From 7797f4956918f1631c52714034ee1690b8fd3fd3 Mon Sep 17 00:00:00 2001
From: Aleksei Sviridkin <f@lex.la>
Date: Thu, 16 Apr 2026 20:20:41 +0300
Subject: [PATCH 01/26] test(api): assert parent HelmRelease Install/Upgrade
 Timeout >= 15m

Adds a failing unit test for convertApplicationToHelmRelease asserting
that Install.Timeout and Upgrade.Timeout are at least 15 minutes. The
default flux helm-controller timeout is too short to cover cold-start
Kamaji control-plane bootstrap (image pull + etcd bootstrap + apiserver
Ready + admin-kubeconfig Secret generation) and causes install
remediation loops.

Assisted-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
---
 .../apps/application/rest_timeout_test.go     | 53 +++++++++++++++++++
 1 file changed, 53 insertions(+)
 create mode 100644 pkg/registry/apps/application/rest_timeout_test.go

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..8ae13183
--- /dev/null
+++ b/pkg/registry/apps/application/rest_timeout_test.go
@@ -0,0 +1,53 @@
+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 TestConvertApplicationToHelmRelease_SetsInstallAndUpgradeTimeout(t *testing.T) {
+	r := &REST{
+		releaseConfig: config.ReleaseConfig{
+			Prefix: "kubernetes-",
+			ChartRef: config.ChartRefConfig{
+				Kind:      "HelmChart",
+				Name:      "kubernetes",
+				Namespace: "cozy-system",
+			},
+		},
+	}
+
+	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 {
+		t.Fatal("Spec.Install must not be nil")
+	}
+	if hr.Spec.Install.Timeout == nil {
+		t.Fatal("Spec.Install.Timeout must be set to cover async admin-kubeconfig provisioning")
+	}
+	if hr.Spec.Install.Timeout.Duration < 15*time.Minute {
+		t.Errorf("Spec.Install.Timeout must be >= 15m (cold bootstrap budget), got %v", hr.Spec.Install.Timeout.Duration)
+	}
+
+	if hr.Spec.Upgrade == nil {
+		t.Fatal("Spec.Upgrade must not be nil")
+	}
+	if hr.Spec.Upgrade.Timeout == nil {
+		t.Fatal("Spec.Upgrade.Timeout must be set to cover async admin-kubeconfig provisioning")
+	}
+	if hr.Spec.Upgrade.Timeout.Duration < 15*time.Minute {
+		t.Errorf("Spec.Upgrade.Timeout must be >= 15m (cold bootstrap budget), got %v", hr.Spec.Upgrade.Timeout.Duration)
+	}
+}

From e4f279f8e2cb08e271130b4a5193d9e0e9d9d3a7 Mon Sep 17 00:00:00 2001
From: Aleksei Sviridkin <f@lex.la>
Date: Thu, 16 Apr 2026 20:21:15 +0300
Subject: [PATCH 02/26] fix(api): set 15m Install/Upgrade Timeout for parent
 HelmRelease

Parent HelmRelease created by cozystack-api for Kubernetes tenants
contains CAPI/Kamaji resources (Cluster, KamajiControlPlane,
MachineDeployment) that asynchronously provision the
*-admin-kubeconfig Secret. Three Deployments in the same chart
(cluster-autoscaler, kccm, kcsi-controller) mount that Secret
directly, so the helm-wait cannot complete until control-plane
bootstrap finishes.

Default flux helm-controller timeout is too short for a cold-node
first-tenant bootstrap (image pull + etcd bootstrap + apiserver Ready
+ admin-kubeconfig generation routinely exceed it). On timeout,
install.remediation triggers uninstall, which removes the Cluster CR
and restarts the cycle indefinitely.

Bumping Install.Timeout and Upgrade.Timeout to 15m gives realistic
bootstrap headroom while remaining bounded.

Assisted-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
---
 pkg/registry/apps/application/rest.go | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/pkg/registry/apps/application/rest.go b/pkg/registry/apps/application/rest.go
index 0728ea13..77d1d458 100644
--- a/pkg/registry/apps/application/rest.go
+++ b/pkg/registry/apps/application/rest.go
@@ -1509,11 +1509,13 @@ func (r *REST) convertApplicationToHelmRelease(app *appsv1alpha1.Application) (*
 			},
 			Interval: metav1.Duration{Duration: 5 * time.Minute},
 			Install: &helmv2.Install{
+				Timeout: &metav1.Duration{Duration: 15 * time.Minute},
 				Remediation: &helmv2.InstallRemediation{
 					Retries: -1,
 				},
 			},
 			Upgrade: &helmv2.Upgrade{
+				Timeout: &metav1.Duration{Duration: 15 * time.Minute},
 				Remediation: &helmv2.UpgradeRemediation{
 					Retries: -1,
 				},

From 3e26234a1c98f138542471a7f4395532e15c7b46 Mon Sep 17 00:00:00 2001
From: Aleksei Sviridkin <f@lex.la>
Date: Thu, 16 Apr 2026 20:24:20 +0300
Subject: [PATCH 03/26] test(kubernetes): assert admin-kubeconfig wait pattern
 and etcd guard

Adds failing helm unittest suite for packages/apps/kubernetes covering:

- cluster-autoscaler, kccm, and csi controller Deployments mount the
  admin-kubeconfig Secret with optional: true
- each of those Deployments has a wait-for-kubeconfig initContainer
  that mounts the same kubeconfig path
- cluster.yaml renders a helm fail with a descriptive message when the
  tenant has no etcd DataStore (empty _namespace.etcd)

Also wires up a test target in the chart Makefile so helm-unit-tests.sh
picks it up.

Assisted-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
---
 packages/apps/kubernetes/Makefile             |  3 +
 .../tests/admin_kubeconfig_wait_test.yaml     | 85 +++++++++++++++++++
 packages/apps/kubernetes/tests/values-ci.yaml |  9 ++
 3 files changed, 97 insertions(+)
 create mode 100644 packages/apps/kubernetes/tests/admin_kubeconfig_wait_test.yaml
 create mode 100644 packages/apps/kubernetes/tests/values-ci.yaml

diff --git a/packages/apps/kubernetes/Makefile b/packages/apps/kubernetes/Makefile
index 01cf736d..0f9e6d57 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
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..e2d9f359
--- /dev/null
+++ b/packages/apps/kubernetes/tests/admin_kubeconfig_wait_test.yaml
@@ -0,0 +1,85 @@
+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
+    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
+    asserts:
+      - equal:
+          path: spec.template.spec.initContainers[0].name
+          value: wait-for-kubeconfig
+
+  - it: csi controller mounts admin-kubeconfig as optional
+    template: templates/csi/deploy.yaml
+    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
+    asserts:
+      - equal:
+          path: spec.template.spec.initContainers[0].name
+          value: wait-for-kubeconfig
+
+  - it: cluster.yaml fails render when tenant has no etcd DataStore
+    template: templates/cluster.yaml
+    set:
+      _namespace:
+        etcd: ""
+        monitoring: ""
+        ingress: ""
+        seaweedfs: ""
+        host: ""
+    asserts:
+      - failedTemplate:
+          errorPattern: "requires a parent tenant with etcd enabled"
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

From ca33cc4e3c4b032293fceec2b7d3c1eb13a241bb Mon Sep 17 00:00:00 2001
From: Aleksei Sviridkin <f@lex.la>
Date: Thu, 16 Apr 2026 20:25:42 +0300
Subject: [PATCH 04/26] fix(kubernetes): wait for admin-kubeconfig before
 starting CP-side pods

Three Deployments in the Kubernetes app chart mount the tenant
admin-kubeconfig Secret directly as a volume: cluster-autoscaler,
kccm, and the kcsi controller. That Secret is provisioned
asynchronously by Kamaji after control-plane bootstrap, so on a fresh
install the pods used to hit FailedMount and the parent HelmRelease
ran out of its wait budget.

Mark the Secret volume optional and add a wait-for-kubeconfig
initContainer that polls the mounted path until the Secret appears.
Kubelet remounts the optional Secret within its sync period once
Kamaji publishes it, the init container exits, and the main container
starts cleanly. The Deployment becomes Available without the helm-wait
ever seeing a FailedMount.

Pins a busybox image for the init container via
images/busybox.tag (same format as the other pinned tags in this
chart).

