- Replace hardcoded tenant-root with {{ .Release.Namespace }} in vmagent
and fluent-bit configs
- Add ExternalName services in cozystack-basics to redirect monitoring
traffic from cozy-monitoring to tenant-root when engine is deployed
- Add missing components to monitoring PackageSource
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: IvanHunters <xorokhotnikov@gmail.com>
Remove components that are already defined in separate PackageSources
to avoid race conditions and HelmRelease conflicts:
- vertical-pod-autoscaler
- postgres-operator
- grafana-operator
- victoria-metrics-operator
- prometheus-operator-crds
These components should be installed via their dedicated PackageSources.
Signed-off-by: IvanHunters <xorokhotnikov@gmail.com>
The monitoring package depends on cozystack.cozystack-controller which
is only installed as part of cozystack-engine. This prevents monitoring
from becoming ready when cozystack-engine is not enabled.
Remove this dependency to allow monitoring to work independently.
Signed-off-by: IvanHunters <xorokhotnikov@gmail.com>
Add migration 26 that re-labels monitoring resources to be owned by
monitoring-system HelmRelease instead of monitoring. This allows
seamless transition from direct resource management in extra/monitoring
to delegated management via system/monitoring chart.
Migration steps:
- Find all monitoring HelmReleases in tenant namespaces
- Suspend HelmRelease to prevent reconciliation
- Delete helm secrets to orphan resources
- Relabel all resources to monitoring-system
- Delete suspended HelmRelease
Also updates:
- monitoring-application.yaml: add monitoring-system component
- helmrelease.yaml: reference monitoring-system ExternalArtifact
- values.yaml: bump targetVersion to 25
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: Andrei Kvapil <kvapss@gmail.com>
## Summary
- Add migration 25 for v1.0 upgrade cleanup:
- Remove legacy `cozystack` installer deployment
- Remove legacy `cozystack-assets` statefulset
- Remove `bootbox` and `bootbox-rd` HelmReleases (will be recreated by
new platform chart if enabled)
- Revert migration 23 to original state (only removes
cozystack-resource-definition-crd)
- Update targetVersion from 24 to 26
- Fix jq syntax error in `hack/migrate-to-version-1.0.sh`
Migration 23 was already executed on existing clusters at version 23+,
so adding cleanup there would never run. The new migration 25 ensures
proper cleanup during v1.0 upgrade.
## Test plan
- [ ] Upgrade from v0.4x to v1.0
- [ ] Verify old `cozystack` deployment is deleted
- [ ] Verify old `cozystack-assets` statefulset is deleted
- [ ] Verify old `bootbox` and `bootbox-rd` HelmReleases are deleted
- [ ] Verify platform installs successfully after migration
- [ ] Run `./hack/migrate-to-version-1.0.sh` without jq errors
Migration 23 was already executed on existing clusters, so the legacy
installer removal never ran. This moves the cleanup to a new migration 25
which will execute during upgrade.
Changes:
- Revert migration 23 to original state (only removes cozystack-resource-definition-crd)
- Add migration 25 for v1.0 upgrade cleanup:
- Remove legacy cozystack installer deployment
- Remove legacy cozystack-assets statefulset
- Remove bootbox and bootbox-rd HelmReleases
- Update targetVersion from 24 to 26
- Fix jq syntax error in migrate-to-version-1.0.sh
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: Andrei Kvapil <kvapss@gmail.com>
## Summary
- Remove duplicate `bootbox-system` component from `bootbox-application`
PackageSource (was duplicating system/bootbox installation)
- Auto-create `cozystack.bootbox-application` Package when bootbox is
enabled in `bundles.enabledPackages`
- This ensures `bootbox-rd` (ApplicationDefinition) is properly
installed as a dependency before `cozystack.bootbox`
## Test plan
- [ ] Enable bootbox in `bundles.enabledPackages`
- [ ] Verify both `cozystack.bootbox-application` and
`cozystack.bootbox` Packages are created
- [ ] Verify `bootbox-rd` (ApplicationDefinition) is installed
- Fix docker build context for migrations image
- Fix yq flags for compatibility with Mike Farah's yq v4
- Add platform image to main build target
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: Andrei Kvapil <kvapss@gmail.com>
Add cleanup of old installer components during migration:
- Delete cozystack deployment
- Delete cozystack-assets statefulset
This ensures no conflict between old installer and new operator
when upgrading to v1.0 architecture.
