TestValidateApplicationName_TenantLengthFallthrough(t *testing.T) {
+	name := strings.Repeat("a", 64) // valid tenant pattern, too long for DNS-1035
+
+	errs := ValidateApplicationName(name, "Tenant", field.NewPath("metadata").Child("name"))
+	if len(errs) == 0 {
+		t.Fatalf("expected DNS-1035 length error for 64-char tenant name, got none")
+	}
+	// This error is the generic DNS-1035 one, NOT the tenant-specific message.
+	// We deliberately do not assert against the exact upstream DNS-1035 text
+	// (that would tie this test to a k8s.io/apimachinery internal string and
+	// break on unrelated upstream wording changes).
+	if strings.Contains(errs[0].Detail, "tenant names must") {
+		t.Errorf("64-char tenant name should surface the generic DNS-1035 error, got tenant-specific: %q", errs[0].Detail)
+	}
+}
diff --git a/pkg/apiserver/apiserver.go b/pkg/apiserver/apiserver.go
index bf02e642..005fc640 100644
--- a/pkg/apiserver/apiserver.go
+++ b/pkg/apiserver/apiserver.go
@@ -21,6 +21,7 @@ import (
 	"fmt"
 	"time"
 
+	cozyv1alpha1 "github.com/cozystack/cozystack/api/v1alpha1"
 	helmv2 "github.com/fluxcd/helm-controller/api/v2"
 	corev1 "k8s.io/api/core/v1"
 	rbacv1 "k8s.io/api/rbac/v1"
@@ -79,6 +80,11 @@ func init() {
 	if err := rbacv1.AddToScheme(mgrScheme); err != nil {
 		panic(fmt.Errorf("Failed to add RBAC types to scheme: %w", err))
 	}
+
+	// Register Cozystack types for WorkloadMonitor queries.
+	if err := cozyv1alpha1.AddToScheme(mgrScheme); err != nil {
+		panic(fmt.Errorf("failed to add Cozystack types to scheme: %w", err))
+	}
 	// Add unversioned types.
 	metav1.AddToGroupVersion(Scheme, schema.GroupVersion{Version: "v1"})
 
@@ -168,6 +174,7 @@ func (c completedConfig) New() (*CozyServer, error) {
 		&corev1.Namespace{},
 		&corev1.Service{},
 		&rbacv1.RoleBinding{},
+		&cozyv1alpha1.WorkloadMonitor{},
 	); err != nil {
 		return nil, fmt.Errorf("failed to get informers: %w", err)
 	}
diff --git a/pkg/cmd/server/start.go b/pkg/cmd/server/start.go
index f7791332..7ebbbc89 100644
--- a/pkg/cmd/server/start.go
+++ b/pkg/cmd/server/start.go
@@ -160,6 +160,34 @@ func (o *CozyServerOptions) Complete() error {
 	// Convert to ResourceConfig
 	o.ResourceConfig = &config.ResourceConfig{}
 	for _, crd := range crdList.Items {
+		release := config.ReleaseConfig{
+			Prefix: crd.Spec.Release.Prefix,
+			Labels: crd.Spec.Release.Labels,
+			ChartRef: config.ChartRefConfig{
+				Kind:      crd.Spec.Release.ChartRef.Kind,
+				Name:      crd.Spec.Release.ChartRef.Name,
+				Namespace: crd.Spec.Release.ChartRef.Namespace,
+			},
+		}
+		// Per-Application HelmRelease Install/Upgrade timeout. Applications
+		// whose parent chart contains asynchronously-provisioned resources
+		// the chart itself depends on (for example, the Kamaji-provisioned
+		// admin-kubeconfig Secret for Kubernetes tenants) need a longer
+		// wait budget than the Flux default. Consumed by the REST storage
+		// layer when building the HelmRelease Spec. The parser rejects
+		// units Flux would reject at webhook time, so a bad annotation
+		// surfaces as a loud startup failure instead of a silent drop to
+		// defaults.
+		d, err := config.ParseHelmInstallTimeoutAnnotation(
+			crd.Annotations[config.HelmInstallTimeoutAnnotation],
+		)
+		if err != nil {
+			return fmt.Errorf(
+				"ApplicationDefinition %q has invalid %s annotation: %w",
+				crd.Name, config.HelmInstallTimeoutAnnotation, err,
+			)
+		}
+		release.HelmInstallTimeout = d
 		resource := config.Resource{
 			Application: config.ApplicationConfig{
 				Kind:          crd.Spec.Application.Kind,
@@ -168,15 +196,7 @@ func (o *CozyServerOptions) Complete() error {
 				ShortNames:    []string{}, // TODO: implement shortnames
 				OpenAPISchema: crd.Spec.Application.OpenAPISchema,
 			},
-			Release: config.ReleaseConfig{
-				Prefix: crd.Spec.Release.Prefix,
-				Labels: crd.Spec.Release.Labels,
-				ChartRef: config.ChartRefConfig{
-					Kind:      crd.Spec.Release.ChartRef.Kind,
-					Name:      crd.Spec.Release.ChartRef.Name,
-					Namespace: crd.Spec.Release.ChartRef.Namespace,
-				},
-			},
+			Release: release,
 		}
 		o.ResourceConfig.Resources = append(o.ResourceConfig.Resources, resource)
 	}
diff --git a/pkg/config/config.go b/pkg/config/config.go
index 1e123e2c..21de27b2 100644
--- a/pkg/config/config.go
+++ b/pkg/config/config.go
@@ -16,6 +16,48 @@ limitations under the License.
 
 package config
 
+import (
+	"fmt"
+	"regexp"
+	"time"
+)
+
+// HelmInstallTimeoutAnnotation is the ApplicationDefinition metadata
+// annotation key that overrides the Flux HelmRelease Install.Timeout and
+// Upgrade.Timeout for a given Application kind.
+const HelmInstallTimeoutAnnotation = "release.cozystack.io/helm-install-timeout"
+
+// helmTimeoutPattern mirrors the CRD validation pattern used by Flux
+// helm-controller on HelmReleaseSpec.Install.Timeout (ms/s/m/h units only).
+// time.ParseDuration accepts ns/us/µs, but Flux rejects them - parsing here
+// with the same shape avoids feeding the controller a value it will later
+// reject at webhook time. See
+// github.com/fluxcd/helm-controller/api/v2 HelmReleaseSpec.Install.Timeout
+// in the go module cache.
+var helmTimeoutPattern = regexp.MustCompile(`^([0-9]+(\.[0-9]+)?(ms|s|m|h))+$`)
+
+// ParseHelmInstallTimeoutAnnotation parses the value of the
+// release.cozystack.io/helm-install-timeout annotation. The empty string is
+// treated as "unset" and returns (0, nil) so callers can leave
+// HelmInstallTimeout zeroed and let flux defaults apply. Values accepted by
+// time.ParseDuration but rejected by Flux (ns/us/µs) return a helpful
+// error instead of silently parsing and failing later at HelmRelease
+// admission.
+func ParseHelmInstallTimeoutAnnotation(raw string) (time.Duration, error) {
+	if raw == "" {
+		return 0, nil
+	}
+	if !helmTimeoutPattern.MatchString(raw) {
+		return 0, fmt.Errorf("must match %s (Flux accepts ms/s/m/h units only), got %q",
+			helmTimeoutPattern, raw)
+	}
+	d, err := time.ParseDuration(raw)
+	if err != nil {
+		return 0, fmt.Errorf("time.ParseDuration(%q): %w", raw, err)
+	}
+	return d, nil
+}
+
 // ResourceConfig represents the structure of the configuration file.
 type ResourceConfig struct {
 	Resources []Resource `yaml:"resources"`
@@ -41,6 +83,12 @@ type ReleaseConfig struct {
 	Prefix   string            `yaml:"prefix"`
 	Labels   map[string]string `yaml:"labels"`
 	ChartRef ChartRefConfig    `yaml:"chartRef"`
+	// HelmInstallTimeout overrides the Flux HelmRelease Install.Timeout and
+	// Upgrade.Timeout for this Application kind. When zero, flux defaults
+	// apply. Populated from the
+	// release.cozystack.io/helm-install-timeout annotation on the
+	// ApplicationDefinition at start-up.
+	HelmInstallTimeout time.Duration `yaml:"helmInstallTimeout,omitempty"`
 }
 
 // ChartRefConfig references a Flux source artifact for the Helm chart.
diff --git a/pkg/config/config_test.go b/pkg/config/config_test.go
new file mode 100644
index 00000000..eb6990af
--- /dev/null
+++ b/pkg/config/config_test.go
@@ -0,0 +1,119 @@
+package config
+
+import (
+	"strings"
+	"testing"
+	"time"
+)
+
+// Cover the annotation parser used by cozystack-api at startup. The parser
+// is consumed by pkg/cmd/server/start.go on every ApplicationDefinition; a
+// typo here silently drops back to flux defaults and the Kubernetes tenant
+// race described in cozystack#2412 reappears, so the table must exercise:
+// - the unset path (empty string treated as "no override"),
+// - every unit Flux accepts (ms, s, m, h),
+// - compound forms (the CRD pattern accepts repeats),
+// - units time.ParseDuration accepts but Flux rejects (ns, us, µs),
+// - outright garbage.
+func TestParseHelmInstallTimeoutAnnotation(t *testing.T) {
+	cases := []struct {
+		name     string
+		input    string
+		want     time.Duration
+		wantErr  bool
+		errMatch string
+	}{
+		{
+			name:  "empty string leaves flux defaults in place",
+			input: "",
+			want:  0,
+		},
+		{
+			name:  "minutes",
+			input: "15m",
+			want:  15 * time.Minute,
+		},
+		{
+			name:  "hours",
+			input: "1h",
+			want:  time.Hour,
+		},
+		{
+			name:  "seconds",
+			input: "45s",
+			want:  45 * time.Second,
+		},
+		{
+			name:  "milliseconds",
+			input: "500ms",
+			want:  500 * time.Millisecond,
+		},
+		{
+			name:  "compound hour and minutes",
+			input: "2h30m",
+			want:  2*time.Hour + 30*time.Minute,
+		},
+		{
+			name:  "decimal minutes",
+			input: "1.5m",
+			want:  90 * time.Second,
+		},
+		{
+			name:     "nanoseconds rejected - Flux CRD pattern excludes ns",
+			input:    "500ns",
+			wantErr:  true,
+			errMatch: "Flux accepts ms/s/m/h units only",
+		},
+		{
+			name:     "microseconds rejected - Flux CRD pattern excludes us",
+			input:    "500us",
+			wantErr:  true,
+			errMatch: "Flux accepts ms/s/m/h units only",
+		},
+		{
+			name:     "microseconds unicode rejected",
+			input:    "500µs",
+			wantErr:  true,
+			errMatch: "Flux accepts ms/s/m/h units only",
+		},
+		{
+			name:     "bare digits rejected",
+			input:    "15",
+			wantErr:  true,
+			errMatch: "Flux accepts ms/s/m/h units only",
+		},
+		{
+			name:     "garbage rejected",
+			input:    "abc",
+			wantErr:  true,
+			errMatch: "Flux accepts ms/s/m/h units only",
+		},
+		{
+			name:     "negative rejected",
+			input:    "-15m",
+			wantErr:  true,
+			errMatch: "Flux accepts ms/s/m/h units only",
+		},
+	}
+
+	for _, tc := range cases {
+		t.Run(tc.name, func(t *testing.T) {
+			got, err := ParseHelmInstallTimeoutAnnotation(tc.input)
+			if tc.wantErr {
+				if err == nil {
+					t.Fatalf("expected error, got duration=%v", got)
+				}
+				if tc.errMatch != "" && !strings.Contains(err.Error(), tc.errMatch) {
+					t.Errorf("error %q does not contain %q", err.Error(), tc.errMatch)
+				}
+				return
+			}
+			if err != nil {
+				t.Fatalf("unexpected error: %v", err)
+			}
+			if got != tc.want {
+				t.Errorf("got %v, want %v", got, tc.want)
+			}
+		})
+	}
+}
diff --git a/pkg/registry/apps/application/rest.go b/pkg/registry/apps/application/rest.go
index 7871bbe7..d233ecc3 100644
--- a/pkg/registry/apps/application/rest.go
+++ b/pkg/registry/apps/application/rest.go
@@ -26,6 +26,7 @@ import (
 	"sync"
 	"time"
 
+	cozyv1alpha1 "github.com/cozystack/cozystack/api/v1alpha1"
 	helmv2 "github.com/fluxcd/helm-controller/api/v2"
 	corev1 "k8s.io/api/core/v1"
 	metainternalversion "k8s.io/apimachinery/pkg/apis/meta/internalversion"
@@ -154,8 +155,8 @@ func (r *REST) Create(ctx context.Context, obj runtime.Object, createValidation
 		return nil, fmt.Errorf("expected *appsv1alpha1.Application object, got %T", obj)
 	}
 
-	// Validate Application name conforms to DNS-1035
-	if errs := validation.ValidateApplicationName(app.Name, field.NewPath("metadata").Child("name")); len(errs) > 0 {
+	// Validate Application name format (DNS-1035 plus any kind-specific rules)
+	if errs := r.validateNameFormat(app.Name); len(errs) > 0 {
 		return nil, apierrors.NewInvalid(r.gvk.GroupKind(), app.Name, errs)
 	}
 
@@ -164,6 +165,17 @@ func (r *REST) Create(ctx context.Context, obj runtime.Object, createValidation
 		return nil, apierrors.NewInvalid(r.gvk.GroupKind(), app.Name, nameLenErrs)
 	}
 
+	// For Tenant applications, also validate that the computed workload
+	// namespace fits within the DNS-1123 label limit. A deeply-nested tenant
+	// can exceed the limit even when its own name passes the per-name Helm
+	// release length check, because the namespace is formed from the full
+	// ancestor chain.
+	if r.kindName == "Tenant" {
+		if nsErrs := r.validateTenantNamespaceLength(app.Namespace, app.Name); len(nsErrs) > 0 {
+			return nil, apierrors.NewInvalid(r.gvk.GroupKind(), app.Name, nsErrs)
+		}
+	}
+
 	// Validate that values don't contain reserved keys (starting with "_")
 	if err := validateNoInternalKeys(app.Spec); err != nil {
 		return nil, apierrors.NewBadRequest(err.Error())
@@ -692,14 +704,49 @@ func (r *REST) Watch(ctx context.Context, options *metainternalversion.ListOptio
 	}
 	customW.underlying = helmWatcher
 