Assisted-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
---
 packages/apps/kubernetes/images/busybox.tag     |  1 +
 .../cluster-autoscaler/deployment.yaml          | 17 +++++++++++++++++
 .../apps/kubernetes/templates/csi/deploy.yaml   | 17 +++++++++++++++++
 .../apps/kubernetes/templates/kccm/manager.yaml | 17 +++++++++++++++++
 4 files changed, 52 insertions(+)
 create mode 100644 packages/apps/kubernetes/images/busybox.tag

diff --git a/packages/apps/kubernetes/images/busybox.tag b/packages/apps/kubernetes/images/busybox.tag
new file mode 100644
index 00000000..e358c12e
--- /dev/null
+++ b/packages/apps/kubernetes/images/busybox.tag
@@ -0,0 +1 @@
+busybox:1.37.0@sha256:1487d0af5f52b4ba31c7e465126ee2123fe3f2305d638e7827681e7cf6c83d5e
diff --git a/packages/apps/kubernetes/templates/cluster-autoscaler/deployment.yaml b/packages/apps/kubernetes/templates/cluster-autoscaler/deployment.yaml
index a00e0155..47c22f76 100644
--- a/packages/apps/kubernetes/templates/cluster-autoscaler/deployment.yaml
+++ b/packages/apps/kubernetes/templates/cluster-autoscaler/deployment.yaml
@@ -23,6 +23,22 @@ spec:
         - key: node-role.kubernetes.io/control-plane
           operator: Exists
           effect: "NoSchedule"
+      initContainers:
+      - name: wait-for-kubeconfig
+        image: "{{ $.Files.Get "images/busybox.tag" | trim }}"
+        command:
+        - sh
+        - -c
+        - |
+          set -eu
+          until [ -s /etc/kubernetes/kubeconfig/super-admin.svc ]; do
+            echo "waiting for admin kubeconfig to be provisioned by Kamaji..."
+            sleep 5
+          done
+        volumeMounts:
+        - name: kubeconfig
+          mountPath: /etc/kubernetes/kubeconfig
+          readOnly: true
       containers:
       - image: "{{ $.Files.Get "images/cluster-autoscaler.tag" | trim }}"
         name: cluster-autoscaler
@@ -56,6 +72,7 @@ spec:
         name: cloud-config
       - secret:
           secretName: {{ .Release.Name }}-admin-kubeconfig
+          optional: true
         name: kubeconfig
       serviceAccountName: {{ .Release.Name }}-cluster-autoscaler
       terminationGracePeriodSeconds: 10
diff --git a/packages/apps/kubernetes/templates/csi/deploy.yaml b/packages/apps/kubernetes/templates/csi/deploy.yaml
index 938b6d67..2979c75c 100644
--- a/packages/apps/kubernetes/templates/csi/deploy.yaml
+++ b/packages/apps/kubernetes/templates/csi/deploy.yaml
@@ -24,6 +24,22 @@ spec:
         - key: node-role.kubernetes.io/control-plane
           operator: Exists
           effect: "NoSchedule"
+      initContainers:
+        - name: wait-for-kubeconfig
+          image: "{{ $.Files.Get "images/busybox.tag" | trim }}"
+          command:
+          - sh
+          - -c
+          - |
+            set -eu
+            until [ -s /etc/kubernetes/kubeconfig/super-admin.svc ]; do
+              echo "waiting for admin kubeconfig to be provisioned by Kamaji..."
+              sleep 5
+            done
+          volumeMounts:
+          - name: kubeconfig
+            mountPath: /etc/kubernetes/kubeconfig
+            readOnly: true
       containers:
         - name: csi-driver
           imagePullPolicy: Always
@@ -234,4 +250,5 @@ spec:
           emptyDir: {}
         - secret:
             secretName: {{ .Release.Name }}-admin-kubeconfig
+            optional: true
           name: kubeconfig
diff --git a/packages/apps/kubernetes/templates/kccm/manager.yaml b/packages/apps/kubernetes/templates/kccm/manager.yaml
index 81426d4e..5c7f15a4 100644
--- a/packages/apps/kubernetes/templates/kccm/manager.yaml
+++ b/packages/apps/kubernetes/templates/kccm/manager.yaml
@@ -22,6 +22,22 @@ spec:
         - key: node-role.kubernetes.io/control-plane
           operator: Exists
           effect: "NoSchedule"
+      initContainers:
+      - name: wait-for-kubeconfig
+        image: "{{ $.Files.Get "images/busybox.tag" | trim }}"
+        command:
+        - sh
+        - -c
+        - |
+          set -eu
+          until [ -s /etc/kubernetes/kubeconfig/super-admin.svc ]; do
+            echo "waiting for admin kubeconfig to be provisioned by Kamaji..."
+            sleep 5
+          done
+        volumeMounts:
+        - name: kubeconfig
+          mountPath: /etc/kubernetes/kubeconfig
+          readOnly: true
       containers:
       - name: kubevirt-cloud-controller-manager
         args:
@@ -55,5 +71,6 @@ spec:
         name: cloud-config
       - secret:
           secretName: {{ .Release.Name }}-admin-kubeconfig
+          optional: true
         name: kubeconfig
       serviceAccountName: {{ .Release.Name }}-kccm

From cac514b60fc00717df37780d2183e1702b07e9c6 Mon Sep 17 00:00:00 2001
From: Aleksei Sviridkin <f@lex.la>
Date: Thu, 16 Apr 2026 20:26:09 +0300
Subject: [PATCH 05/26] fix(kubernetes): fail fast when tenant has no etcd
 DataStore

When a Kubernetes tenant is created without a parent tenant that has
etcd enabled, .Values._namespace.etcd is empty and the rendered
KamajiControlPlane spec carries an empty dataStoreName. The Kamaji
admission webhook then rejects every TenantControlPlane create with
"tenant-root DataStore does not exist" and the control plane never
comes up.

Add a helm template-level guard that fails rendering with a
descriptive, actionable error message before the HelmRelease even
reaches the webhook. This also closes the narrow race where a
Kubernetes HelmRelease reconciles before the etcd HelmRelease has
created the DataStore CR - flux retries on its interval and picks up
the DataStore once it appears.

Assisted-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
---
 packages/apps/kubernetes/templates/cluster.yaml | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/packages/apps/kubernetes/templates/cluster.yaml b/packages/apps/kubernetes/templates/cluster.yaml
index 10d6fd80..99249c1a 100644
--- a/packages/apps/kubernetes/templates/cluster.yaml
+++ b/packages/apps/kubernetes/templates/cluster.yaml
@@ -1,4 +1,7 @@
 {{- $etcd := .Values._namespace.etcd }}
+{{- if not $etcd }}
+{{- fail "Kubernetes tenant requires a parent tenant with etcd enabled: set .Values.etcd=true on tenant-root (or any ancestor tenant), wait for the etcd HelmRelease to reconcile (DataStore CR appears in the tenant namespace), then retry." }}
+{{- end }}
 {{- $ingress := .Values._namespace.ingress }}
 {{- $host := .Values._namespace.host }}
 {{- $kubevirtmachinetemplateNames := list }}

From a7d994365d067dee5d9546a9b95bb555b206a6af Mon Sep 17 00:00:00 2001
From: Aleksei Sviridkin <f@lex.la>
Date: Thu, 16 Apr 2026 20:26:47 +0300
Subject: [PATCH 06/26] test(kubernetes): assert parent HelmRelease did not
 remediate in e2e

Before cleanup, inspect the parent HelmRelease installFailures and
upgradeFailures counters. A non-zero value means flux helm-controller
hit its wait timeout, ran install/upgrade remediation (uninstall),
and re-installed - the exact race condition this PR closes. Fail the
bats test in that case so the signal surfaces in CI instead of being
masked by a green retry.

Assisted-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
---
 hack/e2e-apps/run-kubernetes.sh | 13 +++++++++++++
 1 file changed, 13 insertions(+)

diff --git a/hack/e2e-apps/run-kubernetes.sh b/hack/e2e-apps/run-kubernetes.sh
index be6dcd6f..f05ec370 100644
--- a/hack/e2e-apps/run-kubernetes.sh
+++ b/hack/e2e-apps/run-kubernetes.sh
@@ -320,6 +320,19 @@ 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.
+  install_failures=$(kubectl get hr -n tenant-test "kubernetes-${test_name}" -ojsonpath='{.status.installFailures}')
+  upgrade_failures=$(kubectl get hr -n tenant-test "kubernetes-${test_name}" -ojsonpath='{.status.upgradeFailures}')
+  if [ "${install_failures:-0}" != "0" ] && [ -n "${install_failures}" ] || \
+     [ "${upgrade_failures:-0}" != "0" ] && [ -n "${upgrade_failures}" ]; then
+    echo "Parent HelmRelease entered remediation cycle: installFailures=${install_failures:-0}, upgradeFailures=${upgrade_failures:-0}" >&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}"

From f87834a3edba32945f22e8d419a9592e5303b34b Mon Sep 17 00:00:00 2001
From: Aleksei Sviridkin <f@lex.la>
Date: Thu, 16 Apr 2026 20:48:47 +0300
Subject: [PATCH 07/26] fix(hack): group e2e remediation guard conditions
 correctly

Shell && and || have equal precedence and left-to-right associativity,
so the previous guard parsed as (((A && B) || C) && D) and silently
passed on the canonical failure mode: install_failures=1 with an empty
upgrade_failures.

Extract the check into helmrelease_has_remediation_cycle() in a
dedicated helper sourced from run-kubernetes.sh, and add unit tests
under hack/remediation-guard.bats that pin the expected behavior for
every combination of empty, zero, and positive counter values.