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: Andrei Kvapil <kvapss@gmail.com>
- Remove duplicate bootbox-system component from bootbox-application
PackageSource (was duplicating system/bootbox installation)
- Auto-create cozystack.bootbox-application Package when bootbox is
enabled in bundles.enabledPackages
- This ensures bootbox-rd (ApplicationDefinition) is properly installed
as a dependency before cozystack.bootbox
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: Andrei Kvapil <kvapss@gmail.com>
## What this PR does
Aligns MongoDB configuration with postgres and mysql patterns by
introducing a separate `databases` section.
**Before:**
```yaml
users:
myuser:
db: mydb
roles:
- name: readWrite
db: mydb
- name: dbAdmin
db: mydb
```
**After:**
```yaml
users:
myuser: {}
databases:
mydb:
roles:
admin:
- myuser
readonly:
- reader
```
Changes:
- Add separate `databases` section with `roles.admin` and
`roles.readonly`
- Simplify `users` to only contain optional password field
- All users authenticate via `admin` database (MongoDB production best
practice)
- `admin` role maps to `readWrite` + `dbAdmin` permissions
- `readonly` role maps to `read` permission
## Migration
Includes migration script (`migrations/24`) that automatically converts
existing MongoDB HelmReleases:
- Detects old format users (with `db` and `roles` fields)
- Converts `readWrite`/`dbAdmin` roles → `databases.{db}.roles.admin`
- Converts `read` role → `databases.{db}.roles.readonly`
- Preserves user passwords in new `users` format
- Skips HelmReleases that don't have old format
### Release note
```release-note
[mongodb] Unified users and databases configuration to match postgres and mysql patterns. Users now defined separately from databases, with role assignments in the databases section. Includes automatic migration for existing deployments.
```
Removed reference to .Values.networking.apiServer which was removed
from values.yaml. Use empty defaults and let ConfigMap or
apiServerEndpoint parsing override them.
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
Extract common system packages shared between isp-full and isp-full-generic
into a helper template `cozystack.platform.system.common-packages`.
Packages moved to helper:
- kubeovn-webhook, kubeovn-plunger, cozy-proxy
- multus, metallb, reloader
- linstor-scheduler, snapshot-controller
Packages NOT in helper (differ between variants):
- networking (variant differs: kubeovn-cilium vs kubeovn-cilium-generic)
- linstor (talos.enabled differs)
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
Per review feedback, create separate manifest files for different
deployment targets instead of using templated values:
- cozystack-operator.yaml (Talos): hardcoded localhost:7445 (KubePrism)
- cozystack-operator-generic.yaml: reads from cozystack-operator-config
ConfigMap (user must create before applying)
- cozystack-operator-hosted.yaml: no env override (uses in-cluster SA)
This keeps installation flow clean - users apply the manifest matching
their deployment target without needing to modify original files.
Build system updated to generate all three variants.
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
- Remove apiServer and cilium.cgroup from user-facing values.yaml
- Hardcode Talos-specific values (localhost:7445, cgroup autoMount: false)
directly in isp-full bundle template
- isp-full-generic already hardcodes its own values
- Improve MASTER_NODES comment explaining helm lookup behavior
This hides implementation details from users as requested in review.
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
The enabledPackages list must use fully qualified package names
(e.g., cozystack.velero) to match the helper template check.
Also add backupstrategy-controller which is required by backup-controller.