+	// Start watch on WorkloadMonitor to detect pod readiness changes.
+	// For Tenant applications the WorkloadMonitor lives in a computed child
+	// namespace (see computeTenantNamespace), not in the HelmRelease namespace,
+	// so scoping the watch to `namespace` would miss all events. Use a
+	// cluster-wide watch in that case — label selectors still restrict the
+	// stream to the relevant kind/group.
+	wmLabelSelector := labels.NewSelector().Add(*appKindReq, *appGroupReq)
+	wmList := &cozyv1alpha1.WorkloadMonitorList{}
+	wmListOpts := &client.ListOptions{LabelSelector: wmLabelSelector}
+	if r.kindName != "Tenant" {
+		wmListOpts.Namespace = namespace
+	}
+	wmWatcher, err := r.w.Watch(ctx, wmList, wmListOpts)
+	if err != nil {
+		klog.Warningf("Failed to set up WorkloadMonitor watch, workload status changes won't trigger events: %v", err)
+		// Non-fatal: proceed without WorkloadMonitor watch
+		wmWatcher = nil
+	}
+
 	go func() {
+		// Capture wmWatcher for cleanup; the variable may be set to nil
+		// inside the loop when the channel closes, so defer must use this copy.
+		wmWatcherForCleanup := wmWatcher
 		defer close(customW.resultChan)
 		defer customW.underlying.Stop()
+		if wmWatcherForCleanup != nil {
+			defer wmWatcherForCleanup.Stop()
+		}
 
 		// Track whether we've sent the initial-events-end bookmark
 		initialEventsEndSent := !sendInitialEvents // If not sendInitialEvents, consider it already sent
 		var lastResourceVersion string
 
+		// Buffer of WorkloadMonitor events that arrived before the initial
+		// snapshot finished. The watch-list contract requires the stream to
+		// deliver all ADDED events followed by the initial-events-end bookmark
+		// before any live updates, so we hold WM-triggered Modified events and
+		// replay them once the bookmark has been emitted. Without this, a
+		// workload whose status flips during the snapshot window (after the
+		// Application ADDED but before the bookmark) and then stops changing
+		// would leave the client with a stale WorkloadsReady forever.
+		var pendingWMEvents []watch.Event
+
 		// Get the starting resourceVersion from options
 		// If client provides resourceVersion (e.g., from a previous List), we should skip
 		// objects with resourceVersion <= startingRV (client already has them)
@@ -710,6 +757,19 @@ func (r *REST) Watch(ctx context.Context, options *metainternalversion.ListOptio
 			}
 		}
 
+		drainPendingWMEvents := func() {
+			for _, ev := range pendingWMEvents {
+				select {
+				case customW.resultChan <- ev:
+				case <-customW.stopChan:
+					return
+				case <-ctx.Done():
+					return
+				}
+			}
+			pendingWMEvents = nil
+		}
+
 		// Helper function to send initial-events-end bookmark
 		sendInitialEventsEndBookmark := func() {
 			if initialEventsEndSent {
@@ -734,8 +794,11 @@ func (r *REST) Watch(ctx context.Context, options *metainternalversion.ListOptio
 			select {
 			case customW.resultChan <- bookmarkEvent:
 			case <-customW.stopChan:
+				return
 			case <-ctx.Done():
+				return
 			}
+			drainPendingWMEvents()
 		}
 
 		// Process watch events
@@ -763,8 +826,10 @@ func (r *REST) Watch(ctx context.Context, options *metainternalversion.ListOptio
 							APIVersion: appsv1alpha1.SchemeGroupVersion.String(),
 							Kind:       r.kindName,
 						}
+						justFlipped := false
 						if !initialEventsEndSent {
 							initialEventsEndSent = true
+							justFlipped = true
 							bookmarkApp.SetAnnotations(map[string]string{
 								"k8s.io/initial-events-end": "true",
 							})
@@ -781,6 +846,9 @@ func (r *REST) Watch(ctx context.Context, options *metainternalversion.ListOptio
 						case <-ctx.Done():
 							return
 						}
+						if justFlipped {
+							drainPendingWMEvents()
+						}
 					}
 					continue
 				}
@@ -868,6 +936,104 @@ func (r *REST) Watch(ctx context.Context, options *metainternalversion.ListOptio
 					return
 				}
 
+			case wmEvent, ok := <-wmResultChan(wmWatcher):
+				if !ok {
+					klog.V(4).Info("WorkloadMonitor watcher closed")
+					wmWatcher = nil
+					continue
+				}
+				if wmEvent.Type == watch.Bookmark || wmEvent.Type == watch.Error {
+					if wmEvent.Type == watch.Error {
+						klog.V(4).Infof("WorkloadMonitor watch error event: %v", wmEvent.Object)
+					}
+					continue
+				}
+				wm, ok := wmEvent.Object.(*cozyv1alpha1.WorkloadMonitor)
+				if !ok {
+					continue
+				}
+				// All WM event types (Added/Modified/Deleted) produce a Modified
+				// Application event because the Application itself is what changed
+				// from the client's perspective.
+				wmAppName, hasLabel := wm.Labels[ApplicationNameLabel]
+				if !hasLabel {
+					continue
+				}
+				// Filter: skip WorkloadMonitor events for applications not matching
+				// the watch scope (single-resource or field-selector filtered watches)
+				hrName := r.releaseConfig.Prefix + wmAppName
+				if filterByName != "" && hrName != filterByName {
+					continue
+				}
+				if resourceName != "" && wmAppName != resourceName {
+					continue
+				}
+
+				// Locate the owning HelmRelease. For most application kinds the
+				// WorkloadMonitor and its HelmRelease live in the same namespace,
+				// but Tenant workloads live in a child namespace (see
+				// computeTenantNamespace) while the HelmRelease remains in the
+				// parent/requested namespace — so the WM-to-HR namespace mapping
+				// differs.
+				hrNS := wm.Namespace
+				if r.kindName == "Tenant" {
+					hrNS = namespace
+					// Filter out WM events whose child namespace does not
+					// correspond to the Tenant in our watched namespace, since
+					// the cluster-wide WM watch delivers events for all tenants.
+					if r.computeTenantNamespace(namespace, wmAppName) != wm.Namespace {
+						continue
+					}
+					if !fieldFilter.MatchesNamespace(hrNS) {
+						continue
+					}
+				} else if !fieldFilter.MatchesNamespace(hrNS) {
+					continue
+				}
+				hr := &helmv2.HelmRelease{}
+				if err := r.c.Get(ctx, client.ObjectKey{Namespace: hrNS, Name: hrName}, hr); err != nil {
+					klog.V(4).Infof("Cannot find HelmRelease %s/%s for WorkloadMonitor event: %v", hrNS, hrName, err)
+					continue
+				}
+				// Pass the fresh WorkloadMonitor so conversion uses the latest
+				// operational status even if the cache (r.c) has not yet
+				// observed this watch event.
+				app, err := r.ConvertHelmReleaseToApplicationWithMonitor(ctx, hr, wm)
+				if err != nil {
+					klog.V(4).Infof("Error converting HelmRelease for WorkloadMonitor event: %v", err)
+					continue
+				}
+				// Apply label selector filtering (same as HelmRelease event path)
+				if options.LabelSelector != nil {
+					sel, err := labels.Parse(options.LabelSelector.String())
+					if err != nil {
+						klog.Errorf("Invalid label selector: %v", err)
+						continue
+					}
+					if !sel.Matches(labels.Set(app.Labels)) {
+						continue
+					}
+				}
+				// Use the WorkloadMonitor's ResourceVersion for the emitted event
+				// so clients see a monotonically increasing RV and don't skip this update.
+				app.SetResourceVersion(wm.GetResourceVersion())
+				outEvent := watch.Event{Type: watch.Modified, Object: &app}
+				// Buffer WM-triggered events that arrive before the
+				// initial-events-end bookmark. They will be replayed in order
+				// immediately after the bookmark is emitted.
+				if !initialEventsEndSent {
+					pendingWMEvents = append(pendingWMEvents, outEvent)
+					continue
+				}
+				lastResourceVersion = wm.GetResourceVersion()
+				select {
+				case customW.resultChan <- outEvent:
+				case <-customW.stopChan:
+					return
+				case <-ctx.Done():
+					return
+				}
+
 			case <-customW.stopChan:
 				return
 			case <-ctx.Done():
@@ -880,6 +1046,15 @@ func (r *REST) Watch(ctx context.Context, options *metainternalversion.ListOptio
 	return customW, nil
 }
 
+// wmResultChan returns the result channel of a WorkloadMonitor watcher, or a nil
+// channel (which blocks forever in select) if the watcher is nil.
+func wmResultChan(w watch.Interface) <-chan watch.Event {
+	if w == nil {
+		return nil
+	}
+	return w.ResultChan()
+}
+
 // customWatcher wraps the original watcher and filters/converts events
 type customWatcher struct {
 	resultChan chan watch.Event
@@ -983,12 +1158,21 @@ func filterPrefixedMap(original map[string]string, prefix string) map[string]str
 	return processed
 }
 
-// ConvertHelmReleaseToApplication converts a HelmRelease to an Application
+// ConvertHelmReleaseToApplication converts a HelmRelease to an Application.
 func (r *REST) ConvertHelmReleaseToApplication(ctx context.Context, hr *helmv2.HelmRelease) (appsv1alpha1.Application, error) {
+	return r.ConvertHelmReleaseToApplicationWithMonitor(ctx, hr, nil)
+}
+
+// ConvertHelmReleaseToApplicationWithMonitor converts a HelmRelease to an
+// Application, optionally overriding the cached copy of a WorkloadMonitor with
+// a fresher version received from the watch client. This prevents the emitted
+// Application object from carrying stale WorkloadsReady data when r.c (cache)
+// lags behind r.w (watch).
+func (r *REST) ConvertHelmReleaseToApplicationWithMonitor(ctx context.Context, hr *helmv2.HelmRelease, freshMonitor *cozyv1alpha1.WorkloadMonitor) (appsv1alpha1.Application, error) {
 	klog.V(6).Infof("Converting HelmRelease to Application for resource %s", hr.GetName())
 
 	// Convert HelmRelease struct to Application struct
-	app, err := r.convertHelmReleaseToApplication(ctx, hr)
+	app, err := r.convertHelmReleaseToApplication(ctx, hr, freshMonitor)
 	if err != nil {
 		klog.Errorf("Error converting from HelmRelease to Application: %v", err)
 		return appsv1alpha1.Application{}, err
@@ -1049,6 +1233,19 @@ func validateNoInternalKeys(values *apiextv1.JSON) error {
 // chart-generated resource suffixes within the 63-char DNS-1035 label limit.
 const maxHelmReleaseName = 53
 
+// maxNamespaceName is the DNS-1123 label limit for Kubernetes namespace names.
+// The tenant Helm chart creates a Namespace whose name is the computed
+// workload namespace (parent namespace + "-" + tenant name), so the total
+// must fit inside a single 63-char DNS-1123 label.
+const maxNamespaceName = 63
+
+// validateNameFormat checks an Application name against DNS-1035 and any
+// kind-specific format rules (e.g. Tenant names must be alphanumeric — see
+// validation.ValidateApplicationName for the reasoning).
+func (r *REST) validateNameFormat(name string) field.ErrorList {
+	return validation.ValidateApplicationName(name, r.kindName, field.NewPath("metadata").Child("name"))
+}
+
 // validateNameLength checks that the application name won't exceed Kubernetes limits.
 // prefix + name must fit within the Helm release name limit (53 chars).
 func (r *REST) validateNameLength(name string) field.ErrorList {
@@ -1070,8 +1267,29 @@ func (r *REST) validateNameLength(name string) field.ErrorList {
 	return allErrs
 }
 
-// convertHelmReleaseToApplication implements the actual conversion logic
-func (r *REST) convertHelmReleaseToApplication(ctx context.Context, hr *helmv2.HelmRelease) (appsv1alpha1.Application, error) {
+// validateTenantNamespaceLength checks that the computed workload namespace
+// for a Tenant application fits within the DNS-1123 label limit. The namespace
+// is formed by dash-joining the parent namespace with the tenant name, so
+// deep nesting can exceed the limit even when each individual name passes the
+// per-name Helm release length check.
+func (r *REST) validateTenantNamespaceLength(currentNamespace, tenantName string) field.ErrorList {
+	fldPath := field.NewPath("metadata").Child("name")
+	allErrs := field.ErrorList{}
+
+	computed := r.computeTenantNamespace(currentNamespace, tenantName)
+	if len(computed) > maxNamespaceName {
+		allErrs = append(allErrs, field.Invalid(fldPath, tenantName,
+			fmt.Sprintf("computed tenant namespace %q would be %d characters, which exceeds the %d-character Kubernetes namespace limit; shorten the tenant name or reduce ancestor nesting depth",
+				computed, len(computed), maxNamespaceName)))
+	}
+	return allErrs
+}
+
+// convertHelmReleaseToApplication implements the actual conversion logic.
+// The optional freshMonitor is used to override the cache copy of a
+// WorkloadMonitor when a newer version was delivered via the watch client —
+// see ConvertHelmReleaseToApplicationWithMonitor for the rationale.
+func (r *REST) convertHelmReleaseToApplication(ctx context.Context, hr *helmv2.HelmRelease, freshMonitor *cozyv1alpha1.WorkloadMonitor) (appsv1alpha1.Application, error) {
 	// Filter out internal keys (starting with "_") from spec
 	filteredSpec := filterInternalKeys(hr.Spec.Values)
 