Assisted-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
---
 hack/e2e-apps/remediation-guard.sh | 24 +++++++++
 hack/e2e-apps/run-kubernetes.sh    |  5 +-
 hack/remediation-guard.bats        | 87 ++++++++++++++++++++++++++++++
 3 files changed, 114 insertions(+), 2 deletions(-)
 create mode 100644 hack/e2e-apps/remediation-guard.sh
 create mode 100644 hack/remediation-guard.bats

diff --git a/hack/e2e-apps/remediation-guard.sh b/hack/e2e-apps/remediation-guard.sh
new file mode 100644
index 00000000..6da69f13
--- /dev/null
+++ b/hack/e2e-apps/remediation-guard.sh
@@ -0,0 +1,24 @@
+# Helpers for asserting that a Flux HelmRelease did not fall into an
+# install/upgrade remediation cycle during an e2e run.
+#
+# A non-zero installFailures/upgradeFailures counter means flux
+# helm-controller hit its wait timeout, ran remediation (uninstall),
+# and re-installed. That is exactly the race this guard is meant to
+# catch, so the function returns success (0) when a cycle is detected
+# and failure (1) otherwise.
+#
+# Both arguments may be empty strings, the literal "0", or a positive
+# integer. Shell's && and || have equal precedence with left-to-right
+# associativity, so each half of the disjunction is grouped explicitly
+# to avoid (A && B) || C && D parsing that masks the common
+# install_failures=1, upgrade_failures="" case.
+
+helmrelease_has_remediation_cycle() {
+    install_failures="$1"
+    upgrade_failures="$2"
+    if { [ -n "${install_failures}" ] && [ "${install_failures}" != "0" ]; } || \
+       { [ -n "${upgrade_failures}" ] && [ "${upgrade_failures}" != "0" ]; }; then
+        return 0
+    fi
+    return 1
+}
diff --git a/hack/e2e-apps/run-kubernetes.sh b/hack/e2e-apps/run-kubernetes.sh
index f05ec370..811f42a7 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="$2"
@@ -326,8 +328,7 @@ EOF
   # and churn the Cluster CR.
   install_failures=$(kubectl get hr -n tenant-test "kubernetes-${test_name}" -ojsonpath='{.status.installFailures}')
   upgrade_failures=$(kubectl get hr -n tenant-test "kubernetes-${test_name}" -ojsonpath='{.status.upgradeFailures}')
-  if [ "${install_failures:-0}" != "0" ] && [ -n "${install_failures}" ] || \
-     [ "${upgrade_failures:-0}" != "0" ] && [ -n "${upgrade_failures}" ]; then
+  if helmrelease_has_remediation_cycle "${install_failures}" "${upgrade_failures}"; then
     echo "Parent HelmRelease entered remediation cycle: installFailures=${install_failures:-0}, upgradeFailures=${upgrade_failures:-0}" >&2
     kubectl -n tenant-test describe hr "kubernetes-${test_name}" >&2
     exit 1
diff --git a/hack/remediation-guard.bats b/hack/remediation-guard.bats
new file mode 100644
index 00000000..64aa83af
--- /dev/null
+++ b/hack/remediation-guard.bats
@@ -0,0 +1,87 @@
+#!/usr/bin/env bats
+# -----------------------------------------------------------------------------
+# Unit tests for hack/e2e-apps/remediation-guard.sh
+#
+# helmrelease_has_remediation_cycle is consumed from e2e tests to assert that
+# the parent HelmRelease did not hit flux helm-controller's wait timeout and
+# enter uninstall remediation. The function accepts two arguments (values of
+# .status.installFailures and .status.upgradeFailures) and returns 0 when a
+# remediation cycle is detected, 1 otherwise.
+#
+# Each argument can be empty (controller never populated the field), "0"
+# (populated but never failed), or a positive integer. Shell's && and ||
+# have equal precedence with left-to-right associativity, which used to
+# break this check on the most common failure mode - install_failures=1
+# and upgrade_failures="". These tests pin the correct behavior.
+#
+# cozytest.sh's awk parser recognizes only @test blocks and a bare `}` on
+# its own line; there is no bats `run` or `$status`. Assertions are
+# expressed as direct shell tests that exit non-zero on failure.
+#
+# Run with: hack/cozytest.sh hack/remediation-guard.bats
+# -----------------------------------------------------------------------------
+
+@test "no counters set returns not-detected" {
+    . hack/e2e-apps/remediation-guard.sh
+    rc=0
+    helmrelease_has_remediation_cycle "" "" || rc=$?
+    [ "$rc" -eq 1 ]
+}
+
+@test "both counters zero returns not-detected" {
+    . hack/e2e-apps/remediation-guard.sh
+    rc=0
+    helmrelease_has_remediation_cycle "0" "0" || rc=$?
+    [ "$rc" -eq 1 ]
+}
+
+@test "install zero upgrade empty returns not-detected" {
+    . hack/e2e-apps/remediation-guard.sh
+    rc=0
+    helmrelease_has_remediation_cycle "0" "" || rc=$?
+    [ "$rc" -eq 1 ]
+}
+
+@test "install empty upgrade zero returns not-detected" {
+    . hack/e2e-apps/remediation-guard.sh
+    rc=0
+    helmrelease_has_remediation_cycle "" "0" || rc=$?
+    [ "$rc" -eq 1 ]
+}
+
+@test "install one upgrade empty returns detected" {
+    # Canonical race: first install exceeded helm-wait, remediation fired,
+    # no upgrade has happened yet.
+    . hack/e2e-apps/remediation-guard.sh
+    rc=0
+    helmrelease_has_remediation_cycle "1" "" || rc=$?
+    [ "$rc" -eq 0 ]
+}
+
+@test "install empty upgrade one returns detected" {
+    . hack/e2e-apps/remediation-guard.sh
+    rc=0
+    helmrelease_has_remediation_cycle "" "1" || rc=$?
+    [ "$rc" -eq 0 ]
+}
+
+@test "install two upgrade zero returns detected" {
+    . hack/e2e-apps/remediation-guard.sh
+    rc=0
+    helmrelease_has_remediation_cycle "2" "0" || rc=$?
+    [ "$rc" -eq 0 ]
+}
+
+@test "install zero upgrade two returns detected" {
+    . hack/e2e-apps/remediation-guard.sh
+    rc=0
+    helmrelease_has_remediation_cycle "0" "2" || rc=$?
+    [ "$rc" -eq 0 ]
+}
+
+@test "both counters positive returns detected" {
+    . hack/e2e-apps/remediation-guard.sh
+    rc=0
+    helmrelease_has_remediation_cycle "3" "5" || rc=$?
+    [ "$rc" -eq 0 ]
+}

From 73b80bfb940ec24459a826b32540f676a00bca36 Mon Sep 17 00:00:00 2001
From: Aleksei Sviridkin <f@lex.la>
Date: Thu, 16 Apr 2026 20:51:32 +0300
Subject: [PATCH 08/26] fix(api): scope 15m helm wait budget to Kubernetes
 Application kind

The previous change set Install.Timeout and Upgrade.Timeout to 15m
on every Application's parent HelmRelease, but the admin-kubeconfig
race documented in #2412 is specific to the Kubernetes kind: only its
parent chart creates CAPI/Kamaji resources whose admin-kubeconfig
Secret is asynchronously provisioned and mounted by Deployments in
the same chart. Other kinds (Qdrant, MongoDB, Postgres, ...) have no
such race and should not have their failed installs linger three times
longer before flux triggers remediation.

Gate the timeout on r.kindName == "Kubernetes". Rewrite
rest_timeout_test.go to cover both sides: Kubernetes must get a
>= 15m timeout, other kinds must keep the flux defaults. Both tests
also pin Install/Upgrade Remediation.Retries == -1 so a future edit
that removes unbounded remediation would show up as a red test.

Assisted-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
---
 pkg/registry/apps/application/rest.go         | 16 ++++-
 .../apps/application/rest_timeout_test.go     | 58 ++++++++++++++++---
 2 files changed, 64 insertions(+), 10 deletions(-)

diff --git a/pkg/registry/apps/application/rest.go b/pkg/registry/apps/application/rest.go
index 77d1d458..9b92d2c0 100644
--- a/pkg/registry/apps/application/rest.go
+++ b/pkg/registry/apps/application/rest.go
@@ -1509,13 +1509,11 @@ func (r *REST) convertApplicationToHelmRelease(app *appsv1alpha1.Application) (*
 			},
 			Interval: metav1.Duration{Duration: 5 * time.Minute},
 			Install: &helmv2.Install{
-				Timeout: &metav1.Duration{Duration: 15 * time.Minute},
 				Remediation: &helmv2.InstallRemediation{
 					Retries: -1,
 				},
 			},
 			Upgrade: &helmv2.Upgrade{
-				Timeout: &metav1.Duration{Duration: 15 * time.Minute},
 				Remediation: &helmv2.UpgradeRemediation{
 					Retries: -1,
 				},
@@ -1530,6 +1528,20 @@ func (r *REST) convertApplicationToHelmRelease(app *appsv1alpha1.Application) (*
 		},
 	}
 