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
Add velero to enabledPackages list for isp-full-generic to enable
backup functionality by default.
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
- Allow iaas, paas, naas bundles when using isp-full-generic variant
- Enable all bundles by default in values-isp-full-generic.yaml
- Update bundle templates to accept isp-full-generic alongside isp-full
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
The cozystack-values secret was using the default publishing.host value
from values.yaml (example.org) instead of reading from the cozystack
ConfigMap where the actual root-host is configured.
Add ConfigMap lookup to apps.yaml to read root-host from cozystack
ConfigMap, falling back to chart values if not present.
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
Same issue as lineage-controller-webhook: DaemonSet uses hardcoded
nodeSelector with empty value, but k3s/kubeadm use value "true".
Changes:
- Add nodeSelector to cozystack-api values (default: empty for Talos)
- Update deployment template to use configurable nodeSelector
- Pass nodeSelector for both cozystack-api and lineage-controller-webhook
in isp-full-generic bundle variant
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
On Talos, control-plane nodes have label node-role.kubernetes.io/control-plane
with empty value. On generic k8s (k3s, kubeadm), the same label has value "true".
The lineage-controller-webhook DaemonSet was hardcoded to use empty value,
causing 0 pods scheduled on k3s clusters.
Changes:
- Add nodeSelector to lineage-controller-webhook values (default: empty for Talos)
- Update DaemonSet template to use configurable nodeSelector
- Pass nodeSelector with value "true" for isp-full-generic bundle variant
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
k3s and kubeadm set node-role.kubernetes.io/control-plane=true,
while Talos uses an empty value. KubeOVN node selector needs
the exact label value to match.
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
For isp-full-generic bundle variant, the template now looks up the
cozystack ConfigMap to read cluster-specific settings instead of only
relying on chart default values.
This fixes the issue where Cilium was trying to connect to localhost:7445
(Talos KubePrism default) instead of the actual API server endpoint.
Values read from ConfigMap:
- api-server-endpoint: parsed to extract host/port for Cilium
- ipv4-pod-cidr, ipv4-pod-gateway, ipv4-svc-cidr, ipv4-join-cidr: for KubeOVN
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
Remove trailing dash from {{- end -}} constructs in _helpers.tpl to
preserve trailing newlines after each rendered package YAML document.
Without this fix, consecutive packages are rendered without proper
separators:
variant: default---
apiVersion: cozystack.io/v1alpha1
This causes YAML parse errors: "mapping key 'apiVersion' already
defined" because the YAML parser sees a single document with duplicate
keys instead of two separate documents.
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
The isp-full-generic variant should have iaas, paas, and naas bundles
disabled by default since these bundles require 'isp-full' or 'isp-hosted'
variants as per validation rules in their respective templates.
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
Create dedicated values file for isp-full-generic bundle variant with
bundles.system.variant set to "isp-full-generic" instead of "isp-full".
This fixes the ArtifactGenerator generating identical digests for both
isp-full and isp-full-generic variants, which caused the wrong bundle
template to be selected.
Also adds isp-full-generic variant to installer PackageSource definition.
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
The isp-full-generic bundle now automatically extracts host and port
from publishing.apiServerEndpoint URL for Cilium and KubeOVN
configuration. This eliminates the need for manual networking.apiServer
values when api-server-endpoint is set in the ConfigMap.
Changes:
- Parse apiServerEndpoint URL to extract host and port for Cilium
- Auto-detect MASTER_NODES for KubeOVN from apiServerEndpoint if not
explicitly set
- Hardcode cgroup.autoMount=true for generic k8s (non-Talos)
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
Add isp-full-generic bundle variant for deploying Cozystack on non-Talos
Kubernetes clusters (kubeadm, k3s, RKE2, etc.).