@@ -1108,10 +1326,81 @@ func (r *REST) convertHelmReleaseToApplication(ctx context.Context, hr *helmv2.H
 			})
 		}
 	}
+	// Enrich conditions with WorkloadMonitor operational status.
+	// Tenant workloads live in a child namespace (computed from the Tenant name),
+	// not in the same namespace as the owning HelmRelease — look there instead.
+	workloadsNS := hr.Namespace
+	if r.kindName == "Tenant" {
+		workloadsNS = r.computeTenantNamespace(hr.Namespace, app.Name)
+	}
+	ws, wsErr := r.getWorkloadsOperational(ctx, workloadsNS, app.Name, freshMonitor)
+	// Derive a stable LastTransitionTime: use the owning HelmRelease's own
+	// condition update time (or CreationTimestamp as a floor) so that repeated
+	// conversions of the same underlying state produce identical timestamps.
+	wrTransition := hr.CreationTimestamp
+	for _, c := range hr.GetConditions() {
+		if c.LastTransitionTime.After(wrTransition.Time) {
+			wrTransition = c.LastTransitionTime
+		}
+	}
+	if ws.transitionTime.After(wrTransition.Time) {
+		wrTransition = ws.transitionTime
+	}
+	if wrTransition.IsZero() {
+		// Fallback for objects that somehow have no timestamps at all
+		// (e.g. hand-crafted test fixtures). In production HelmReleases
+		// always carry a CreationTimestamp, so the stable branch above
+		// is used.
+		wrTransition = metav1.Now()
+	}
+	if wsErr != nil {
+		// Fail-open: if we can't query WorkloadMonitors (e.g., informer cache not ready),
+		// don't override Ready. Prefer operational availability over safety.
+		// The WorkloadsReady=Unknown condition still signals the issue to the user.
+		klog.Warningf("Failed to check workload monitors for %s/%s: %v", workloadsNS, app.Name, wsErr)
+		conditions = append(conditions, metav1.Condition{
+			Type:               "WorkloadsReady",
+			Status:             metav1.ConditionUnknown,
+			LastTransitionTime: wrTransition,
+			Reason:             "Error",
+			Message:            fmt.Sprintf("Failed to check workload status: %v", wsErr),
+		})
+	} else if ws.found {
+		workloadsCondition := metav1.Condition{
+			Type:               "WorkloadsReady",
+			LastTransitionTime: wrTransition,
+			Reason:             "WorkloadMonitorCheck",
+		}
+		switch {
+		case !ws.operational:
+			// Concrete failure takes priority over unknown/pending state
+			workloadsCondition.Status = metav1.ConditionFalse
+			workloadsCondition.Message = "One or more workloads are not operational"
+		case ws.unknown:
+			workloadsCondition.Status = metav1.ConditionUnknown
+			workloadsCondition.Reason = "Pending"
+			workloadsCondition.Message = "One or more workloads have not been reconciled yet"
+		default:
+			workloadsCondition.Status = metav1.ConditionTrue
+			workloadsCondition.Message = "All workloads are operational"
+		}
+		conditions = append(conditions, workloadsCondition)
+
+		// Intentionally do NOT override the Ready condition based on WorkloadsReady.
+		// Ready continues to reflect HelmRelease state only, which:
+		//   - preserves backward compatibility with existing tooling (kubectl wait,
+		//     GitOps health checks) that expect Ready to match HelmRelease
+		//   - avoids false-negative Ready=False during normal startup windows where
+		//     pods are still coming up but WorkloadMonitor has already reported
+		//     Operational=false due to availableReplicas < MinReplicas
+		// WorkloadsReady is a separate condition that surfaces workload health
+		// independently — users and dashboards can observe it for operational visibility.
+	}
+
 	app.SetConditions(conditions)
 
 	// Add namespace field for Tenant applications
-	if r.kindName == "Tenant" {
+	if r.kindName == validation.TenantKind {
 		app.Status.Namespace = r.computeTenantNamespace(hr.Namespace, app.Name)
 		externalIPsCount, err := r.countTenantExternalIPs(ctx, app.Status.Namespace)
 		if err != nil {
@@ -1124,6 +1413,79 @@ func (r *REST) convertHelmReleaseToApplication(ctx context.Context, hr *helmv2.H
 	return app, nil
 }
 
+// workloadsStatus holds the aggregated operational status of WorkloadMonitors.
+type workloadsStatus struct {
+	operational bool
+	found       bool
+	unknown     bool // true when at least one monitor has nil Operational (not yet reconciled)
+	// transitionTime is the most recent metadata update time across the
+	// matching monitors. Used as WorkloadsReady.LastTransitionTime so that
+	// repeated conversions for the same underlying state produce stable
+	// timestamps (preserving the Kubernetes contract that identical
+	// resource versions represent identical content).
+	transitionTime metav1.Time
+}
+
+// getWorkloadsOperational checks WorkloadMonitor resources for an application and returns
+// aggregated operational status. If no monitors exist, returns found=false.
+// When freshOverride is non-nil, its status replaces the cached copy for the
+// corresponding monitor — this keeps the result consistent with watch events
+// when the cache (r.c) lags behind the watch client (r.w).
+func (r *REST) getWorkloadsOperational(ctx context.Context, namespace, appName string, freshOverride *cozyv1alpha1.WorkloadMonitor) (workloadsStatus, error) {
+	monitors := &cozyv1alpha1.WorkloadMonitorList{}
+	if err := r.c.List(ctx, monitors,
+		client.InNamespace(namespace),
+		client.MatchingLabels{
+			appsv1alpha1.ApplicationKindLabel:  r.kindName,
+			appsv1alpha1.ApplicationGroupLabel: r.gvk.Group,
+			appsv1alpha1.ApplicationNameLabel:  appName,
+		},
+	); err != nil {
+		return workloadsStatus{}, err
+	}
+	// Ensure the freshOverride is represented in the aggregation even when
+	// the cache has not yet observed it (brand-new resource) or is behind.
+	replaced := false
+	if freshOverride != nil {
+		for i := range monitors.Items {
+			if monitors.Items[i].UID == freshOverride.UID ||
+				(monitors.Items[i].Name == freshOverride.Name && monitors.Items[i].Namespace == freshOverride.Namespace) {
+				monitors.Items[i] = *freshOverride
+				replaced = true
+				break
+			}
+		}
+		if !replaced {
+			monitors.Items = append(monitors.Items, *freshOverride)
+		}
+	}
+	if len(monitors.Items) == 0 {
+		return workloadsStatus{operational: true, found: false}, nil
+	}
+	operational := true
+	unknown := false
+	var latest metav1.Time
+	for _, m := range monitors.Items {
+		if m.Status.Operational == nil {
+			unknown = true
+		} else if !*m.Status.Operational {
+			operational = false
+		}
+		// Pick the most recent monitor mtime as a stable transition time.
+		if t := latestMonitorTime(&m); t.After(latest.Time) {
+			latest = t
+		}
+	}
+	return workloadsStatus{operational: operational, found: true, unknown: unknown, transitionTime: latest}, nil
+}
+
+// latestMonitorTime returns the most recent timestamp associated with a
+// WorkloadMonitor — currently only the object creation time is guaranteed.
+// Status does not carry a transition time, so we fall back to CreationTimestamp.
+func latestMonitorTime(m *cozyv1alpha1.WorkloadMonitor) metav1.Time {
+	return m.CreationTimestamp
+}
+
 // convertApplicationToHelmRelease implements the actual conversion logic
 func (r *REST) convertApplicationToHelmRelease(app *appsv1alpha1.Application) (*helmv2.HelmRelease, error) {
 	helmRelease := &helmv2.HelmRelease{
@@ -1166,6 +1528,24 @@ func (r *REST) convertApplicationToHelmRelease(app *appsv1alpha1.Application) (*
 		},
 	}
 
+	// Per-Application HelmRelease wait budget. The mechanism is generic:
+	// an ApplicationDefinition that sets
+	// release.cozystack.io/helm-install-timeout gets Install.Timeout and
+	// Upgrade.Timeout populated from ReleaseConfig.HelmInstallTimeout
+	// (parsed at startup). Applications that leave it unset keep flux
+	// defaults so their failed installs remediate on the normal cadence.
+	// Today only kubernetes-rd carries the annotation because the
+	// Kubernetes Application's parent chart contains CAPI/Kamaji
+	// resources whose admin-kubeconfig Secret is provisioned
+	// asynchronously and Kamaji cold-start routinely exceeds flux's
+	// default wait budget; any future kind with the same shape can opt
+	// in by setting the same annotation.
+	if r.releaseConfig.HelmInstallTimeout > 0 {
+		timeout := metav1.Duration{Duration: r.releaseConfig.HelmInstallTimeout}
+		helmRelease.Spec.Install.Timeout = &timeout
+		helmRelease.Spec.Upgrade.Timeout = &timeout
+	}
+
 	return helmRelease, nil
 }
 