+	// The Kubernetes Application's parent chart creates CAPI/Kamaji resources
+	// whose admin-kubeconfig Secret is provisioned asynchronously and mounted
+	// by Deployments in the same chart. On a cold node, Kamaji control-plane
+	// bootstrap routinely exceeds flux helm-controller's default wait budget,
+	// so install remediation loops uninstall the Cluster CR and churn. Extend
+	// the wait budget to 15m for Kubernetes only - other Application kinds
+	// without this race keep flux defaults, so their failed installs do not
+	// linger unnecessarily before remediation fires.
+	if r.kindName == "Kubernetes" {
+		timeout := metav1.Duration{Duration: 15 * time.Minute}
+		helmRelease.Spec.Install.Timeout = &timeout
+		helmRelease.Spec.Upgrade.Timeout = &timeout
+	}
+
 	return helmRelease, nil
 }
 
diff --git a/pkg/registry/apps/application/rest_timeout_test.go b/pkg/registry/apps/application/rest_timeout_test.go
index 8ae13183..fa6d5427 100644
--- a/pkg/registry/apps/application/rest_timeout_test.go
+++ b/pkg/registry/apps/application/rest_timeout_test.go
@@ -10,18 +10,22 @@ import (
 	"github.com/cozystack/cozystack/pkg/config"
 )
 
-func TestConvertApplicationToHelmRelease_SetsInstallAndUpgradeTimeout(t *testing.T) {
-	r := &REST{
+func newRESTForKind(kind, prefix string) *REST {
+	return &REST{
+		kindName: kind,
 		releaseConfig: config.ReleaseConfig{
-			Prefix: "kubernetes-",
+			Prefix: prefix,
 			ChartRef: config.ChartRefConfig{
 				Kind:      "HelmChart",
-				Name:      "kubernetes",
+				Name:      "x",
 				Namespace: "cozy-system",
 			},
 		},
 	}
+}
 
+func TestConvertApplicationToHelmRelease_KubernetesKindGetsLongTimeout(t *testing.T) {
+	r := newRESTForKind("Kubernetes", "kubernetes-")
 	app := &appsv1alpha1.Application{
 		ObjectMeta: metav1.ObjectMeta{Name: "example", Namespace: "tenant-root"},
 	}
@@ -35,19 +39,57 @@ func TestConvertApplicationToHelmRelease_SetsInstallAndUpgradeTimeout(t *testing
 		t.Fatal("Spec.Install must not be nil")
 	}
 	if hr.Spec.Install.Timeout == nil {
-		t.Fatal("Spec.Install.Timeout must be set to cover async admin-kubeconfig provisioning")
+		t.Fatal("Spec.Install.Timeout must be set for Kubernetes kind")
 	}
 	if hr.Spec.Install.Timeout.Duration < 15*time.Minute {
-		t.Errorf("Spec.Install.Timeout must be >= 15m (cold bootstrap budget), got %v", hr.Spec.Install.Timeout.Duration)
+		t.Errorf("Spec.Install.Timeout must be >= 15m for Kubernetes, got %v", hr.Spec.Install.Timeout.Duration)
 	}
 
 	if hr.Spec.Upgrade == nil {
 		t.Fatal("Spec.Upgrade must not be nil")
 	}
 	if hr.Spec.Upgrade.Timeout == nil {
-		t.Fatal("Spec.Upgrade.Timeout must be set to cover async admin-kubeconfig provisioning")
+		t.Fatal("Spec.Upgrade.Timeout must be set for Kubernetes kind")
 	}
 	if hr.Spec.Upgrade.Timeout.Duration < 15*time.Minute {
-		t.Errorf("Spec.Upgrade.Timeout must be >= 15m (cold bootstrap budget), got %v", hr.Spec.Upgrade.Timeout.Duration)
+		t.Errorf("Spec.Upgrade.Timeout must be >= 15m for Kubernetes, 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)
+	}
+}
+
+func TestConvertApplicationToHelmRelease_NonKubernetesKindKeepsFluxDefaults(t *testing.T) {
+	// For Applications whose parent chart has no admin-kubeconfig race
+	// (Qdrant, MongoDB, Postgres, etc.), do NOT extend the helm-wait
+	// budget - otherwise failed installs would block three times as long
+	// before Flux starts remediating.
+	r := newRESTForKind("Qdrant", "qdrant-")
+	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.Install.Timeout != nil {
+		t.Errorf("Spec.Install.Timeout must be unset for non-Kubernetes kinds, got %v", hr.Spec.Install.Timeout.Duration)
+	}
+	if hr.Spec.Upgrade != nil && hr.Spec.Upgrade.Timeout != nil {
+		t.Errorf("Spec.Upgrade.Timeout must be unset for non-Kubernetes kinds, got %v", hr.Spec.Upgrade.Timeout.Duration)
+	}
+
+	// But remediation must still be -1 across the board.
+	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)
 	}
 }

From 03606091df5672b6e857057a349875d534990f3a Mon Sep 17 00:00:00 2001
From: Aleksei Sviridkin <f@lex.la>
Date: Thu, 16 Apr 2026 20:53:17 +0300
Subject: [PATCH 09/26] chore(kubernetes): align busybox image with project
 convention

Adds images/busybox/Dockerfile and an image-busybox Makefile target
that mirror the same pattern as the rest of this chart's images (the
Dockerfile pins the upstream busybox by digest; the Makefile target
builds and tags for ghcr.io/cozystack/cozystack/busybox the same way
cluster-autoscaler et al. are handled). Also wires it into the
umbrella image target so 'make image' rebuilds everything.

Until the first release build runs image-busybox and rewrites the
.tag to point at ghcr.io, the .tag keeps a fully-qualified
docker.io/library/busybox:1.37.0@sha256:... reference so pods do not
silently resolve the short name via the default registry and pulls
remain immutable by digest. The release workflow overwrites this
file the same way it does for the other images.

Assisted-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
---
 packages/apps/kubernetes/Makefile                  | 14 +++++++++++++-
 packages/apps/kubernetes/images/busybox.tag        |  2 +-
 packages/apps/kubernetes/images/busybox/Dockerfile |  1 +
 3 files changed, 15 insertions(+), 2 deletions(-)
 create mode 100644 packages/apps/kubernetes/images/busybox/Dockerfile

diff --git a/packages/apps/kubernetes/Makefile b/packages/apps/kubernetes/Makefile
index 0f9e6d57..eb23c2ef 100644
--- a/packages/apps/kubernetes/Makefile
+++ b/packages/apps/kubernetes/Makefile
@@ -15,7 +15,7 @@ update:
 	hack/update-versions.sh
 	make generate
 
-image: image-ubuntu-container-disk image-kubevirt-cloud-provider image-kubevirt-csi-driver image-cluster-autoscaler
+image: image-ubuntu-container-disk image-kubevirt-cloud-provider image-kubevirt-csi-driver image-cluster-autoscaler image-busybox
 
 image-ubuntu-container-disk:
 	$(foreach ver,$(KUBERNETES_VERSIONS), \
@@ -70,3 +70,15 @@ image-cluster-autoscaler:
 	echo "$(REGISTRY)/cluster-autoscaler:$(call settag,$(KUBERNETES_PKG_TAG))@$$(yq e '."containerimage.digest"' images/cluster-autoscaler.json -o json -r)" \
 		> images/cluster-autoscaler.tag
 	rm -f images/cluster-autoscaler.json
+
+image-busybox:
+	docker buildx build images/busybox \
+		--tag $(REGISTRY)/busybox:$(call settag,$(KUBERNETES_PKG_TAG)) \
+		--tag $(REGISTRY)/busybox:$(call settag,$(KUBERNETES_PKG_TAG)-$(TAG)) \
+		--cache-from type=registry,ref=$(REGISTRY)/busybox:latest \
+		--cache-to type=inline \
+		--metadata-file images/busybox.json \
+		$(BUILDX_ARGS)
+	echo "$(REGISTRY)/busybox:$(call settag,$(KUBERNETES_PKG_TAG))@$$(yq e '."containerimage.digest"' images/busybox.json -o json -r)" \
+		> images/busybox.tag
+	rm -f images/busybox.json
diff --git a/packages/apps/kubernetes/images/busybox.tag b/packages/apps/kubernetes/images/busybox.tag
index e358c12e..39de220a 100644
--- a/packages/apps/kubernetes/images/busybox.tag
+++ b/packages/apps/kubernetes/images/busybox.tag
@@ -1 +1 @@
-busybox:1.37.0@sha256:1487d0af5f52b4ba31c7e465126ee2123fe3f2305d638e7827681e7cf6c83d5e
+docker.io/library/busybox:1.37.0@sha256:1487d0af5f52b4ba31c7e465126ee2123fe3f2305d638e7827681e7cf6c83d5e
diff --git a/packages/apps/kubernetes/images/busybox/Dockerfile b/packages/apps/kubernetes/images/busybox/Dockerfile
new file mode 100644
index 00000000..0c57fd39
--- /dev/null
+++ b/packages/apps/kubernetes/images/busybox/Dockerfile
@@ -0,0 +1 @@
+FROM docker.io/library/busybox:1.37.0@sha256:1487d0af5f52b4ba31c7e465126ee2123fe3f2305d638e7827681e7cf6c83d5e

From 97696b2b036be79ecafd8e96401dc2af3bc39d32 Mon Sep 17 00:00:00 2001
From: Aleksei Sviridkin <f@lex.la>
Date: Thu, 16 Apr 2026 20:54:10 +0300
Subject: [PATCH 10/26] test(kubernetes): add positive cluster render test and
 pin document kind

Adds an assertion that cluster.yaml renders successfully when
_namespace.etcd is set and produces a KamajiControlPlane whose
dataStoreName equals the tenant's etcd DataStore name. Without this
positive case a future edit that inverts or removes the existing etcd
guard would pass the suite as long as the negative case still fails.