This variant:
- Uses kubeovn-cilium-generic networking (no values-talos.yaml)
- Passes talos.enabled=false to linstor (keeps DRBD init containers)
- Supports all standard networking configuration via platform values
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
Add cilium-generic and kubeovn-cilium-generic variants that exclude
values-talos.yaml. These variants are suitable for kubeadm, k3s, RKE2
and other non-Talos Kubernetes distributions.
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
- Add networking.kubeovn.MASTER_NODES to platform values.yaml
- Conditionally include MASTER_NODES in kubeovn values when set
- Enables KubeOVN deployment on k3s/kubeadm without control-plane node labels
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
Add networking.apiServer.host and networking.apiServer.port to platform values
to allow overriding the default Talos KubePrism settings (localhost:7445).
Also add networking.cilium.cgroup.autoMount for non-Talos clusters that need
Cilium to mount cgroups automatically.
Changes:
- packages/core/platform/values.yaml: Add apiServer and cilium.cgroup settings
- packages/core/platform/templates/bundles/system.yaml: Pass cilium values
- packages/core/installer/values.yaml: Add kubernetesServiceHost/Port
- packages/core/installer/templates/cozystack-operator.yaml: Template env vars
Closes: #1933
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
## What this PR does
Fixes v1.0 migration script
### Release note
```release-note
V1.0 migration script fixed after an upgrade was performed.
```
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **Chores**
* Updated bundle naming convention (paas → isp).
* Restructured configuration schema: flattened branding, renamed
scheduling field, and moved authentication redirect setting.
* Minor output text and prompt wording adjustments.
* **New Features**
* Added support for external IPs in generated configuration with safe
empty fallback.
* Consolidated branding and scheduling entries for simpler output.
* **Bug Fixes**
* Migration script now validates prerequisites and produces improved,
YAML-friendly output.
<sub>✏️ Tip: You can customize this high-level summary in your review
settings.</sub>
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
Add migration script to convert existing MongoDB HelmReleases
from old users format to new users + databases format.
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: Andrei Kvapil <kvapss@gmail.com>
## What this PR does
Fixes v1.0 packagesources.
### Release note
```release-note
V1.0 packagesources fixed after an upgrade was performed.
```
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **Chores**
* Updated platform component configurations for cozy-proxy and
metrics-server deployments.
* Reorganized monitoring component namespace allocation.
* Added metrics server as a dependency in the monitoring pipeline.
* Integrated new system components (linstor-scheduler and
metrics-server) into the platform templates.
* Refined HelmRelease chart references for vertical pod autoscaler.
<sub>✏️ Tip: You can customize this high-level summary in your review
settings.</sub>
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
## What this PR does
Adds MongoDB as a managed application to Cozystack using Percona
Operator for MongoDB (Apache 2.0).
**New packages:**
- `packages/system/mongodb-operator/` - Percona psmdb-operator
deployment
- `packages/apps/mongodb/` - MongoDB application chart
- `packages/system/mongodb-rd/` - CozystackResourceDefinition for
dashboard
**Features:**
- Replica set mode with configurable replica count
- Sharded cluster mode (config servers, mongos routers, multiple shards)
- S3 backup integration with cron scheduling
- Point-in-time recovery from backups
- External access via LoadBalancer
- Multiple MongoDB versions (v6, v7, v8)
- Resource presets (nano to 2xlarge)
- User management with role-based access
### Release note
```release-note
[apps] Add MongoDB as a managed application with replica set, sharding, and S3 backup support
```
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **New Features**
* Managed MongoDB service and operator packaging; replica sets with
optional sharding, external access, per-user credentials, workload
monitoring, and version mappings for provider images
* Backup & restore with S3 support and point-in-time recovery
* **Documentation**
* Comprehensive README plus Helm values and schema for configuration
* **Tests**
* Extensive unit/template tests and end-to-end tests covering
credentials, services, backup, restore, CR rendering, and operator
install
* **Packaging**
* Charts and bundle entries added for operator and managed service
installation
<sub>✏️ Tip: You can customize this high-level summary in your review
settings.</sub>
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
Signed-off-by: Andrei Kvapil <kvapss@gmail.com>
<!-- 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].