diff --git a/pkg/registry/apps/application/rest_conditions_test.go b/pkg/registry/apps/application/rest_conditions_test.go
new file mode 100644
index 00000000..185f59b2
--- /dev/null
+++ b/pkg/registry/apps/application/rest_conditions_test.go
@@ -0,0 +1,331 @@
+package application
+
+import (
+	"context"
+	"testing"
+
+	cozyv1alpha1 "github.com/cozystack/cozystack/api/v1alpha1"
+	helmv2 "github.com/fluxcd/helm-controller/api/v2"
+	appsv1alpha1 "github.com/cozystack/cozystack/pkg/apis/apps/v1alpha1"
+	"github.com/cozystack/cozystack/pkg/config"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/apimachinery/pkg/runtime"
+	"k8s.io/utils/ptr"
+	"sigs.k8s.io/controller-runtime/pkg/client/fake"
+)
+
+func findCondition(conditions []metav1.Condition, condType string) *metav1.Condition {
+	for i := range conditions {
+		if conditions[i].Type == condType {
+			return &conditions[i]
+		}
+	}
+	return nil
+}
+
+func makeHelmRelease(name, namespace string) *helmv2.HelmRelease {
+	hr := &helmv2.HelmRelease{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      name,
+			Namespace: namespace,
+			Labels: map[string]string{
+				ApplicationKindLabel:  "PostgreSQL",
+				ApplicationGroupLabel: "apps.cozystack.io",
+				ApplicationNameLabel:  "mydb",
+			},
+		},
+	}
+	return hr
+}
+
+func TestConvertConditions_WorkloadsReadyAdded(t *testing.T) {
+	monitor := &cozyv1alpha1.WorkloadMonitor{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "mon-1",
+			Namespace: "default",
+			Labels: map[string]string{
+				appsv1alpha1.ApplicationKindLabel:  "PostgreSQL",
+				appsv1alpha1.ApplicationGroupLabel: "apps.cozystack.io",
+				appsv1alpha1.ApplicationNameLabel:  "mydb",
+			},
+		},
+		Status: cozyv1alpha1.WorkloadMonitorStatus{
+			Operational: ptr.To(true),
+		},
+	}
+
+	r := newTestRESTWithSchemes(monitor)
+	hr := makeHelmRelease("postgresql-mydb", "default")
+	hr.Status.Conditions = []metav1.Condition{
+		{Type: "Ready", Status: metav1.ConditionTrue, Reason: "Succeeded", Message: "Release applied"},
+		{Type: "Released", Status: metav1.ConditionTrue, Reason: "Succeeded", Message: "Released"},
+	}
+
+	app, err := r.convertHelmReleaseToApplication(context.TODO(), hr, nil)
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+
+	wc := findCondition(app.GetConditions(), "WorkloadsReady")
+	if wc == nil {
+		t.Fatal("expected WorkloadsReady condition to be present")
+	}
+	if wc.Status != metav1.ConditionTrue {
+		t.Errorf("expected WorkloadsReady=True, got %s", wc.Status)
+	}
+
+	rc := findCondition(app.GetConditions(), "Ready")
+	if rc == nil {
+		t.Fatal("expected Ready condition to be present")
+	}
+	if rc.Status != metav1.ConditionTrue {
+		t.Errorf("expected Ready=True, got %s", rc.Status)
+	}
+}
+
+func TestConvertConditions_ReadyNotOverriddenWhenWorkloadsNotReady(t *testing.T) {
+	// Ready must reflect HelmRelease state only. WorkloadsReady is a separate
+	// signal that users/dashboards can observe independently.
+	monitor := &cozyv1alpha1.WorkloadMonitor{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "mon-1",
+			Namespace: "default",
+			Labels: map[string]string{
+				appsv1alpha1.ApplicationKindLabel:  "PostgreSQL",
+				appsv1alpha1.ApplicationGroupLabel: "apps.cozystack.io",
+				appsv1alpha1.ApplicationNameLabel:  "mydb",
+			},
+		},
+		Status: cozyv1alpha1.WorkloadMonitorStatus{
+			Operational: ptr.To(false),
+		},
+	}
+
+	r := newTestRESTWithSchemes(monitor)
+	hr := makeHelmRelease("postgresql-mydb", "default")
+	hr.Status.Conditions = []metav1.Condition{
+		{Type: "Ready", Status: metav1.ConditionTrue, Reason: "Succeeded", Message: "Release applied"},
+	}
+
+	app, err := r.convertHelmReleaseToApplication(context.TODO(), hr, nil)
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+
+	wc := findCondition(app.GetConditions(), "WorkloadsReady")
+	if wc == nil {
+		t.Fatal("expected WorkloadsReady condition to be present")
+	}
+	if wc.Status != metav1.ConditionFalse {
+		t.Errorf("expected WorkloadsReady=False, got %s", wc.Status)
+	}
+
+	rc := findCondition(app.GetConditions(), "Ready")
+	if rc == nil {
+		t.Fatal("expected Ready condition to be present")
+	}
+	if rc.Status != metav1.ConditionTrue {
+		t.Errorf("expected Ready=True (reflects HelmRelease only), got %s", rc.Status)
+	}
+	if rc.Reason != "Succeeded" {
+		t.Errorf("expected Ready.Reason=Succeeded (unchanged), got %s", rc.Reason)
+	}
+}
+
+func TestConvertConditions_NoOverrideWhenNoMonitors(t *testing.T) {
+	r := newTestRESTWithSchemes()
+	hr := makeHelmRelease("postgresql-mydb", "default")
+	hr.Status.Conditions = []metav1.Condition{
+		{Type: "Ready", Status: metav1.ConditionTrue, Reason: "Succeeded", Message: "Release applied"},
+	}
+
+	app, err := r.convertHelmReleaseToApplication(context.TODO(), hr, nil)
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+
+	wc := findCondition(app.GetConditions(), "WorkloadsReady")
+	if wc != nil {
+		t.Error("expected no WorkloadsReady condition when no monitors exist")
+	}
+
+	rc := findCondition(app.GetConditions(), "Ready")
+	if rc == nil {
+		t.Fatal("expected Ready condition to be present")
+	}
+	if rc.Status != metav1.ConditionTrue {
+		t.Errorf("expected Ready=True unchanged, got %s", rc.Status)
+	}
+}
+
+func TestConvertConditions_ReadyStaysTrue_WhenAllOperational(t *testing.T) {
+	monitor := &cozyv1alpha1.WorkloadMonitor{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "mon-1",
+			Namespace: "default",
+			Labels: map[string]string{
+				appsv1alpha1.ApplicationKindLabel:  "PostgreSQL",
+				appsv1alpha1.ApplicationGroupLabel: "apps.cozystack.io",
+				appsv1alpha1.ApplicationNameLabel:  "mydb",
+			},
+		},
+		Status: cozyv1alpha1.WorkloadMonitorStatus{
+			Operational: ptr.To(true),
+		},
+	}
+
+	r := newTestRESTWithSchemes(monitor)
+	hr := makeHelmRelease("postgresql-mydb", "default")
+	hr.Status.Conditions = []metav1.Condition{
+		{Type: "Ready", Status: metav1.ConditionTrue, Reason: "Succeeded", Message: "Release applied"},
+		{Type: "Released", Status: metav1.ConditionTrue, Reason: "Succeeded", Message: "Released"},
+	}
+
+	app, err := r.convertHelmReleaseToApplication(context.TODO(), hr, nil)
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+
+	rc := findCondition(app.GetConditions(), "Ready")
+	if rc == nil {
+		t.Fatal("expected Ready condition")
+	}
+	if rc.Status != metav1.ConditionTrue {
+		t.Errorf("expected Ready=True when all workloads operational, got %s", rc.Status)
+	}
+	if rc.Reason != "Succeeded" {
+		t.Errorf("expected Ready.Reason=Succeeded (unchanged), got %s", rc.Reason)
+	}
+}
+
+func TestConvertConditions_WorkloadsReadyTimestampIsNonZero(t *testing.T) {
+	monitor := &cozyv1alpha1.WorkloadMonitor{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "mon-1",
+			Namespace: "default",
+			Labels: map[string]string{
+				appsv1alpha1.ApplicationKindLabel:  "PostgreSQL",
+				appsv1alpha1.ApplicationGroupLabel: "apps.cozystack.io",
+				appsv1alpha1.ApplicationNameLabel:  "mydb",
+			},
+		},
+		Status: cozyv1alpha1.WorkloadMonitorStatus{
+			Operational: ptr.To(true),
+		},
+	}
+
+	r := newTestRESTWithSchemes(monitor)
+	hr := makeHelmRelease("postgresql-mydb", "default")
+	// No Ready condition — HR still being reconciled
+	hr.Status.Conditions = []metav1.Condition{}
+
+	app, err := r.convertHelmReleaseToApplication(context.TODO(), hr, nil)
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+
+	wc := findCondition(app.GetConditions(), "WorkloadsReady")
+	if wc == nil {
+		t.Fatal("expected WorkloadsReady condition")
+	}
+	if wc.LastTransitionTime.IsZero() {
+		t.Error("expected non-zero LastTransitionTime")
+	}
+}
+
+func TestConvertConditions_WorkloadsReadyUnknownWhenNilOperational(t *testing.T) {
+	monitor := &cozyv1alpha1.WorkloadMonitor{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "mon-1",
+			Namespace: "default",
+			Labels: map[string]string{
+				appsv1alpha1.ApplicationKindLabel:  "PostgreSQL",
+				appsv1alpha1.ApplicationGroupLabel: "apps.cozystack.io",
+				appsv1alpha1.ApplicationNameLabel:  "mydb",
+			},
+		},
+		Status: cozyv1alpha1.WorkloadMonitorStatus{
+			Operational: nil, // Not yet reconciled
+		},
+	}
+
+	r := newTestRESTWithSchemes(monitor)
+	hr := makeHelmRelease("postgresql-mydb", "default")
+	hr.Status.Conditions = []metav1.Condition{
+		{Type: "Ready", Status: metav1.ConditionTrue, Reason: "Succeeded", Message: "ok"},
+	}
+
+	app, err := r.convertHelmReleaseToApplication(context.TODO(), hr, nil)
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+
+	wc := findCondition(app.GetConditions(), "WorkloadsReady")
+	if wc == nil {
+		t.Fatal("expected WorkloadsReady condition")
+	}
+	if wc.Status != metav1.ConditionUnknown {
+		t.Errorf("expected WorkloadsReady=Unknown for nil Operational, got %s", wc.Status)
+	}
+
+	// Ready should NOT be overridden for unknown — prefer availability during startup
+	rc := findCondition(app.GetConditions(), "Ready")
+	if rc == nil {
+		t.Fatal("expected Ready condition")
+	}
+	if rc.Status != metav1.ConditionTrue {
+		t.Errorf("expected Ready=True when workloads unknown (pending), got %s", rc.Status)
+	}
+}
+
+func TestConvertConditions_WorkloadsReadyUnknownOnError(t *testing.T) {
+	// Create a client with only HelmRelease scheme — WorkloadMonitor List will fail
+	scheme := runtime.NewScheme()
+	_ = helmv2.AddToScheme(scheme)
+	// Deliberately NOT registering cozyv1alpha1 so that List returns an error
+
+	c := fake.NewClientBuilder().WithScheme(scheme).Build()
+	r := NewREST(c, nil, &config.Resource{
+		Application: config.ApplicationConfig{
+			Kind:     "PostgreSQL",
+			Plural:   "postgresqls",
+			Singular: "postgresql",
+		},
+		Release: config.ReleaseConfig{
+			Prefix: "postgresql-",
+		},
+	})
+
+	hr := makeHelmRelease("postgresql-mydb", "default")
+	hr.CreationTimestamp = metav1.Now()
+	hr.Status.Conditions = []metav1.Condition{
+		{Type: "Ready", Status: metav1.ConditionTrue, Reason: "Succeeded", Message: "ok"},
+	}
+
+	app, err := r.convertHelmReleaseToApplication(context.TODO(), hr, nil)
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+
+	wc := findCondition(app.GetConditions(), "WorkloadsReady")
+	if wc == nil {
+		t.Fatal("expected WorkloadsReady condition with Unknown status on error")
+	}
+	if wc.Status != metav1.ConditionUnknown {
+		t.Errorf("expected WorkloadsReady=Unknown, got %s", wc.Status)
+	}
+	if wc.Reason != "Error" {
+		t.Errorf("expected reason=Error, got %s", wc.Reason)
+	}
+
+	// Ready should NOT be overridden on error (fail-open: prefer availability)
+	rc := findCondition(app.GetConditions(), "Ready")
+	if rc == nil {
+		t.Fatal("expected Ready condition")
+	}
+	if rc.Status != metav1.ConditionTrue {
+		t.Errorf("expected Ready=True (fail-open on error), got %s", rc.Status)
+	}
+}
+
+
diff --git a/pkg/registry/apps/application/rest_timeout_test.go b/pkg/registry/apps/application/rest_timeout_test.go
new file mode 100644
index 00000000..85ac0b97
--- /dev/null
+++ b/pkg/registry/apps/application/rest_timeout_test.go
@@ -0,0 +1,116 @@
+package application
+
+import (
+	"testing"
+	"time"
+
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+
+	appsv1alpha1 "github.com/cozystack/cozystack/pkg/apis/apps/v1alpha1"
+	"github.com/cozystack/cozystack/pkg/config"
+)
+
+func newRESTForTimeout(kind, prefix string, helmInstallTimeout time.Duration) *REST {
+	return &REST{
+		kindName: kind,
+		releaseConfig: config.ReleaseConfig{
+			Prefix: prefix,
+			ChartRef: config.ChartRefConfig{
+				Kind:      "HelmChart",
+				Name:      "x",
+				Namespace: "cozy-system",
+			},
+			HelmInstallTimeout: helmInstallTimeout,
+		},
+	}
+}
+
+// Table-driven: every Application kind carries a per-CRD HelmRelease wait
+// budget. The Kubernetes kind's parent chart contains CAPI/Kamaji resources
+// whose admin-kubeconfig Secret is provisioned asynchronously, so its
+// ApplicationDefinition sets release.cozystack.io/helm-install-timeout=15m
+// (or longer). Other kinds leave the annotation unset and keep flux defaults
+// so their failed installs remediate on the normal cadence. The test must
+// cover both paths: a kind with the timeout set and one without.
+func TestConvertApplicationToHelmRelease_AppliesReleaseConfigTimeout(t *testing.T) {
+	cases := []struct {
+		name       string
+		kind       string
+		prefix     string
+		configured time.Duration
+		wantSet    bool
+	}{
+		{
+			name:       "Kubernetes kind with 15m configured gets Install and Upgrade Timeout",
+			kind:       "Kubernetes",
+			prefix:     "kubernetes-",
+			configured: 15 * time.Minute,
+			wantSet:    true,
+		},
+		{
+			// Fictional kind on purpose: the test is about the unset path
+			// regardless of which real Application kind ends up needing a
+			// timeout override. Using a real kind name would create false
+			// coupling to that Application's ApplicationDefinition.
+			name:       "unrelated kind without configured timeout keeps flux defaults",
+			kind:       "PlaceholderKindForDefaults",
+			prefix:     "placeholder-",
+			configured: 0,
+			wantSet:    false,
+		},
+		{
+			name:       "arbitrary future kind with 20m configured gets 20m",
+			kind:       "TalosCluster",
+			prefix:     "talos-",
+			configured: 20 * time.Minute,
+			wantSet:    true,
+		},
+	}
+
+	for _, tc := range cases {
+		t.Run(tc.name, func(t *testing.T) {
+			r := newRESTForTimeout(tc.kind, tc.prefix, tc.configured)
+			app := &appsv1alpha1.Application{
+				ObjectMeta: metav1.ObjectMeta{Name: "example", Namespace: "tenant-root"},
+			}
+
+			hr, err := r.convertApplicationToHelmRelease(app)
+			if err != nil {
+				t.Fatalf("convertApplicationToHelmRelease returned error: %v", err)
+			}
+
+			if hr.Spec.Install == nil || hr.Spec.Upgrade == nil {
+				t.Fatalf("Spec.Install/Upgrade must be non-nil")
+			}
+
+			if tc.wantSet {
+				if hr.Spec.Install.Timeout == nil {
+					t.Fatalf("Spec.Install.Timeout must be set when HelmInstallTimeout=%v", tc.configured)
+				}
+				if hr.Spec.Install.Timeout.Duration != tc.configured {
+					t.Errorf("Spec.Install.Timeout = %v, want %v", hr.Spec.Install.Timeout.Duration, tc.configured)
+				}
+				if hr.Spec.Upgrade.Timeout == nil {
+					t.Fatalf("Spec.Upgrade.Timeout must be set when HelmInstallTimeout=%v", tc.configured)
+				}
+				if hr.Spec.Upgrade.Timeout.Duration != tc.configured {
+					t.Errorf("Spec.Upgrade.Timeout = %v, want %v", hr.Spec.Upgrade.Timeout.Duration, tc.configured)
+				}
+			} else {
+				if hr.Spec.Install.Timeout != nil {
+					t.Errorf("Spec.Install.Timeout must be nil when HelmInstallTimeout=0, got %v", hr.Spec.Install.Timeout.Duration)
+				}
+				if hr.Spec.Upgrade.Timeout != nil {
+					t.Errorf("Spec.Upgrade.Timeout must be nil when HelmInstallTimeout=0, got %v", hr.Spec.Upgrade.Timeout.Duration)
+				}
+			}
+
+			if hr.Spec.Install.Remediation == nil || hr.Spec.Install.Remediation.Retries != -1 {
+				t.Errorf("Spec.Install.Remediation.Retries must remain -1, got %+v", hr.Spec.Install.Remediation)
+			}
+			if hr.Spec.Upgrade.Remediation == nil || hr.Spec.Upgrade.Remediation.Retries != -1 {
+				t.Errorf("Spec.Upgrade.Remediation.Retries must remain -1, got %+v", hr.Spec.Upgrade.Remediation)
+			}
+		})
+	}
+}
diff --git a/pkg/registry/apps/application/rest_validation_test.go b/pkg/registry/apps/application/rest_validation_test.go
index c5d331b2..b806161c 100644
--- a/pkg/registry/apps/application/rest_validation_test.go
+++ b/pkg/registry/apps/application/rest_validation_test.go
@@ -17,12 +17,207 @@ limitations under the License.
 package application
 
 import (
+	"context"
+	"fmt"
 	"strings"
 	"testing"
 
+	helmv2 "github.com/fluxcd/helm-controller/api/v2"
+	corev1 "k8s.io/api/core/v1"
+	apierrors "k8s.io/apimachinery/pkg/api/errors"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/apimachinery/pkg/runtime"
+	"k8s.io/apimachinery/pkg/runtime/schema"
+	"k8s.io/apiserver/pkg/endpoints/request"
+	"k8s.io/apiserver/pkg/registry/rest"
+	"sigs.k8s.io/controller-runtime/pkg/client/fake"
+
+	appsv1alpha1 "github.com/cozystack/cozystack/pkg/apis/apps/v1alpha1"
+	"github.com/cozystack/cozystack/pkg/apis/apps/validation"
 	"github.com/cozystack/cozystack/pkg/config"
 )
 