Also adds documentSelector: kind=Deployment to the kccm and csi
controller assertions so the jsonpath filter operates on a single
document, matching the cluster-autoscaler case and removing
reliance on helm-unittest filter-vs-single-value coercion.

Assisted-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
---
 .../tests/admin_kubeconfig_wait_test.yaml     | 34 +++++++++++++++++++
 1 file changed, 34 insertions(+)

diff --git a/packages/apps/kubernetes/tests/admin_kubeconfig_wait_test.yaml b/packages/apps/kubernetes/tests/admin_kubeconfig_wait_test.yaml
index e2d9f359..3b2b1d72 100644
--- a/packages/apps/kubernetes/tests/admin_kubeconfig_wait_test.yaml
+++ b/packages/apps/kubernetes/tests/admin_kubeconfig_wait_test.yaml
@@ -39,6 +39,9 @@ tests:
 
   - 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
@@ -49,13 +52,25 @@ tests:
 
   - 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
@@ -66,10 +81,29 @@ tests:
 
   - 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: 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 fails render when tenant has no etcd DataStore
     template: templates/cluster.yaml

From 6afc0eb370f827c55f478045d1aff2a329b25100 Mon Sep 17 00:00:00 2001
From: Aleksei Sviridkin <f@lex.la>
Date: Thu, 16 Apr 2026 20:55:17 +0300
Subject: [PATCH 11/26] fix(kubernetes): bound init wait and reword fail-fast
 message

Cap the wait-for-kubeconfig init container at 20m. If Kamaji genuinely
fails to produce the admin-kubeconfig Secret (misconfigured tenant,
etcd outage after the guard already passed, Kamaji crash-loop), the
pod now exits non-zero and goes CrashLoopBackOff so the failure is
visible in dashboards, instead of silently sleeping in Init forever
and leaving only the flux helm-wait timeout to surface the problem.

Reword the etcd DataStore guard to reference the parent Tenant
application's etcd flag (not .Values.etcd of the Kubernetes chart,
which is a different chart). Update the helm unittest errorPattern
accordingly.

Assisted-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
---
 .../kubernetes/templates/cluster-autoscaler/deployment.yaml  | 5 +++++
 packages/apps/kubernetes/templates/cluster.yaml              | 2 +-
 packages/apps/kubernetes/templates/csi/deploy.yaml           | 5 +++++
 packages/apps/kubernetes/templates/kccm/manager.yaml         | 5 +++++
 .../apps/kubernetes/tests/admin_kubeconfig_wait_test.yaml    | 2 +-
 5 files changed, 17 insertions(+), 2 deletions(-)

diff --git a/packages/apps/kubernetes/templates/cluster-autoscaler/deployment.yaml b/packages/apps/kubernetes/templates/cluster-autoscaler/deployment.yaml
index 47c22f76..277b6df8 100644
--- a/packages/apps/kubernetes/templates/cluster-autoscaler/deployment.yaml
+++ b/packages/apps/kubernetes/templates/cluster-autoscaler/deployment.yaml
@@ -31,7 +31,12 @@ spec:
         - -c
         - |
           set -eu
+          deadline=$(( $(date +%s) + 1200 ))
           until [ -s /etc/kubernetes/kubeconfig/super-admin.svc ]; do
+            if [ "$(date +%s)" -ge "$deadline" ]; then
+              echo "admin kubeconfig was not provisioned within 20m; exiting so the pod goes CrashLoopBackOff and surfaces in dashboards" >&2
+              exit 1
+            fi
             echo "waiting for admin kubeconfig to be provisioned by Kamaji..."
             sleep 5
           done
diff --git a/packages/apps/kubernetes/templates/cluster.yaml b/packages/apps/kubernetes/templates/cluster.yaml
index 99249c1a..1747df4e 100644
--- a/packages/apps/kubernetes/templates/cluster.yaml
+++ b/packages/apps/kubernetes/templates/cluster.yaml
@@ -1,6 +1,6 @@
 {{- $etcd := .Values._namespace.etcd }}
 {{- if not $etcd }}
-{{- fail "Kubernetes tenant requires a parent tenant with etcd enabled: set .Values.etcd=true on tenant-root (or any ancestor tenant), wait for the etcd HelmRelease to reconcile (DataStore CR appears in the tenant namespace), then retry." }}
+{{- fail "Kubernetes tenant requires a parent Tenant application with etcd: true so the etcd module deploys a DataStore CR in the tenant namespace. Set spec.etcd: true on the root Tenant (or any ancestor Tenant), wait for the etcd HelmRelease to reconcile and the DataStore CR to appear, then retry." }}
 {{- end }}
 {{- $ingress := .Values._namespace.ingress }}
 {{- $host := .Values._namespace.host }}
diff --git a/packages/apps/kubernetes/templates/csi/deploy.yaml b/packages/apps/kubernetes/templates/csi/deploy.yaml
index 2979c75c..ff170cb8 100644
--- a/packages/apps/kubernetes/templates/csi/deploy.yaml
+++ b/packages/apps/kubernetes/templates/csi/deploy.yaml
@@ -32,7 +32,12 @@ spec:
           - -c
           - |
             set -eu
+            deadline=$(( $(date +%s) + 1200 ))
             until [ -s /etc/kubernetes/kubeconfig/super-admin.svc ]; do
+              if [ "$(date +%s)" -ge "$deadline" ]; then
+                echo "admin kubeconfig was not provisioned within 20m; exiting so the pod goes CrashLoopBackOff and surfaces in dashboards" >&2
+                exit 1
+              fi
               echo "waiting for admin kubeconfig to be provisioned by Kamaji..."
               sleep 5
             done
diff --git a/packages/apps/kubernetes/templates/kccm/manager.yaml b/packages/apps/kubernetes/templates/kccm/manager.yaml
index 5c7f15a4..67321647 100644
--- a/packages/apps/kubernetes/templates/kccm/manager.yaml
+++ b/packages/apps/kubernetes/templates/kccm/manager.yaml
@@ -30,7 +30,12 @@ spec:
         - -c
         - |
           set -eu
+          deadline=$(( $(date +%s) + 1200 ))
           until [ -s /etc/kubernetes/kubeconfig/super-admin.svc ]; do
+            if [ "$(date +%s)" -ge "$deadline" ]; then
+              echo "admin kubeconfig was not provisioned within 20m; exiting so the pod goes CrashLoopBackOff and surfaces in dashboards" >&2
+              exit 1
+            fi
             echo "waiting for admin kubeconfig to be provisioned by Kamaji..."
             sleep 5
           done
diff --git a/packages/apps/kubernetes/tests/admin_kubeconfig_wait_test.yaml b/packages/apps/kubernetes/tests/admin_kubeconfig_wait_test.yaml
index 3b2b1d72..2a43e13f 100644
--- a/packages/apps/kubernetes/tests/admin_kubeconfig_wait_test.yaml
+++ b/packages/apps/kubernetes/tests/admin_kubeconfig_wait_test.yaml
@@ -116,4 +116,4 @@ tests:
         host: ""
     asserts:
       - failedTemplate:
-          errorPattern: "requires a parent tenant with etcd enabled"
+          errorPattern: "requires a parent Tenant application with etcd: true"

From b38ae605495ff30a71013037af94839c4944e815 Mon Sep 17 00:00:00 2001
From: Aleksei Sviridkin <f@lex.la>
Date: Thu, 16 Apr 2026 21:05:23 +0300
Subject: [PATCH 12/26] fix(kubernetes): soft-skip cluster resources when
 tenant has no DataStore

A hard helm fail in cluster.yaml made every cold bootstrap racy: if
the parent Tenant chart had not yet populated _namespace.etcd in
cozystack-values when the Kubernetes HelmRelease first reconciled, the
fail fired, install.remediation triggered, installFailures incremented
and the new e2e remediation-guard flagged it as a bug. That directly
contradicts the race the rest of this PR is trying to close.

Replace fail with a graceful skip: render only a status-beacon
ConfigMap (test-awaiting-etcd) when etcd is empty, wrap all
CAPI/Kamaji resources in {{ if $etcd }}. The HelmRelease installs
successfully and goes Ready; flux retries on its 5m interval and picks
up the DataStore as soon as the Tenant chart finishes reconciling.

Update the helm unittest: positive test still asserts dataStoreName on
KamajiControlPlane; the negative test now asserts exactly one
ConfigMap document with status=awaiting-etcd, no Cluster / KCP /
KubevirtCluster / WorkloadMonitor rendered.