- 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.
-->
## What this PR does
### Release note
<!-- Write a release note:
- Explain what has changed internally and for users.
- Start with the same [label] as in the PR title
- Follow the guidelines at
https://github.com/kubernetes/community/blob/master/contributors/guide/release-notes.md.
-->
```release-note
[kilo] Introduce kilo
```
## What this PR does
This PR introduces a new flux-plunger controller that automatically
fixes HelmRelease resources experiencing "has no deployed releases"
errors by cleaning up corrupted Helm release secrets.
### Problem
When Helm releases fail with "has no deployed releases" error, FluxCD
cannot recover automatically. This typically happens when:
- Helm release secrets are corrupted
- Failed upgrades leave the release in an inconsistent state
- Manual intervention is required to delete the problematic secret
### Solution
The flux-plunger controller:
1. Watches HelmRelease resources for "has no deployed releases" errors
2. Suspends the HelmRelease (with `flux-client-side-apply` field
manager)
3. Identifies and deletes the latest (problematic) Helm release secret
4. Tracks processed versions via annotation to prevent recursive
deletion
5. Unsuspends the HelmRelease to allow FluxCD to retry
**Key features:**
- Only deletes one secret at a time (prevents cascading deletions)
- Uses version tracking: if `latest+1 == processed`, skips (already
handled)
- Handles suspended HelmReleases: only unsuspends if it was suspended by
flux-plunger itself
- Gracefully handles invalid annotations by treating them as unprocessed
### Components
- **Controller:** `internal/controller/fluxplunger/flux_plunger.go`
- **Main:** `cmd/flux-plunger/main.go`
- **Helm Chart:** `packages/system/flux-plunger/`
- **Platform Integration:** Added to `paas-full.yaml` bundle
### Release note
```release-note
[platform] Add flux-plunger controller to automatically fix HelmRelease resources with "has no deployed releases" error by cleaning up corrupted Helm secrets
```
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
## Release Notes
* **New Features**
* Introduced a new controller that automatically resolves HelmRelease
resource failures with built-in health probes and graceful recovery
procedures.
* Enabled by default in system installations.
<sub>✏️ Tip: You can customize this high-level summary in your review
settings.</sub>
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
## What this PR does
This PR adds a new package `local-ccm` to the Cozystack platform.
The local-ccm package provides a Cloud Controller Manager for local
development and testing environments. It integrates the upstream Helm
chart from https://github.com/cozystack/local-ccm (version 0.2.0).
**Changes:**
- Created package structure in `packages/system/local-ccm/`
- Added Makefile with automated upstream chart synchronization from
GitHub
- Added PackageSource definition in
`packages/core/platform/sources/local-ccm.yaml`
- Configured dependency on networking components
- Set privileged mode as required for cloud controller operations
**Package details:**
- Namespace: `cozy-local-ccm`
- Release name: `local-ccm`
- Upstream: https://github.com/cozystack/local-ccm
### Release note
```release-note
[local-ccm] Add local-ccm package for cloud controller manager in local environments
```
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **New Features**
* Added Local Cloud Controller Manager package providing automatic node
IP detection, internal and external IP support, initialization taint
removal, and continuous node reconciliation with configurable intervals.
* **Chores**
* Updated system component tolerations and scheduling policies to
improve compatibility with cloud infrastructure, including better
handling of node readiness, network initialization, and cloud provider
states.
<sub>✏️ Tip: You can customize this high-level summary in your review
settings.</sub>
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
Add new package local-ccm from https://github.com/cozystack/local-ccm
- Create package structure in packages/system/local-ccm
- Add Makefile with upstream chart sync from GitHub
- Add PackageSource definition in packages/core/platform/sources/
- Package provides cloud controller manager for local environments
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: Andrei Kvapil <kvapss@gmail.com>