+func TestValidateNameFormat(t *testing.T) {
+	tests := []struct {
+		name      string
+		kindName  string
+		appName   string
+		wantError bool
+	}{
+		// Non-tenant kinds follow DNS-1035 only — hyphens are allowed.
+		{"non-tenant accepts hyphen", "MySQL", "my-db", false},
+		{"non-tenant accepts double hyphen", "MySQL", "my--db", false},
+		{"non-tenant rejects uppercase", "MySQL", "MyDB", true},
+
+		// Tenant kind enforces alphanumeric-only — see
+		// packages/apps/tenant/templates/_helpers.tpl for the reason.
+		{"tenant accepts alphanumeric", validation.TenantKind, "foo", false},
+		{"tenant accepts digits", validation.TenantKind, "foo123", false},
+		{"tenant rejects single hyphen", validation.TenantKind, "foo-bar", true},
+		{"tenant rejects leading hyphen", validation.TenantKind, "-foo", true},
+		{"tenant rejects trailing hyphen", validation.TenantKind, "foo-", true},
+		{"tenant rejects uppercase", validation.TenantKind, "Foo", true},
+		{"tenant rejects underscore", validation.TenantKind, "foo_bar", true},
+		{"tenant rejects empty", validation.TenantKind, "", true},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			r := &REST{kindName: tt.kindName}
+
+			errs := r.validateNameFormat(tt.appName)
+
+			if tt.wantError && len(errs) == 0 {
+				t.Errorf("expected error for name %q (kind=%q), got none", tt.appName, tt.kindName)
+			}
+			if !tt.wantError && len(errs) > 0 {
+				t.Errorf("unexpected error for name %q (kind=%q): %v", tt.appName, tt.kindName, errs)
+			}
+		})
+	}
+}
+
+// TestUpdate_ForceAllowCreate_RejectsTenantDashName pins the wiring from the
+// Update → Create fall-through path. When a user runs `kubectl apply` and
+// the object does not yet exist, Kubernetes routes the request through
+// REST.Update with forceAllowCreate=true, which delegates to REST.Create
+// (rest.go:452). Without this test, a future refactor could quietly reroute
+// that delegation and bypass the tenant name check — unit tests of the
+// pure r.validateNameFormat method would still pass while upsert-style
+// kubectl apply regressed back to accepting tenant names with dashes.
+func TestUpdate_ForceAllowCreate_RejectsTenantDashName(t *testing.T) {
+	scheme := runtime.NewScheme()
+	if err := helmv2.AddToScheme(scheme); err != nil {
+		t.Fatalf("register helmv2 scheme: %v", err)
+	}
+	// Register the dynamic Tenant kind so the Application type round-trips
+	// through the scheme the same way the real aggregated API server wires
+	// it at startup.
+	resourceCfg := &config.ResourceConfig{
+		Resources: []config.Resource{
+			{
+				Application: config.ApplicationConfig{Kind: validation.TenantKind},
+			},
+		},
+	}
+	if err := appsv1alpha1.RegisterDynamicTypes(scheme, resourceCfg); err != nil {
+		t.Fatalf("register dynamic Tenant type: %v", err)
+	}
+
+	fakeClient := fake.NewClientBuilder().WithScheme(scheme).Build()
+
+	r := &REST{
+		c: fakeClient,
+		gvr: schema.GroupVersionResource{
+			Group:    appsv1alpha1.GroupName,
+			Version:  "v1alpha1",
+			Resource: "tenants",
+		},
+		gvk: schema.GroupVersionKind{
+			Group:   appsv1alpha1.GroupName,
+			Version: "v1alpha1",
+			Kind:    validation.TenantKind,
+		},
+		kindName: validation.TenantKind,
+		releaseConfig: config.ReleaseConfig{
+			Prefix: "tenant-",
+		},
+	}
+
+	newApp := &appsv1alpha1.Application{
+		TypeMeta: metav1.TypeMeta{
+			APIVersion: "apps.cozystack.io/v1alpha1",
+			Kind:       validation.TenantKind,
+		},
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "foo-bar",
+			Namespace: "tenant-root",
+		},
+	}
+
+	ctx := request.WithNamespace(context.Background(), "tenant-root")
+
+	_, _, err := r.Update(
+		ctx,
+		"foo-bar",
+		rest.DefaultUpdatedObjectInfo(newApp),
+		nil,                      // createValidation
+		nil,                      // updateValidation
+		true,                     // forceAllowCreate → routes through Create on NotFound
+		&metav1.UpdateOptions{},
+	)
+	if err == nil {
+		t.Fatalf("expected Update to reject tenant name with dashes, got no error")
+	}
+	if !apierrors.IsInvalid(err) {
+		t.Errorf("expected Invalid status error, got %T: %v", err, err)
+	}
+	if !strings.Contains(err.Error(), "tenant names must") {
+		t.Errorf("expected tenant-specific error in %q", err.Error())
+	}
+}
+
+// TestConvertHelmReleaseToApplication_TenantNamespaceKindGate pins the
+// behavior that convertHelmReleaseToApplication fills Status.Namespace only
+// when the kind is Tenant. This path is gated on r.kindName matching a
+// specific literal — the test uses the validation.TenantKind constant so
+// that if the source of truth for the tenant kind string is ever renamed,
+// the gate and the constant drift together (or the test fails).
+func TestConvertHelmReleaseToApplication_TenantNamespaceKindGate(t *testing.T) {
+	scheme := runtime.NewScheme()
+	if err := helmv2.AddToScheme(scheme); err != nil {
+		t.Fatalf("register helmv2 scheme: %v", err)
+	}
+	if err := corev1.AddToScheme(scheme); err != nil {
+		t.Fatalf("register corev1 scheme: %v", err)
+	}
+
+	fakeClient := fake.NewClientBuilder().WithScheme(scheme).Build()
+
+	hr := &helmv2.HelmRelease{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "tenant-foo",
+			Namespace: "tenant-root",
+		},
+	}
+
+	t.Run("tenant kind fills Status.Namespace", func(t *testing.T) {
+		r := &REST{
+			c:        fakeClient,
+			kindName: validation.TenantKind,
+			releaseConfig: config.ReleaseConfig{
+				Prefix: "tenant-",
+			},
+		}
+
+		app, err := r.convertHelmReleaseToApplication(context.Background(), hr, nil)
+		if err != nil {
+			t.Fatalf("convertHelmReleaseToApplication: %v", err)
+		}
+		if app.Status.Namespace == "" {
+			t.Errorf("expected Status.Namespace to be populated for tenant kind, got empty")
+		}
+	})
+
+	t.Run("non-tenant kind leaves Status.Namespace empty", func(t *testing.T) {
+		r := &REST{
+			c:        fakeClient,
+			kindName: "MySQL",
+			releaseConfig: config.ReleaseConfig{
+				Prefix: "mysql-",
+			},
+		}
+
+		app, err := r.convertHelmReleaseToApplication(context.Background(), hr, nil)
+		if err != nil {
+			t.Fatalf("convertHelmReleaseToApplication: %v", err)
+		}
+		if app.Status.Namespace != "" {
+			t.Errorf("expected Status.Namespace to be empty for non-tenant kind, got %q", app.Status.Namespace)
+		}
+	})
+}
+
 func TestValidateNameLength(t *testing.T) {
 	tests := []struct {
 		name      string
@@ -102,3 +297,101 @@ func TestValidateNameLength(t *testing.T) {
 		})
 	}
 }
+
+// TestValidateTenantNamespaceLength covers the check that the computed
+// workload namespace for a Tenant application fits inside the 63-char
+// DNS-1123 label limit. The namespace is formed by dash-joining the
+// parent namespace with the tenant name, so deep nesting can exceed the
+// limit even when each individual name passes the per-name Helm release
+// length check.
+//
+// The "tenant-root" branches of computeTenantNamespace do not get a
+// dedicated overflow case: that branch produces "tenant-<name>" (7 +
+// len(name)), so for the computed result to exceed 63 chars the name
+// would need to exceed 56 chars, which is already blocked by
+// validateNameLength (max 46 for the "tenant-" prefix). The invariant
+// holds by arithmetic.
+func TestValidateTenantNamespaceLength(t *testing.T) {
+	tests := []struct {
+		name             string
+		currentNamespace string
+		tenantName       string
+		wantError        bool
+	}{
+		{
+			name:             "tenant-root with short name passes",
+			currentNamespace: "tenant-root",
+			tenantName:       "alpha",
+			wantError:        false,
+		},
+		{
+			name:             "root tenant inside root namespace passes",
+			currentNamespace: "tenant-root",
+			tenantName:       "root",
+			wantError:        false,
+		},
+		{
+			name:             "short parent and short name passes",
+			currentNamespace: "tenant-foo",
+			tenantName:       "bar",
+			wantError:        false,
+		},
+		{
+			name:             "exactly at 63-char limit passes",
+			currentNamespace: "tenant-" + strings.Repeat("a", 45), // 52 chars
+			tenantName:       strings.Repeat("b", 10),             // 10 chars -> 52+1+10 = 63
+			wantError:        false,
+		},
+		{
+			name:             "one char over the limit fails",
+			currentNamespace: "tenant-" + strings.Repeat("a", 45), // 52 chars
+			tenantName:       strings.Repeat("b", 11),             // 11 chars -> 52+1+11 = 64
+			wantError:        true,
+		},
+		{
+			name:             "deeply nested failure with issue-style names",
+			currentNamespace: "tenant-" + strings.Repeat("a", 35), // 42 chars
+			tenantName:       strings.Repeat("b", 25),             // 25 chars -> 42+1+25 = 68
+			wantError:        true,
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			r := &REST{
+				kindName: "Tenant",
+				releaseConfig: config.ReleaseConfig{
+					Prefix: "tenant-",
+				},
+			}
+
+			errs := r.validateTenantNamespaceLength(tt.currentNamespace, tt.tenantName)
+
+			if tt.wantError && len(errs) == 0 {
+				computed := r.computeTenantNamespace(tt.currentNamespace, tt.tenantName)
+				t.Errorf("expected error for parent=%q name=%q (computed=%q, len=%d), got none",
+					tt.currentNamespace, tt.tenantName, computed, len(computed))
+				return
+			}
+			if !tt.wantError && len(errs) > 0 {
+				t.Errorf("unexpected error for parent=%q name=%q: %v",
+					tt.currentNamespace, tt.tenantName, errs)
+				return
+			}
+
+			// For failing cases, verify the error message surfaces the
+			// computed namespace string and its actual length so a
+			// regression in the message format is caught.
+			if tt.wantError {
+				computed := r.computeTenantNamespace(tt.currentNamespace, tt.tenantName)
+				msg := errs.ToAggregate().Error()
+				if !strings.Contains(msg, computed) {
+					t.Errorf("error message must contain computed namespace %q, got: %s", computed, msg)
+				}
+				if !strings.Contains(msg, fmt.Sprintf("%d characters", len(computed))) {
+					t.Errorf("error message must contain the computed length %d, got: %s", len(computed), msg)
+				}
+			}
+		})
+	}
+}
diff --git a/pkg/registry/apps/application/rest_watch_test.go b/pkg/registry/apps/application/rest_watch_test.go
new file mode 100644
index 00000000..93a51f3c
--- /dev/null
+++ b/pkg/registry/apps/application/rest_watch_test.go
@@ -0,0 +1,223 @@
+package application
+
+import (
+	"context"
+	"testing"
+	"time"
+
+	cozyv1alpha1 "github.com/cozystack/cozystack/api/v1alpha1"
+	appsv1alpha1 "github.com/cozystack/cozystack/pkg/apis/apps/v1alpha1"
+	"github.com/cozystack/cozystack/pkg/config"
+	helmv2 "github.com/fluxcd/helm-controller/api/v2"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/apimachinery/pkg/runtime"
+	"k8s.io/apimachinery/pkg/types"
+	"k8s.io/apimachinery/pkg/watch"
+	"k8s.io/utils/ptr"
+	"sigs.k8s.io/controller-runtime/pkg/client"
+	"sigs.k8s.io/controller-runtime/pkg/client/fake"
+)
+
+func TestWmResultChan_NilWatcher(t *testing.T) {
+	ch := wmResultChan(nil)
+	if ch != nil {
+		t.Error("expected nil channel for nil watcher")
+	}
+}
+
+func TestWmResultChan_ValidWatcher(t *testing.T) {
+	fw := watch.NewFake()
+	ch := wmResultChan(fw)
+	if ch == nil {
+		t.Error("expected non-nil channel for valid watcher")
+	}
+	fw.Stop()
+}
+
+// TestWatchIntegration_WorkloadMonitorTriggersModifiedEvent verifies the
+// full path: WM event → label lookup → HelmRelease Get → Application conversion.
+func TestWatchIntegration_WorkloadMonitorTriggersModifiedEvent(t *testing.T) {
+	scheme := runtime.NewScheme()
+	_ = cozyv1alpha1.AddToScheme(scheme)
+	_ = helmv2.AddToScheme(scheme)
+
+	hr := &helmv2.HelmRelease{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "postgresql-mydb",
+			Namespace: "default",
+			Labels: map[string]string{
+				ApplicationKindLabel:  "PostgreSQL",
+				ApplicationGroupLabel: "apps.cozystack.io",
+				ApplicationNameLabel:  "mydb",
+			},
+		},
+	}
+	hr.Status.Conditions = []metav1.Condition{
+		{Type: "Ready", Status: metav1.ConditionTrue, Reason: "Succeeded", Message: "ok"},
+	}
+
+	wm := &cozyv1alpha1.WorkloadMonitor{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "mon-1",
+			Namespace: "default",
+			Labels: map[string]string{
+				appsv1alpha1.ApplicationKindLabel:  "PostgreSQL",
+				appsv1alpha1.ApplicationGroupLabel: "apps.cozystack.io",
+				appsv1alpha1.ApplicationNameLabel:  "mydb",
+			},
+		},
+		Status: cozyv1alpha1.WorkloadMonitorStatus{
+			Operational: ptr.To(false),
+		},
+	}
+
+	c := fake.NewClientBuilder().WithScheme(scheme).WithRuntimeObjects(hr, wm).Build()
+
+	r := newTestRESTWithSchemesFromClient(c)
+
+	// Simulate the Watch goroutine path: extract app name from WM labels,
+	// construct HelmRelease name, look up HR, convert to Application
+	wmAppName := wm.Labels[ApplicationNameLabel]
+	hrName := r.releaseConfig.Prefix + wmAppName
+
+	foundHR := &helmv2.HelmRelease{}
+	if err := c.Get(context.TODO(), types.NamespacedName{Namespace: "default", Name: hrName}, foundHR); err != nil {
+		t.Fatalf("failed to get HelmRelease: %v", err)
+	}
+	app, err := r.ConvertHelmReleaseToApplication(context.TODO(), foundHR)
+	if err != nil {
+		t.Fatalf("failed to convert: %v", err)
+	}
+	if app.Name != "mydb" {
+		t.Errorf("expected app name 'mydb', got %q", app.Name)
+	}
+
+	// Verify WorkloadsReady is False due to non-operational WM
+	wc := findCondition(app.GetConditions(), "WorkloadsReady")
+	if wc == nil {
+		t.Fatal("expected WorkloadsReady condition")
+	}
+	if wc.Status != metav1.ConditionFalse {
+		t.Errorf("expected WorkloadsReady=False, got %s", wc.Status)
+	}
+}
+
+// TestWatchIntegration_MonitorDeletionDropsWorkloadsReady verifies that when a
+// WorkloadMonitor is deleted, the Application's WorkloadsReady condition
+// disappears. Ready condition always reflects HelmRelease state regardless.
+func TestWatchIntegration_MonitorDeletionDropsWorkloadsReady(t *testing.T) {
+	scheme := runtime.NewScheme()
+	_ = cozyv1alpha1.AddToScheme(scheme)
+	_ = helmv2.AddToScheme(scheme)
+
+	hr := &helmv2.HelmRelease{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "postgresql-mydb",
+			Namespace: "default",
+			Labels: map[string]string{
+				ApplicationKindLabel:  "PostgreSQL",
+				ApplicationGroupLabel: "apps.cozystack.io",
+				ApplicationNameLabel:  "mydb",