Assisted-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
---
 .../apps/kubernetes/templates/cluster.yaml    | 37 +++++++++++++++++--
 .../tests/admin_kubeconfig_wait_test.yaml     | 19 ++++++++--
 2 files changed, 50 insertions(+), 6 deletions(-)

diff --git a/packages/apps/kubernetes/templates/cluster.yaml b/packages/apps/kubernetes/templates/cluster.yaml
index 1747df4e..12da9818 100644
--- a/packages/apps/kubernetes/templates/cluster.yaml
+++ b/packages/apps/kubernetes/templates/cluster.yaml
@@ -1,7 +1,15 @@
 {{- $etcd := .Values._namespace.etcd }}
-{{- if not $etcd }}
-{{- fail "Kubernetes tenant requires a parent Tenant application with etcd: true so the etcd module deploys a DataStore CR in the tenant namespace. Set spec.etcd: true on the root Tenant (or any ancestor Tenant), wait for the etcd HelmRelease to reconcile and the DataStore CR to appear, then retry." }}
-{{- end }}
+{{- /*
+  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 }}
@@ -87,6 +95,28 @@ 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 }}
+  annotations:
+    cozystack.io/status-beacon: "true"
+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
@@ -407,3 +437,4 @@ metadata:
 spec:
   {{- .spec | toYaml | nindent 2 }}
 {{- end }}
+{{- end }}
diff --git a/packages/apps/kubernetes/tests/admin_kubeconfig_wait_test.yaml b/packages/apps/kubernetes/tests/admin_kubeconfig_wait_test.yaml
index 2a43e13f..e7c03bd1 100644
--- a/packages/apps/kubernetes/tests/admin_kubeconfig_wait_test.yaml
+++ b/packages/apps/kubernetes/tests/admin_kubeconfig_wait_test.yaml
@@ -105,7 +105,12 @@ tests:
           path: spec.dataStoreName
           value: tenant-root
 
-  - it: cluster.yaml fails render when tenant has no etcd DataStore
+  - 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:
@@ -115,5 +120,13 @@ tests:
         seaweedfs: ""
         host: ""
     asserts:
-      - failedTemplate:
-          errorPattern: "requires a parent Tenant application with etcd: true"
+      - hasDocuments:
+          count: 1
+      - isKind:
+          of: ConfigMap
+      - equal:
+          path: metadata.name
+          value: test-awaiting-etcd
+      - equal:
+          path: data.status
+          value: awaiting-etcd

From 1757567218ac313a030da1f82e64541a3f9bfbae Mon Sep 17 00:00:00 2001
From: Aleksei Sviridkin <f@lex.la>
Date: Thu, 16 Apr 2026 21:06:47 +0300
Subject: [PATCH 13/26] refactor(kubernetes): extract wait-for-kubeconfig init
 into shared helper

The three control-plane-side Deployments (cluster-autoscaler, kccm,
kcsi-controller) carried three copies of the same 20-line init
container. That already drifted: the CSI copy used 4-space nesting
while the other two used 2-space. Any future update to the image,
the deadline, or the poll script had to land in three places or
silently diverge.

Extract the block into a new kubernetes.waitForAdminKubeconfig helper
in templates/_helpers.tpl and include it at each call site. Tighten
the deadline from 20m to 10m so it stays strictly below the 15m
HelmRelease Install.Timeout and the CrashLoopBackOff surfaces in
dashboards before flux remediation can fire. Also clarify the wait
message so operators debugging a stuck init container do not chase
Kamaji for what is actually kubelet's optional-Secret refresh cadence.

Assisted-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
---
 .../apps/kubernetes/templates/_helpers.tpl    | 40 +++++++++++++++++++
 .../cluster-autoscaler/deployment.yaml        | 21 +---------
 .../apps/kubernetes/templates/csi/deploy.yaml | 21 +---------
 .../kubernetes/templates/kccm/manager.yaml    | 21 +---------
 4 files changed, 43 insertions(+), 60 deletions(-)

diff --git a/packages/apps/kubernetes/templates/_helpers.tpl b/packages/apps/kubernetes/templates/_helpers.tpl
index 36c06b64..89f06934 100644
--- a/packages/apps/kubernetes/templates/_helpers.tpl
+++ b/packages/apps/kubernetes/templates/_helpers.tpl
@@ -49,3 +49,43 @@ 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 scoped to the Kubernetes Application kind so the
+CrashLoopBackOff surfaces before flux remediation fires and uninstalls
+the Cluster CR.
+
+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: "{{ .Files.Get "images/busybox.tag" | trim }}"
+  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 277b6df8..348b017d 100644
--- a/packages/apps/kubernetes/templates/cluster-autoscaler/deployment.yaml
+++ b/packages/apps/kubernetes/templates/cluster-autoscaler/deployment.yaml
@@ -24,26 +24,7 @@ spec:
           operator: Exists
           effect: "NoSchedule"
       initContainers:
-      - name: wait-for-kubeconfig
-        image: "{{ $.Files.Get "images/busybox.tag" | trim }}"
-        command:
-        - sh
-        - -c
-        - |
-          set -eu
-          deadline=$(( $(date +%s) + 1200 ))
-          until [ -s /etc/kubernetes/kubeconfig/super-admin.svc ]; do
-            if [ "$(date +%s)" -ge "$deadline" ]; then
-              echo "admin kubeconfig was not provisioned within 20m; exiting so the pod goes CrashLoopBackOff and surfaces in dashboards" >&2
-              exit 1
-            fi
-            echo "waiting for admin kubeconfig to be provisioned by Kamaji..."
-            sleep 5
-          done
-        volumeMounts:
-        - name: kubeconfig
-          mountPath: /etc/kubernetes/kubeconfig
-          readOnly: true
+      {{- include "kubernetes.waitForAdminKubeconfig" $ | nindent 6 }}
       containers:
       - image: "{{ $.Files.Get "images/cluster-autoscaler.tag" | trim }}"
         name: cluster-autoscaler
diff --git a/packages/apps/kubernetes/templates/csi/deploy.yaml b/packages/apps/kubernetes/templates/csi/deploy.yaml
index ff170cb8..c1af7f13 100644
--- a/packages/apps/kubernetes/templates/csi/deploy.yaml
+++ b/packages/apps/kubernetes/templates/csi/deploy.yaml
@@ -25,26 +25,7 @@ spec:
           operator: Exists
           effect: "NoSchedule"
       initContainers:
-        - name: wait-for-kubeconfig
-          image: "{{ $.Files.Get "images/busybox.tag" | trim }}"
-          command:
-          - sh
-          - -c
-          - |
-            set -eu
-            deadline=$(( $(date +%s) + 1200 ))
-            until [ -s /etc/kubernetes/kubeconfig/super-admin.svc ]; do
-              if [ "$(date +%s)" -ge "$deadline" ]; then
-                echo "admin kubeconfig was not provisioned within 20m; exiting so the pod goes CrashLoopBackOff and surfaces in dashboards" >&2
-                exit 1
-              fi
-              echo "waiting for admin kubeconfig to be provisioned by Kamaji..."
-              sleep 5
-            done
-          volumeMounts:
-          - name: kubeconfig
-            mountPath: /etc/kubernetes/kubeconfig
-            readOnly: true
+      {{- include "kubernetes.waitForAdminKubeconfig" $ | nindent 6 }}
       containers:
         - name: csi-driver
           imagePullPolicy: Always
diff --git a/packages/apps/kubernetes/templates/kccm/manager.yaml b/packages/apps/kubernetes/templates/kccm/manager.yaml
index 67321647..20ac0a1e 100644
--- a/packages/apps/kubernetes/templates/kccm/manager.yaml
+++ b/packages/apps/kubernetes/templates/kccm/manager.yaml
@@ -23,26 +23,7 @@ spec:
           operator: Exists
           effect: "NoSchedule"
       initContainers:
-      - name: wait-for-kubeconfig
-        image: "{{ $.Files.Get "images/busybox.tag" | trim }}"
-        command:
-        - sh
-        - -c
-        - |
-          set -eu
-          deadline=$(( $(date +%s) + 1200 ))
-          until [ -s /etc/kubernetes/kubeconfig/super-admin.svc ]; do
-            if [ "$(date +%s)" -ge "$deadline" ]; then
-              echo "admin kubeconfig was not provisioned within 20m; exiting so the pod goes CrashLoopBackOff and surfaces in dashboards" >&2
-              exit 1
-            fi
-            echo "waiting for admin kubeconfig to be provisioned by Kamaji..."
-            sleep 5
-          done
-        volumeMounts:
-        - name: kubeconfig
-          mountPath: /etc/kubernetes/kubeconfig
-          readOnly: true
+      {{- include "kubernetes.waitForAdminKubeconfig" $ | nindent 6 }}
       containers:
       - name: kubevirt-cloud-controller-manager
         args:

From 7b146cbe56c26511d2228c17cdb7c2201f480395 Mon Sep 17 00:00:00 2001
From: Aleksei Sviridkin <f@lex.la>
Date: Thu, 16 Apr 2026 21:09:35 +0300
Subject: [PATCH 14/26] feat(api): make HelmRelease Install/Upgrade timeout
 per-Application

Replace the hardcoded r.kindName == "Kubernetes" switch in rest.go
with a config-driven path. The ApplicationDefinition CR now accepts a
release.cozystack.io/helm-install-timeout annotation that is parsed
at cozystack-api startup into config.ReleaseConfig.HelmInstallTimeout
and applied to both Install.Timeout and Upgrade.Timeout on the
rendered HelmRelease. Applications that leave the annotation unset
keep flux defaults so their failed installs remediate on the normal
cadence - only the Kubernetes kind carries the override and gets a
15m budget. New kinds with a similar race can opt in by setting the
same annotation; no rest.go patch needed.

Kubernetes-rd sets the annotation to 15m. Table-driven test in
rest_timeout_test.go covers three cases: Kubernetes with 15m, Qdrant
unset, and an arbitrary future kind with 20m - all of which pin the
Remediation.Retries == -1 contract as well.

Assisted-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
---
 .../kubernetes-rd/cozyrds/kubernetes.yaml     |   8 +
 pkg/cmd/server/start.go                       |  35 +++--
 pkg/config/config.go                          |   8 +
 pkg/registry/apps/application/rest.go         |  21 +--
 .../apps/application/rest_timeout_test.go     | 143 ++++++++++--------
 5 files changed, 133 insertions(+), 82 deletions(-)

diff --git a/packages/system/kubernetes-rd/cozyrds/kubernetes.yaml b/packages/system/kubernetes-rd/cozyrds/kubernetes.yaml
index 5e9e8f94..1128abe8 100644
--- a/packages/system/kubernetes-rd/cozyrds/kubernetes.yaml
+++ b/packages/system/kubernetes-rd/cozyrds/kubernetes.yaml
@@ -2,6 +2,14 @@ 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 applied by cozystack-api
+    # to the HelmRelease Spec.Install.Timeout and Spec.Upgrade.Timeout.
+    release.cozystack.io/helm-install-timeout: "15m"
 spec:
   application:
     kind: Kubernetes
diff --git a/pkg/cmd/server/start.go b/pkg/cmd/server/start.go
index f7791332..3e905735 100644
--- a/pkg/cmd/server/start.go
+++ b/pkg/cmd/server/start.go
@@ -160,6 +160,31 @@ 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.
+		if raw, ok := crd.Annotations["release.cozystack.io/helm-install-timeout"]; ok && raw != "" {
+			d, err := time.ParseDuration(raw)
+			if err != nil {
+				return fmt.Errorf(
+					"ApplicationDefinition %q has invalid release.cozystack.io/helm-install-timeout %q: %w",
+					crd.Name, raw, err,
+				)
+			}
+			release.HelmInstallTimeout = d
+		}
 		resource := config.Resource{
 			Application: config.ApplicationConfig{
 				Kind:          crd.Spec.Application.Kind,
@@ -168,15 +193,7 @@ 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..16cf4f0c 100644
--- a/pkg/config/config.go
+++ b/pkg/config/config.go
@@ -16,6 +16,8 @@ limitations under the License.
 
 package config
 
+import "time"
+
 // ResourceConfig represents the structure of the configuration file.
 type ResourceConfig struct {
 	Resources []Resource `yaml:"resources"`
@@ -41,6 +43,12 @@ 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/registry/apps/application/rest.go b/pkg/registry/apps/application/rest.go
index 9b92d2c0..406d3738 100644
--- a/pkg/registry/apps/application/rest.go
+++ b/pkg/registry/apps/application/rest.go
@@ -1528,16 +1528,17 @@ func (r *REST) convertApplicationToHelmRelease(app *appsv1alpha1.Application) (*
 		},
 	}
 
-	// The Kubernetes Application's parent chart creates CAPI/Kamaji resources
-	// whose admin-kubeconfig Secret is provisioned asynchronously and mounted
-	// by Deployments in the same chart. On a cold node, Kamaji control-plane
-	// bootstrap routinely exceeds flux helm-controller's default wait budget,
-	// so install remediation loops uninstall the Cluster CR and churn. Extend
-	// the wait budget to 15m for Kubernetes only - other Application kinds
-	// without this race keep flux defaults, so their failed installs do not
-	// linger unnecessarily before remediation fires.
-	if r.kindName == "Kubernetes" {
-		timeout := metav1.Duration{Duration: 15 * time.Minute}
+	// Per-Application HelmRelease wait budget. When an ApplicationDefinition
+	// sets release.cozystack.io/helm-install-timeout, the annotation is
+	// parsed at startup into ReleaseConfig.HelmInstallTimeout and applied
+	// to both Install and Upgrade here. Applications that leave it unset
+	// (the common case) keep flux defaults, so their failed installs
+	// remediate on the normal cadence. Needed for the Kubernetes kind
+	// because its parent chart contains CAPI/Kamaji resources whose
+	// admin-kubeconfig Secret is provisioned asynchronously and Kamaji
+	// cold-start routinely exceeds flux's default wait budget.
+	if r.releaseConfig.HelmInstallTimeout > 0 {
+		timeout := metav1.Duration{Duration: r.releaseConfig.HelmInstallTimeout}
 		helmRelease.Spec.Install.Timeout = &timeout
 		helmRelease.Spec.Upgrade.Timeout = &timeout
 	}
diff --git a/pkg/registry/apps/application/rest_timeout_test.go b/pkg/registry/apps/application/rest_timeout_test.go
index fa6d5427..8474d1cb 100644
--- a/pkg/registry/apps/application/rest_timeout_test.go
+++ b/pkg/registry/apps/application/rest_timeout_test.go
@@ -10,7 +10,7 @@ import (
 	"github.com/cozystack/cozystack/pkg/config"
 )
 
-func newRESTForKind(kind, prefix string) *REST {
+func newRESTForTimeout(kind, prefix string, helmInstallTimeout time.Duration) *REST {
 	return &REST{
 		kindName: kind,
 		releaseConfig: config.ReleaseConfig{
@@ -20,76 +20,93 @@ func newRESTForKind(kind, prefix string) *REST {
 				Name:      "x",
 				Namespace: "cozy-system",
 			},
+			HelmInstallTimeout: helmInstallTimeout,
 		},
 	}
 }
 
-func TestConvertApplicationToHelmRelease_KubernetesKindGetsLongTimeout(t *testing.T) {
-	r := newRESTForKind("Kubernetes", "kubernetes-")
-	app := &appsv1alpha1.Application{
-		ObjectMeta: metav1.ObjectMeta{Name: "example", Namespace: "tenant-root"},
+// 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,
+		},
+		{
+			name:       "Qdrant kind without configured timeout keeps flux defaults",
+			kind:       "Qdrant",
+			prefix:     "qdrant-",
+			configured: 0,
+			wantSet:    false,
+		},
+		{
+			name:       "arbitrary future kind with 20m configured gets 20m",
+			kind:       "TalosCluster",
+			prefix:     "talos-",
+			configured: 20 * time.Minute,
+			wantSet:    true,
+		},
 	}
 
-	hr, err := r.convertApplicationToHelmRelease(app)
-	if err != nil {
-		t.Fatalf("convertApplicationToHelmRelease returned error: %v", err)
-	}
+	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"},
+			}
 
-	if hr.Spec.Install == nil {
-		t.Fatal("Spec.Install must not be nil")
-	}
-	if hr.Spec.Install.Timeout == nil {
-		t.Fatal("Spec.Install.Timeout must be set for Kubernetes kind")
-	}
-	if hr.Spec.Install.Timeout.Duration < 15*time.Minute {
-		t.Errorf("Spec.Install.Timeout must be >= 15m for Kubernetes, got %v", hr.Spec.Install.Timeout.Duration)
-	}
+			hr, err := r.convertApplicationToHelmRelease(app)
+			if err != nil {
+				t.Fatalf("convertApplicationToHelmRelease returned error: %v", err)
+			}
 
-	if hr.Spec.Upgrade == nil {
-		t.Fatal("Spec.Upgrade must not be nil")
-	}
-	if hr.Spec.Upgrade.Timeout == nil {
-		t.Fatal("Spec.Upgrade.Timeout must be set for Kubernetes kind")
-	}
-	if hr.Spec.Upgrade.Timeout.Duration < 15*time.Minute {
-		t.Errorf("Spec.Upgrade.Timeout must be >= 15m for Kubernetes, got %v", hr.Spec.Upgrade.Timeout.Duration)
-	}
+			if hr.Spec.Install == nil || hr.Spec.Upgrade == nil {
+				t.Fatalf("Spec.Install/Upgrade must be non-nil")
+			}
 