+			},
+		},
+	}
+	hr.Status.Conditions = []metav1.Condition{
+		{Type: "Ready", Status: metav1.ConditionTrue, Reason: "Succeeded", Message: "ok"},
+	}
+
+	nonOpMonitor := &cozyv1alpha1.WorkloadMonitor{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "mon-1",
+			Namespace: "default",
+			Labels: map[string]string{
+				appsv1alpha1.ApplicationKindLabel:  "PostgreSQL",
+				appsv1alpha1.ApplicationGroupLabel: "apps.cozystack.io",
+				appsv1alpha1.ApplicationNameLabel:  "mydb",
+			},
+		},
+		Status: cozyv1alpha1.WorkloadMonitorStatus{
+			Operational: ptr.To(false),
+		},
+	}
+
+	c := fake.NewClientBuilder().WithScheme(scheme).WithRuntimeObjects(hr, nonOpMonitor).Build()
+	r := newTestRESTWithSchemesFromClient(c)
+
+	// Step 1: With non-operational monitor, WorkloadsReady=False, Ready=True
+	app1, err := r.convertHelmReleaseToApplication(context.TODO(), hr, nil)
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	wc1 := findCondition(app1.GetConditions(), "WorkloadsReady")
+	if wc1 == nil || wc1.Status != metav1.ConditionFalse {
+		t.Fatalf("expected WorkloadsReady=False with non-operational monitor, got %v", wc1)
+	}
+	rc1 := findCondition(app1.GetConditions(), "Ready")
+	if rc1 == nil || rc1.Status != metav1.ConditionTrue {
+		t.Fatalf("expected Ready=True (reflects HelmRelease), got %v", rc1)
+	}
+
+	// Step 2: Delete the monitor
+	if err := c.Delete(context.TODO(), nonOpMonitor); err != nil {
+		t.Fatalf("failed to delete monitor: %v", err)
+	}
+
+	// Step 3: WorkloadsReady should disappear, Ready stays True
+	app2, err := r.convertHelmReleaseToApplication(context.TODO(), hr, nil)
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	wc2 := findCondition(app2.GetConditions(), "WorkloadsReady")
+	if wc2 != nil {
+		t.Error("expected no WorkloadsReady condition after monitor deletion")
+	}
+	rc2 := findCondition(app2.GetConditions(), "Ready")
+	if rc2 == nil {
+		t.Fatal("expected Ready condition")
+	}
+	if rc2.Status != metav1.ConditionTrue {
+		t.Errorf("expected Ready=True after monitor deletion, got %s", rc2.Status)
+	}
+}
+
+// TestWatchIntegration_WMWatcherCloseProducesNilChannel verifies that
+// after wmWatcher is set to nil, wmResultChan returns nil channel.
+func TestWatchIntegration_WMWatcherCloseProducesNilChannel(t *testing.T) {
+	fw := watch.NewFake()
+	ch := wmResultChan(fw)
+	if ch == nil {
+		t.Fatal("expected non-nil channel before close")
+	}
+
+	fw.Stop()
+
+	timeout := time.After(time.Second)
+	for {
+		select {
+		case _, ok := <-ch:
+			if !ok {
+				var nilWatcher watch.Interface
+				nilCh := wmResultChan(nilWatcher)
+				if nilCh != nil {
+					t.Error("expected nil channel after watcher set to nil")
+				}
+				return
+			}
+		case <-timeout:
+			t.Fatal("timeout waiting for watcher channel to close")
+		}
+	}
+}
+
+func newTestRESTWithSchemesFromClient(c client.Client) *REST {
+	return NewREST(c, nil, &config.Resource{
+		Application: config.ApplicationConfig{
+			Kind:     "PostgreSQL",
+			Plural:   "postgresqls",
+			Singular: "postgresql",
+		},
+		Release: config.ReleaseConfig{
+			Prefix: "postgresql-",
+		},
+	})
+}
diff --git a/pkg/registry/apps/application/rest_workloads_test.go b/pkg/registry/apps/application/rest_workloads_test.go
new file mode 100644
index 00000000..b706c323
--- /dev/null
+++ b/pkg/registry/apps/application/rest_workloads_test.go
@@ -0,0 +1,342 @@
+package application
+
+import (
+	"context"
+	"testing"
+
+	cozyv1alpha1 "github.com/cozystack/cozystack/api/v1alpha1"
+	appsv1alpha1 "github.com/cozystack/cozystack/pkg/apis/apps/v1alpha1"
+	"github.com/cozystack/cozystack/pkg/config"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/apimachinery/pkg/runtime"
+	"k8s.io/utils/ptr"
+	"sigs.k8s.io/controller-runtime/pkg/client/fake"
+
+	helmv2 "github.com/fluxcd/helm-controller/api/v2"
+)
+
+func newTestRESTWithSchemes(objs ...runtime.Object) *REST {
+	scheme := runtime.NewScheme()
+	_ = cozyv1alpha1.AddToScheme(scheme)
+	_ = helmv2.AddToScheme(scheme)
+
+	builder := fake.NewClientBuilder().WithScheme(scheme)
+	for _, obj := range objs {
+		builder = builder.WithRuntimeObjects(obj)
+	}
+	c := builder.Build()
+
+	return NewREST(c, nil, &config.Resource{
+		Application: config.ApplicationConfig{
+			Kind:     "PostgreSQL",
+			Plural:   "postgresqls",
+			Singular: "postgresql",
+		},
+		Release: config.ReleaseConfig{
+			Prefix: "postgresql-",
+		},
+	})
+}
+
+func TestGetWorkloadsOperational_NoMonitors(t *testing.T) {
+	r := newTestRESTWithSchemes()
+	ws, err := r.getWorkloadsOperational(context.TODO(), "default", "mydb", nil)
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	if ws.found {
+		t.Error("expected found=false when no monitors exist")
+	}
+	if !ws.operational {
+		t.Error("expected operational=true when no monitors exist")
+	}
+}
+
+func TestGetWorkloadsOperational_AllOperational(t *testing.T) {
+	m1 := &cozyv1alpha1.WorkloadMonitor{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "mon-1",
+			Namespace: "default",
+			Labels: map[string]string{
+				appsv1alpha1.ApplicationKindLabel:  "PostgreSQL",
+				appsv1alpha1.ApplicationGroupLabel: "apps.cozystack.io",
+				appsv1alpha1.ApplicationNameLabel:  "mydb",
+			},
+		},
+		Status: cozyv1alpha1.WorkloadMonitorStatus{
+			Operational: ptr.To(true),
+		},
+	}
+	m2 := &cozyv1alpha1.WorkloadMonitor{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "mon-2",
+			Namespace: "default",
+			Labels: map[string]string{
+				appsv1alpha1.ApplicationKindLabel:  "PostgreSQL",
+				appsv1alpha1.ApplicationGroupLabel: "apps.cozystack.io",
+				appsv1alpha1.ApplicationNameLabel:  "mydb",
+			},
+		},
+		Status: cozyv1alpha1.WorkloadMonitorStatus{
+			Operational: ptr.To(true),
+		},
+	}
+
+	r := newTestRESTWithSchemes(m1, m2)
+	ws, err := r.getWorkloadsOperational(context.TODO(), "default", "mydb", nil)
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	if !ws.found {
+		t.Error("expected found=true")
+	}
+	if !ws.operational {
+		t.Error("expected operational=true when all monitors are operational")
+	}
+}
+
+func TestGetWorkloadsOperational_SomeNotOperational(t *testing.T) {
+	m1 := &cozyv1alpha1.WorkloadMonitor{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "mon-1",
+			Namespace: "default",
+			Labels: map[string]string{
+				appsv1alpha1.ApplicationKindLabel:  "PostgreSQL",
+				appsv1alpha1.ApplicationGroupLabel: "apps.cozystack.io",
+				appsv1alpha1.ApplicationNameLabel:  "mydb",
+			},
+		},
+		Status: cozyv1alpha1.WorkloadMonitorStatus{
+			Operational: ptr.To(true),
+		},
+	}
+	m2 := &cozyv1alpha1.WorkloadMonitor{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "mon-2",
+			Namespace: "default",
+			Labels: map[string]string{
+				appsv1alpha1.ApplicationKindLabel:  "PostgreSQL",
+				appsv1alpha1.ApplicationGroupLabel: "apps.cozystack.io",
+				appsv1alpha1.ApplicationNameLabel:  "mydb",
+			},
+		},
+		Status: cozyv1alpha1.WorkloadMonitorStatus{
+			Operational: ptr.To(false),
+		},
+	}
+
+	r := newTestRESTWithSchemes(m1, m2)
+	ws, err := r.getWorkloadsOperational(context.TODO(), "default", "mydb", nil)
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	if !ws.found {
+		t.Error("expected found=true")
+	}
+	if ws.operational {
+		t.Error("expected operational=false when at least one monitor is not operational")
+	}
+}
+
+func TestGetWorkloadsOperational_OperationalNil(t *testing.T) {
+	m := &cozyv1alpha1.WorkloadMonitor{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "mon-1",
+			Namespace: "default",
+			Labels: map[string]string{
+				appsv1alpha1.ApplicationKindLabel:  "PostgreSQL",
+				appsv1alpha1.ApplicationGroupLabel: "apps.cozystack.io",
+				appsv1alpha1.ApplicationNameLabel:  "mydb",
+			},
+		},
+		Status: cozyv1alpha1.WorkloadMonitorStatus{
+			Operational: nil, // Not yet reconciled
+		},
+	}
+
+	r := newTestRESTWithSchemes(m)
+	ws, err := r.getWorkloadsOperational(context.TODO(), "default", "mydb", nil)
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	if !ws.found {
+		t.Error("expected found=true")
+	}
+	if !ws.unknown {
+		t.Error("expected unknown=true when Operational is nil")
+	}
+}
+
+func TestGetWorkloadsOperational_MixedNilAndOperational(t *testing.T) {
+	m1 := &cozyv1alpha1.WorkloadMonitor{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "mon-1",
+			Namespace: "default",
+			Labels: map[string]string{
+				appsv1alpha1.ApplicationKindLabel:  "PostgreSQL",
+				appsv1alpha1.ApplicationGroupLabel: "apps.cozystack.io",
+				appsv1alpha1.ApplicationNameLabel:  "mydb",
+			},
+		},
+		Status: cozyv1alpha1.WorkloadMonitorStatus{
+			Operational: ptr.To(true),
+		},
+	}
+	m2 := &cozyv1alpha1.WorkloadMonitor{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "mon-2",
+			Namespace: "default",
+			Labels: map[string]string{
+				appsv1alpha1.ApplicationKindLabel:  "PostgreSQL",
+				appsv1alpha1.ApplicationGroupLabel: "apps.cozystack.io",
+				appsv1alpha1.ApplicationNameLabel:  "mydb",
+			},
+		},
+		Status: cozyv1alpha1.WorkloadMonitorStatus{
+			Operational: nil,
+		},
+	}
+
+	r := newTestRESTWithSchemes(m1, m2)
+	ws, err := r.getWorkloadsOperational(context.TODO(), "default", "mydb", nil)
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	if !ws.unknown {
+		t.Error("expected unknown=true when at least one monitor has nil Operational")
+	}
+}
+
+func TestGetWorkloadsOperational_MixedFailedAndPending(t *testing.T) {
+	m1 := &cozyv1alpha1.WorkloadMonitor{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "mon-1",
+			Namespace: "default",
+			Labels: map[string]string{
+				appsv1alpha1.ApplicationKindLabel:  "PostgreSQL",
+				appsv1alpha1.ApplicationGroupLabel: "apps.cozystack.io",
+				appsv1alpha1.ApplicationNameLabel:  "mydb",
+			},
+		},
+		Status: cozyv1alpha1.WorkloadMonitorStatus{
+			Operational: ptr.To(false), // Confirmed failure
+		},
+	}
+	m2 := &cozyv1alpha1.WorkloadMonitor{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "mon-2",
+			Namespace: "default",
+			Labels: map[string]string{
+				appsv1alpha1.ApplicationKindLabel:  "PostgreSQL",
+				appsv1alpha1.ApplicationGroupLabel: "apps.cozystack.io",
+				appsv1alpha1.ApplicationNameLabel:  "mydb",
+			},
+		},
+		Status: cozyv1alpha1.WorkloadMonitorStatus{
+			Operational: nil, // Not yet reconciled
+		},
+	}
+
+	r := newTestRESTWithSchemes(m1, m2)
+	ws, err := r.getWorkloadsOperational(context.TODO(), "default", "mydb", nil)
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	if ws.operational {
+		t.Error("expected operational=false when at least one monitor is explicitly failed")
+	}
+	if !ws.unknown {
+		t.Error("expected unknown=true when at least one monitor has nil Operational")
+	}
+}
+
+func TestConvertConditions_MixedFailedAndPendingShowsFalse(t *testing.T) {
+	mFailed := &cozyv1alpha1.WorkloadMonitor{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "mon-failed",
+			Namespace: "default",
+			Labels: map[string]string{
+				appsv1alpha1.ApplicationKindLabel:  "PostgreSQL",
+				appsv1alpha1.ApplicationGroupLabel: "apps.cozystack.io",
+				appsv1alpha1.ApplicationNameLabel:  "mydb",
+			},
+		},
+		Status: cozyv1alpha1.WorkloadMonitorStatus{
+			Operational: ptr.To(false),
+		},
+	}
+	mPending := &cozyv1alpha1.WorkloadMonitor{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "mon-pending",
+			Namespace: "default",
+			Labels: map[string]string{
+				appsv1alpha1.ApplicationKindLabel:  "PostgreSQL",
+				appsv1alpha1.ApplicationGroupLabel: "apps.cozystack.io",
+				appsv1alpha1.ApplicationNameLabel:  "mydb",
+			},
+		},
+		Status: cozyv1alpha1.WorkloadMonitorStatus{
+			Operational: nil,
+		},
+	}
+
+	r := newTestRESTWithSchemes(mFailed, mPending)
+	hr := &helmv2.HelmRelease{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "postgresql-mydb",
+			Namespace: "default",
+			Labels: map[string]string{
+				ApplicationKindLabel:  "PostgreSQL",
+				ApplicationGroupLabel: "apps.cozystack.io",
+				ApplicationNameLabel:  "mydb",
+			},
+		},
+	}
+	hr.Status.Conditions = []metav1.Condition{
+		{Type: "Ready", Status: metav1.ConditionTrue, Reason: "Succeeded", Message: "ok"},
+	}
+
+	app, err := r.convertHelmReleaseToApplication(context.TODO(), hr, nil)
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+
+	// Concrete failure should take priority over unknown
+	wc := findCondition(app.GetConditions(), "WorkloadsReady")
+	if wc == nil {
+		t.Fatal("expected WorkloadsReady condition")
+	}
+	if wc.Status != metav1.ConditionFalse {
+		t.Errorf("expected WorkloadsReady=False (failure takes priority over unknown), got %s", wc.Status)
+	}
+}
+
+func TestGetWorkloadsOperational_DifferentApp_NotFound(t *testing.T) {
+	m := &cozyv1alpha1.WorkloadMonitor{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "mon-1",
+			Namespace: "default",
+			Labels: map[string]string{
+				appsv1alpha1.ApplicationKindLabel:  "PostgreSQL",
+				appsv1alpha1.ApplicationGroupLabel: "apps.cozystack.io",
+				appsv1alpha1.ApplicationNameLabel:  "other-db",
+			},
+		},
+		Status: cozyv1alpha1.WorkloadMonitorStatus{
+			Operational: ptr.To(false),
+		},
+	}
+
+	r := newTestRESTWithSchemes(m)
+	ws, err := r.getWorkloadsOperational(context.TODO(), "default", "mydb", nil)
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	if ws.found {
+		t.Error("expected found=false for different app name")
+	}
+	if !ws.operational {
+		t.Error("expected operational=true when no matching monitors found")
+	}
+}
+
diff --git a/pkg/registry/core/tenantnamespace/rest.go b/pkg/registry/core/tenantnamespace/rest.go
index f1ed3fab..4ee97eb0 100644
--- a/pkg/registry/core/tenantnamespace/rest.go
+++ b/pkg/registry/core/tenantnamespace/rest.go
@@ -17,6 +17,8 @@ import (
 	apierrors "k8s.io/apimachinery/pkg/api/errors"
 	metainternal "k8s.io/apimachinery/pkg/apis/meta/internalversion"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/apimachinery/pkg/fields"
+	"k8s.io/apimachinery/pkg/labels"
 	"k8s.io/apimachinery/pkg/runtime"
 	"k8s.io/apimachinery/pkg/runtime/schema"
 	"k8s.io/apimachinery/pkg/types"
@@ -24,6 +26,7 @@ import (
 	"k8s.io/apimachinery/pkg/watch"
 	"k8s.io/apiserver/pkg/endpoints/request"
 	"k8s.io/apiserver/pkg/registry/rest"
+	"k8s.io/klog/v2"
 	"sigs.k8s.io/controller-runtime/pkg/client"
 
 	corev1alpha1 "github.com/cozystack/cozystack/pkg/apis/core/v1alpha1"
@@ -123,8 +126,18 @@ func (r *REST) Get(
 		return nil, apierrors.NewNotFound(r.gvr.GroupResource(), name)
 	}
 
+	// Check if user has access to this namespace
+	hasAccess, err := r.hasAccessToNamespace(ctx, name)
+	if err != nil {
+		return nil, err
+	}
+	if !hasAccess {
+		// Return Forbidden to follow standard K8s RBAC behavior
+		return nil, apierrors.NewForbidden(r.gvr.GroupResource(), name, fmt.Errorf("access denied"))
+	}
+
 	ns := &corev1.Namespace{}
-	err := r.c.Get(ctx, types.NamespacedName{Namespace: "", Name: name}, ns, &client.GetOptions{Raw: opts})
+	err = r.c.Get(ctx, types.NamespacedName{Namespace: "", Name: name}, ns, &client.GetOptions{Raw: opts})
 	if err != nil {
 		return nil, err
 	}
@@ -143,11 +156,33 @@ func (r *REST) Get(
 // -----------------------------------------------------------------------------
 
 func (r *REST) Watch(ctx context.Context, opts *metainternal.ListOptions) (watch.Interface, error) {
+	// Extract user identity once for the lifetime of the watch — it does not
+	// change between events and rebuilding it per event is wasteful.
+	u, ok := request.UserFrom(ctx)
+	if !ok {
+		return nil, apierrors.NewUnauthorized("user missing in context")
+	}
+	username := u.GetName()
+	groups := make(map[string]struct{})
+	for _, group := range u.GetGroups() {
+		groups[group] = struct{}{}
+	}
+
 	nsList := &corev1.NamespaceList{}
-	nsWatch, err := r.w.Watch(ctx, nsList, &client.ListOptions{Raw: &metav1.ListOptions{
+
+	// Build upstream watch options with field and label selectors
+	rawOpts := &metav1.ListOptions{
 		Watch:           true,
 		ResourceVersion: opts.ResourceVersion,
-	}})
+	}
+	if opts.FieldSelector != nil {
+		rawOpts.FieldSelector = opts.FieldSelector.String()
+	}
+	if opts.LabelSelector != nil {
+		rawOpts.LabelSelector = opts.LabelSelector.String()
+	}
+
+	nsWatch, err := r.w.Watch(ctx, nsList, &client.ListOptions{Raw: rawOpts})
 	if err != nil {
 		return nil, err
 	}
@@ -189,6 +224,30 @@ func (r *REST) Watch(ctx context.Context, opts *metainternal.ListOptions) (watch
 				continue
 			}
 
+			// Apply defensive filtering for field and label selectors
+			if opts.FieldSelector != nil {
+				if !opts.FieldSelector.Matches(fields.Set{"metadata.name": ns.Name}) {
+					continue
+				}
+			}
+			if opts.LabelSelector != nil {
+				if !opts.LabelSelector.Matches(labels.Set(ns.Labels)) {
+					continue
+				}
+			}
+
+			// Check if user has access to this namespace using the cached
+			// identity — avoids re-extracting user/groups on every event.
+			hasAccess, err := r.hasAccessToNamespaceForUser(ctx, ns.Name, username, groups)
+			if err != nil {
+				klog.ErrorS(err, "Failed to check access for namespace in watch", "namespace", ns.Name)
+				continue
+			}
+			if !hasAccess {
+				// User doesn't have access, skip this event
+				continue
+			}
+
 			out := &corev1alpha1.TenantNamespace{
 				TypeMeta: metav1.TypeMeta{
 					APIVersion: corev1alpha1.SchemeGroupVersion.String(),
@@ -309,6 +368,26 @@ func (r *REST) makeList(src *corev1.NamespaceList, allowed []string) *corev1alph
 	return out
 }
 
+// matchesSubject checks if a RoleBinding subject matches the user's identity.
+// It handles Group, User, and ServiceAccount subjects with proper namespace fallback.
+func matchesSubject(subj rbacv1.Subject, bindingNamespace, username string, groups map[string]struct{}) bool {
+	switch subj.Kind {
+	case "Group":
+		_, ok := groups[subj.Name]
+		return ok
+	case "User":
+		return subj.Name == username
+	case "ServiceAccount":
+		saNamespace := subj.Namespace
+		if saNamespace == "" {
+			saNamespace = bindingNamespace
+		}
+		return username == fmt.Sprintf("system:serviceaccount:%s:%s", saNamespace, subj.Name)
+	default:
+		return false
+	}
+}
+
 func (r *REST) filterAccessible(
 	ctx context.Context,
 	names []string,
@@ -347,22 +426,9 @@ func (r *REST) filterAccessible(
 	subjectLoop:
 		for j := range rbs.Items[i].Subjects {
 			subj := rbs.Items[i].Subjects[j]
-			switch subj.Kind {
-			case "Group":
-				if _, ok = groups[subj.Name]; ok {
-					allowedNameSet[rbs.Items[i].Namespace] = struct{}{}
-					break subjectLoop
-				}
-			case "User":
-				if subj.Name == u.GetName() {
-					allowedNameSet[rbs.Items[i].Namespace] = struct{}{}
-					break subjectLoop
-				}
-			case "ServiceAccount":
-				if u.GetName() == fmt.Sprintf("system:serviceaccount:%s:%s", subj.Namespace, subj.Name) {
-					allowedNameSet[rbs.Items[i].Namespace] = struct{}{}
-					break subjectLoop
-				}
+			if matchesSubject(subj, rbs.Items[i].Namespace, u.GetName(), groups) {
+				allowedNameSet[rbs.Items[i].Namespace] = struct{}{}
+				break subjectLoop
 			}
 		}
 	}
@@ -373,6 +439,60 @@ func (r *REST) filterAccessible(
 	return allowed, nil
 }
 
+// hasAccessToNamespace checks if the user has access to a single namespace.
+// This is optimized for Get/Watch operations where we check one namespace at a time.
+// It lists RoleBindings only in the target namespace instead of all cluster RoleBindings.
+func (r *REST) hasAccessToNamespace(
+	ctx context.Context,
+	namespace string,
+) (bool, error) {
+	u, ok := request.UserFrom(ctx)
+	if !ok {
+		return false, fmt.Errorf("user missing in context")
+	}
+	groups := make(map[string]struct{})
+	for _, group := range u.GetGroups() {
+		groups[group] = struct{}{}
+	}
+	return r.hasAccessToNamespaceForUser(ctx, namespace, u.GetName(), groups)
+}
+
+// hasAccessToNamespaceForUser is the inner check that does not re-extract user
+// identity from context. Use this in hot paths (e.g. the Watch loop) where the
+// caller has already cached the user name and groups.
+func (r *REST) hasAccessToNamespaceForUser(
+	ctx context.Context,
+	namespace, username string,
+	groups map[string]struct{},
+) (bool, error) {
+	// Check privileged groups
+	if _, ok := groups["system:masters"]; ok {
+		return true, nil
+	}
+	if _, ok := groups["cozystack-cluster-admin"]; ok {
+		return true, nil
+	}
+
+	// List RoleBindings only in the target namespace
+	rbs := &rbacv1.RoleBindingList{}
+	err := r.c.List(ctx, rbs, client.InNamespace(namespace))
+	if err != nil {
+		return false, fmt.Errorf("failed to list rolebindings in %s: %w", namespace, err)
+	}
+
+	// Check if user is in any RoleBinding subjects
+	for i := range rbs.Items {
+		for j := range rbs.Items[i].Subjects {
+			subj := rbs.Items[i].Subjects[j]
+			if matchesSubject(subj, rbs.Items[i].Namespace, username, groups) {
+				return true, nil
+			}
+		}
+	}
+
+	return false, nil
+}
+
 // -----------------------------------------------------------------------------
 // Boiler-plate
 // -----------------------------------------------------------------------------
diff --git a/pkg/registry/core/tenantnamespace/rest_test.go b/pkg/registry/core/tenantnamespace/rest_test.go
index 7f2979bc..52d2a075 100644
--- a/pkg/registry/core/tenantnamespace/rest_test.go
+++ b/pkg/registry/core/tenantnamespace/rest_test.go
@@ -3,10 +3,20 @@
 package tenantnamespace
 
 import (
+	"context"
 	"testing"
 
 	corev1 "k8s.io/api/core/v1"
+	rbacv1 "k8s.io/api/rbac/v1"
+	apierrors "k8s.io/apimachinery/pkg/api/errors"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/apimachinery/pkg/runtime"
+	"k8s.io/apimachinery/pkg/runtime/schema"
+	"k8s.io/apiserver/pkg/authentication/user"
+	"k8s.io/apiserver/pkg/endpoints/request"
+	"sigs.k8s.io/controller-runtime/pkg/client/fake"
+
+	corev1alpha1 "github.com/cozystack/cozystack/pkg/apis/core/v1alpha1"
 )
 
 func TestMakeListSortsAlphabetically(t *testing.T) {
@@ -38,3 +48,466 @@ func TestMakeListSortsAlphabetically(t *testing.T) {
 		}
 	}
 }
+
+// Security tests for IDOR fix
+
+func TestHasAccessToNamespace_WithUserAccess(t *testing.T) {
+	scheme := runtime.NewScheme()
+	_ = corev1.AddToScheme(scheme)
+	_ = rbacv1.AddToScheme(scheme)
+
+	ns := &corev1.Namespace{
+		ObjectMeta: metav1.ObjectMeta{Name: "tenant-test"},
+	}
+
+	rb := &rbacv1.RoleBinding{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "test-binding",
+			Namespace: "tenant-test",
+		},
+		Subjects: []rbacv1.Subject{
+			{Kind: "User", Name: "test-user", APIGroup: "rbac.authorization.k8s.io"},
+		},
+		RoleRef: rbacv1.RoleRef{
+			APIGroup: "rbac.authorization.k8s.io",
+			Kind:     "Role",
+			Name:     "test-role",
+		},
+	}
+
+	client := fake.NewClientBuilder().
+		WithScheme(scheme).
+		WithObjects(ns, rb).
+		Build()
+
+	r := &REST{
+		c:   client,
+		gvr: schema.GroupVersionResource{Group: "core.cozystack.io", Version: "v1alpha1", Resource: "tenantnamespaces"},
+	}
+
+	u := &user.DefaultInfo{
+		Name:   "test-user",
+		Groups: []string{"system:authenticated"},
+	}
+	ctx := request.WithUser(context.Background(), u)
+
+	hasAccess, err := r.hasAccessToNamespace(ctx, "tenant-test")
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	if !hasAccess {
+		t.Error("expected user to have access, but got false")
+	}
+}
+
+func TestHasAccessToNamespace_WithoutAccess(t *testing.T) {
+	scheme := runtime.NewScheme()
+	_ = corev1.AddToScheme(scheme)
+	_ = rbacv1.AddToScheme(scheme)
+
+	ns := &corev1.Namespace{
+		ObjectMeta: metav1.ObjectMeta{Name: "tenant-test"},
+	}
+
+	// RoleBinding for different user
+	rb := &rbacv1.RoleBinding{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "test-binding",
+			Namespace: "tenant-test",
+		},
+		Subjects: []rbacv1.Subject{
+			{Kind: "User", Name: "other-user", APIGroup: "rbac.authorization.k8s.io"},
+		},
+		RoleRef: rbacv1.RoleRef{
+			APIGroup: "rbac.authorization.k8s.io",
+			Kind:     "Role",
+			Name:     "test-role",
+		},
+	}
+
+	client := fake.NewClientBuilder().
+		WithScheme(scheme).
+		WithObjects(ns, rb).
+		Build()
+
+	r := &REST{
+		c:   client,
+		gvr: schema.GroupVersionResource{Group: "core.cozystack.io", Version: "v1alpha1", Resource: "tenantnamespaces"},
+	}
+
+	u := &user.DefaultInfo{
+		Name:   "test-user",
+		Groups: []string{"system:authenticated"},
+	}
+	ctx := request.WithUser(context.Background(), u)
+
+	hasAccess, err := r.hasAccessToNamespace(ctx, "tenant-test")
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	if hasAccess {
+		t.Error("expected user to NOT have access, but got true")
+	}
+}
+
+func TestHasAccessToNamespace_WithGroupAccess(t *testing.T) {
+	scheme := runtime.NewScheme()
+	_ = corev1.AddToScheme(scheme)
+	_ = rbacv1.AddToScheme(scheme)
+
+	ns := &corev1.Namespace{
+		ObjectMeta: metav1.ObjectMeta{Name: "tenant-test"},
+	}
+
+	rb := &rbacv1.RoleBinding{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "test-binding",
+			Namespace: "tenant-test",
+		},
+		Subjects: []rbacv1.Subject{
+			{Kind: "Group", Name: "test-group", APIGroup: "rbac.authorization.k8s.io"},
+		},
+		RoleRef: rbacv1.RoleRef{
+			APIGroup: "rbac.authorization.k8s.io",
+			Kind:     "Role",
+			Name:     "test-role",
+		},
+	}
+
+	client := fake.NewClientBuilder().
+		WithScheme(scheme).
+		WithObjects(ns, rb).
+		Build()
+
+	r := &REST{
+		c:   client,
+		gvr: schema.GroupVersionResource{Group: "core.cozystack.io", Version: "v1alpha1", Resource: "tenantnamespaces"},
+	}
+
+	u := &user.DefaultInfo{
+		Name:   "test-user",
+		Groups: []string{"system:authenticated", "test-group"},
+	}
+	ctx := request.WithUser(context.Background(), u)
+
+	hasAccess, err := r.hasAccessToNamespace(ctx, "tenant-test")
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	if !hasAccess {
+		t.Error("expected user to have access via group, but got false")
+	}
+}
+
+func TestHasAccessToNamespace_SystemMasters(t *testing.T) {
+	scheme := runtime.NewScheme()
+	_ = corev1.AddToScheme(scheme)
+	_ = rbacv1.AddToScheme(scheme)
+
+	ns := &corev1.Namespace{
+		ObjectMeta: metav1.ObjectMeta{Name: "tenant-test"},
+	}
+
+	client := fake.NewClientBuilder().
+		WithScheme(scheme).
+		WithObjects(ns).
+		Build()
+
+	r := &REST{
+		c:   client,
+		gvr: schema.GroupVersionResource{Group: "core.cozystack.io", Version: "v1alpha1", Resource: "tenantnamespaces"},
+	}
+
+	u := &user.DefaultInfo{
+		Name:   "admin",
+		Groups: []string{"system:masters"},
+	}
+	ctx := request.WithUser(context.Background(), u)
+
+	hasAccess, err := r.hasAccessToNamespace(ctx, "tenant-test")
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	if !hasAccess {
+		t.Error("expected system:masters to have access, but got false")
+	}
+}
+
+func TestHasAccessToNamespace_CozyAdminGroup(t *testing.T) {
+	scheme := runtime.NewScheme()
+	_ = corev1.AddToScheme(scheme)
+	_ = rbacv1.AddToScheme(scheme)
+
+	ns := &corev1.Namespace{
+		ObjectMeta: metav1.ObjectMeta{Name: "tenant-test"},
+	}
+
+	client := fake.NewClientBuilder().
+		WithScheme(scheme).
+		WithObjects(ns).
+		Build()
+
+	r := &REST{
+		c:   client,
+		gvr: schema.GroupVersionResource{Group: "core.cozystack.io", Version: "v1alpha1", Resource: "tenantnamespaces"},
+	}
+
+	u := &user.DefaultInfo{
+		Name:   "cozy-admin",
+		Groups: []string{"cozystack-cluster-admin"},
+	}
+	ctx := request.WithUser(context.Background(), u)
+
+	hasAccess, err := r.hasAccessToNamespace(ctx, "tenant-test")
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	if !hasAccess {
+		t.Error("expected cozystack-cluster-admin to have access, but got false")
+	}
+}
+
+func TestHasAccessToNamespace_ServiceAccount(t *testing.T) {
+	scheme := runtime.NewScheme()
+	_ = corev1.AddToScheme(scheme)
+	_ = rbacv1.AddToScheme(scheme)
+
+	ns := &corev1.Namespace{
+		ObjectMeta: metav1.ObjectMeta{Name: "tenant-test"},
+	}
+
+	rb := &rbacv1.RoleBinding{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "test-binding",
+			Namespace: "tenant-test",
+		},
+		Subjects: []rbacv1.Subject{
+			{Kind: "ServiceAccount", Name: "test-sa", Namespace: "tenant-test"},
+		},
+		RoleRef: rbacv1.RoleRef{
+			APIGroup: "rbac.authorization.k8s.io",
+			Kind:     "Role",
+			Name:     "test-role",
+		},
+	}
+
+	client := fake.NewClientBuilder().
+		WithScheme(scheme).
+		WithObjects(ns, rb).
+		Build()
+
+	r := &REST{
+		c:   client,
+		gvr: schema.GroupVersionResource{Group: "core.cozystack.io", Version: "v1alpha1", Resource: "tenantnamespaces"},
+	}
+
+	u := &user.DefaultInfo{
+		Name:   "system:serviceaccount:tenant-test:test-sa",
+		Groups: []string{"system:authenticated"},
+	}
+	ctx := request.WithUser(context.Background(), u)
+
+	hasAccess, err := r.hasAccessToNamespace(ctx, "tenant-test")
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	if !hasAccess {
+		t.Error("expected service account to have access, but got false")
+	}
+}
+
+func TestHasAccessToNamespace_ServiceAccountEmptyNamespace(t *testing.T) {
+	scheme := runtime.NewScheme()
+	_ = corev1.AddToScheme(scheme)
+	_ = rbacv1.AddToScheme(scheme)
+
+	ns := &corev1.Namespace{
+		ObjectMeta: metav1.ObjectMeta{Name: "tenant-test"},
+	}
+
+	// ServiceAccount subject with empty namespace should default to RoleBinding namespace
+	rb := &rbacv1.RoleBinding{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "test-binding",
+			Namespace: "tenant-test",
+		},
+		Subjects: []rbacv1.Subject{
+			{Kind: "ServiceAccount", Name: "test-sa", Namespace: ""}, // Empty namespace
+		},
+		RoleRef: rbacv1.RoleRef{
+			APIGroup: "rbac.authorization.k8s.io",
+			Kind:     "Role",
+			Name:     "test-role",
+		},
+	}
+
+	client := fake.NewClientBuilder().
+		WithScheme(scheme).
+		WithObjects(ns, rb).
+		Build()
+
+	r := &REST{
+		c:   client,
+		gvr: schema.GroupVersionResource{Group: "core.cozystack.io", Version: "v1alpha1", Resource: "tenantnamespaces"},
+	}
+
+	u := &user.DefaultInfo{
+		Name:   "system:serviceaccount:tenant-test:test-sa",
+		Groups: []string{"system:authenticated"},
+	}
+	ctx := request.WithUser(context.Background(), u)
+
+	hasAccess, err := r.hasAccessToNamespace(ctx, "tenant-test")
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	if !hasAccess {
+		t.Error("expected service account with empty namespace to have access, but got false")
+	}
+}
+
+func TestGet_WithAccess(t *testing.T) {
+	scheme := runtime.NewScheme()
+	_ = corev1.AddToScheme(scheme)
+	_ = rbacv1.AddToScheme(scheme)
+
+	ns := &corev1.Namespace{
+		ObjectMeta: metav1.ObjectMeta{Name: "tenant-test"},
+	}
+
+	rb := &rbacv1.RoleBinding{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "test-binding",
+			Namespace: "tenant-test",
+		},
+		Subjects: []rbacv1.Subject{
+			{Kind: "User", Name: "test-user", APIGroup: "rbac.authorization.k8s.io"},
+		},
+		RoleRef: rbacv1.RoleRef{
+			APIGroup: "rbac.authorization.k8s.io",
+			Kind:     "Role",
+			Name:     "test-role",
+		},
+	}
+
+	client := fake.NewClientBuilder().
+		WithScheme(scheme).
+		WithObjects(ns, rb).
+		Build()
+
+	r := &REST{
+		c:   client,
+		gvr: schema.GroupVersionResource{Group: "core.cozystack.io", Version: "v1alpha1", Resource: "tenantnamespaces"},
+	}
+
+	u := &user.DefaultInfo{
+		Name:   "test-user",
+		Groups: []string{"system:authenticated"},
+	}
+	ctx := request.WithUser(context.Background(), u)
+
+	obj, err := r.Get(ctx, "tenant-test", &metav1.GetOptions{})
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	tn, ok := obj.(*corev1alpha1.TenantNamespace)
+	if !ok {
+		t.Fatalf("expected *TenantNamespace, got %T", obj)
+	}
+	if tn.Name != "tenant-test" {
+		t.Errorf("expected name %q, got %q", "tenant-test", tn.Name)
+	}
+	if tn.Kind != "TenantNamespace" {
+		t.Errorf("expected Kind=TenantNamespace, got %q", tn.Kind)
+	}
+	if tn.APIVersion != corev1alpha1.SchemeGroupVersion.String() {
+		t.Errorf("expected APIVersion=%q, got %q", corev1alpha1.SchemeGroupVersion.String(), tn.APIVersion)
+	}
+}
+
+func TestGet_WithoutAccess(t *testing.T) {
+	scheme := runtime.NewScheme()
+	_ = corev1.AddToScheme(scheme)
+	_ = rbacv1.AddToScheme(scheme)
+
+	ns := &corev1.Namespace{
+		ObjectMeta: metav1.ObjectMeta{Name: "tenant-test"},
+	}
+
+	// RoleBinding for different user
+	rb := &rbacv1.RoleBinding{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "test-binding",
+			Namespace: "tenant-test",
+		},
+		Subjects: []rbacv1.Subject{
+			{Kind: "User", Name: "other-user", APIGroup: "rbac.authorization.k8s.io"},
+		},
+		RoleRef: rbacv1.RoleRef{
+			APIGroup: "rbac.authorization.k8s.io",
+			Kind:     "Role",
+			Name:     "test-role",
+		},
+	}
+
+	client := fake.NewClientBuilder().
+		WithScheme(scheme).
+		WithObjects(ns, rb).
+		Build()
+
+	r := &REST{
+		c:   client,
+		gvr: schema.GroupVersionResource{Group: "core.cozystack.io", Version: "v1alpha1", Resource: "tenantnamespaces"},
+	}
+
+	u := &user.DefaultInfo{
+		Name:   "test-user",
+		Groups: []string{"system:authenticated"},
+	}
+	ctx := request.WithUser(context.Background(), u)
+
+	obj, err := r.Get(ctx, "tenant-test", &metav1.GetOptions{})
+	if err == nil {
+		t.Fatal("expected error, got nil")
+	}
+	if obj != nil {
+		t.Errorf("expected nil object, got %v", obj)
+	}
+
+	// Verify it returns Forbidden to follow standard K8s RBAC behavior
+	if !apierrors.IsForbidden(err) {
+		t.Errorf("expected Forbidden error, got %v", err)
+	}
+}
+
+func TestGet_NonTenantNamespace(t *testing.T) {
+	scheme := runtime.NewScheme()
+	_ = corev1.AddToScheme(scheme)
+	_ = rbacv1.AddToScheme(scheme)
+
+	client := fake.NewClientBuilder().
+		WithScheme(scheme).
+		Build()
+
+	r := &REST{
+		c:   client,
+		gvr: schema.GroupVersionResource{Group: "core.cozystack.io", Version: "v1alpha1", Resource: "tenantnamespaces"},
+	}
+
+	u := &user.DefaultInfo{
+		Name:   "test-user",
+		Groups: []string{"system:masters"},
+	}
+	ctx := request.WithUser(context.Background(), u)
+
+	obj, err := r.Get(ctx, "default", &metav1.GetOptions{})
+	if err == nil {
+		t.Fatal("expected error for non-tenant namespace, got nil")
+	}
+	if obj != nil {
+		t.Errorf("expected nil object, got %v", obj)
+	}
+	if !apierrors.IsNotFound(err) {
+		t.Errorf("expected NotFound error, got %v", err)
+	}
+}