-	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)
-	}
-}
-
-func TestConvertApplicationToHelmRelease_NonKubernetesKindKeepsFluxDefaults(t *testing.T) {
-	// For Applications whose parent chart has no admin-kubeconfig race
-	// (Qdrant, MongoDB, Postgres, etc.), do NOT extend the helm-wait
-	// budget - otherwise failed installs would block three times as long
-	// before Flux starts remediating.
-	r := newRESTForKind("Qdrant", "qdrant-")
-	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.Install.Timeout != nil {
-		t.Errorf("Spec.Install.Timeout must be unset for non-Kubernetes kinds, got %v", hr.Spec.Install.Timeout.Duration)
-	}
-	if hr.Spec.Upgrade != nil && hr.Spec.Upgrade.Timeout != nil {
-		t.Errorf("Spec.Upgrade.Timeout must be unset for non-Kubernetes kinds, got %v", hr.Spec.Upgrade.Timeout.Duration)
-	}
-
-	// But remediation must still be -1 across the board.
-	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)
+			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)
+			}
+		})
 	}
 }

From bc5473d4fc9bb1b43be79ea5eb8d370d9e6c00b7 Mon Sep 17 00:00:00 2001
From: Aleksei Sviridkin <f@lex.la>
Date: Thu, 16 Apr 2026 21:11:38 +0300
Subject: [PATCH 15/26] test(kubernetes): chart-wide invariant for
 admin-kubeconfig guards

The per-template unittests in packages/apps/kubernetes/tests/ assert
that cluster-autoscaler, kccm, and the csi controller each mount the
admin-kubeconfig Secret optional and carry the wait-for-kubeconfig
init. That locks in today's three Deployments by name - a fourth
Deployment that mounts the same Secret but forgets the guard would
slip past them.

Add a bats-unit test that renders the entire chart, enumerates every
Deployment whose spec mounts a Secret ending in -admin-kubeconfig, and
asserts optional:true plus wait-for-kubeconfig init on all of them.
Verified by temporarily removing optional:true from csi/deploy.yaml:
the test correctly flagged invariant-kcsi-controller as an offender.

Assisted-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
---
 hack/admin-kubeconfig-invariant.bats | 72 ++++++++++++++++++++++++++++
 1 file changed, 72 insertions(+)
 create mode 100644 hack/admin-kubeconfig-invariant.bats

diff --git a/hack/admin-kubeconfig-invariant.bats b/hack/admin-kubeconfig-invariant.bats
new file mode 100644
index 00000000..d699de83
--- /dev/null
+++ b/hack/admin-kubeconfig-invariant.bats
@@ -0,0 +1,72 @@
+#!/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)"
+}

From 03426fbd718029e40d6680f299189e97f915a614 Mon Sep 17 00:00:00 2001
From: Aleksei Sviridkin <f@lex.la>
Date: Thu, 16 Apr 2026 21:13:06 +0300
Subject: [PATCH 16/26] test(hack): pin HelmRelease v2 status shape used by
 remediation guard

run-kubernetes.sh extracts .status.installFailures and
.status.upgradeFailures via kubectl -o jsonpath. If a future flux
release renames the counters, kubectl returns empty, the guard reports
no cycle, and e2e silently misses real remediation loops.

Add a bats unit test that feeds a pinned HelmRelease v2 status snippet
through the same jsonpath and asserts the extraction still yields the
expected values. Also leave a pointer comment in run-kubernetes.sh so a
future flux bump surfaces the version coupling in review.

Assisted-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
---
 hack/e2e-apps/run-kubernetes.sh |  7 ++++++
 hack/remediation-guard.bats     | 44 +++++++++++++++++++++++++++++++++
 2 files changed, 51 insertions(+)

diff --git a/hack/e2e-apps/run-kubernetes.sh b/hack/e2e-apps/run-kubernetes.sh
index 811f42a7..e6002b53 100644
--- a/hack/e2e-apps/run-kubernetes.sh
+++ b/hack/e2e-apps/run-kubernetes.sh
@@ -326,6 +326,13 @@ EOF
   # 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 status shape: .status.installFailures and
+  # .status.upgradeFailures are counters populated by the controller on
+  # every failed install/upgrade. If a future flux release renames them,
+  # kubectl returns the empty string and the guard silently passes. The
+  # shape is pinned by hack/remediation-guard.bats (see that file for
+  # details), and the vendored API types live under
+  # vendor/github.com/fluxcd/helm-controller/api/v2.
   install_failures=$(kubectl get hr -n tenant-test "kubernetes-${test_name}" -ojsonpath='{.status.installFailures}')
   upgrade_failures=$(kubectl get hr -n tenant-test "kubernetes-${test_name}" -ojsonpath='{.status.upgradeFailures}')
   if helmrelease_has_remediation_cycle "${install_failures}" "${upgrade_failures}"; then
diff --git a/hack/remediation-guard.bats b/hack/remediation-guard.bats
index 64aa83af..9c33452a 100644
--- a/hack/remediation-guard.bats
+++ b/hack/remediation-guard.bats
@@ -85,3 +85,47 @@
     helmrelease_has_remediation_cycle "3" "5" || rc=$?
     [ "$rc" -eq 0 ]
 }
+
+@test "installFailures and upgradeFailures extraction pins HR v2 status shape" {
+    # Pins the Flux HelmRelease v2 status shape that run-kubernetes.sh relies
+    # on. If a future Flux version renames .status.installFailures (or
+    # .status.upgradeFailures), kubectl get -o jsonpath returns an empty
+    # string, the guard quietly says "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. yq
+    # evaluates the same json-ish jsonpath against a pinned HR snippet, so
+    # the test fails loudly if the field ever disappears or moves. Cross
+    # reference: vendor/github.com/fluxcd/helm-controller/api/v2/ status
+    # struct field tags.
+    tmp=$(mktemp -d)
+    trap 'rm -rf "$tmp"' EXIT
+
+    cat > "$tmp/hr.yaml" <<'YAML'
+apiVersion: helm.toolkit.fluxcd.io/v2
+kind: HelmRelease
+metadata:
+  name: kubernetes-test
+  namespace: tenant-test
+spec:
+  interval: 5m
+status:
+  installFailures: 2
+  upgradeFailures: 0
+  conditions:
+    - type: Ready
+      status: "False"
+      reason: UninstallSucceeded
+YAML
+
+    install_failures=$(yq '.status.installFailures' "$tmp/hr.yaml")
+    upgrade_failures=$(yq '.status.upgradeFailures' "$tmp/hr.yaml")
+
+    [ "$install_failures" = "2" ]
+    [ "$upgrade_failures" = "0" ]
+
+    . hack/e2e-apps/remediation-guard.sh
+    rc=0
+    helmrelease_has_remediation_cycle "$install_failures" "$upgrade_failures" || rc=$?
+    [ "$rc" -eq 0 ]
+}

From 12632c60c76b77639c8e2edd7ab5bdb248398a7e Mon Sep 17 00:00:00 2001
From: Aleksei Sviridkin <f@lex.la>
Date: Thu, 16 Apr 2026 21:25:50 +0300
Subject: [PATCH 17/26] fix(kubernetes): gate CP-side Deployments on tenant
 etcd DataStore

The soft-skip wrap in cluster.yaml only silenced Cluster and
KamajiControlPlane rendering when _namespace.etcd is empty. The three
CP-side Deployments (cluster-autoscaler, kccm, kcsi-controller) still
rendered, their wait-for-kubeconfig init containers CrashLoopBackOff'd
forever (no KamajiControlPlane = no admin-kubeconfig Secret),
HelmRelease hit its 15m wait timeout and triggered the very install
remediation cycle the rest of this PR prevents. Self-contradiction.

Wrap each of the three Deployment templates in
{{- if .Values._namespace.etcd }}...{{- end }} so they render only
when there is a DataStore to back them. Add an invariant bats test
that renders the whole chart with etcd empty and asserts zero
Deployments reference *-admin-kubeconfig.

Assisted-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
---
 hack/admin-kubeconfig-invariant.bats          | 45 +++++++++++++++++++
 .../cluster-autoscaler/deployment.yaml        | 12 +++++
 .../apps/kubernetes/templates/csi/deploy.yaml |  2 +
 .../kubernetes/templates/kccm/manager.yaml    |  2 +
 4 files changed, 61 insertions(+)

diff --git a/hack/admin-kubeconfig-invariant.bats b/hack/admin-kubeconfig-invariant.bats
index d699de83..c027e2c4 100644
--- a/hack/admin-kubeconfig-invariant.bats
+++ b/hack/admin-kubeconfig-invariant.bats
@@ -70,3 +70,48 @@
 
   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 "$tmp"' EXIT
+
+  helm template invariant packages/apps/kubernetes \
+    --namespace tenant-root \
+    --set _namespace.etcd="" \
+    --set _namespace.monitoring="" \
+    --set _namespace.ingress="" \
+    --set _namespace.seaweedfs="" \
+    --set _namespace.host="" \
+    --set _cluster.cluster-domain=cozy.local \
+    --set 'nodeGroups=null' \
+    2>/dev/null > "$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)"
+}
diff --git a/packages/apps/kubernetes/templates/cluster-autoscaler/deployment.yaml b/packages/apps/kubernetes/templates/cluster-autoscaler/deployment.yaml
index 348b017d..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
@@ -108,3 +119,4 @@ rules:
     - list
     - update
     - watch
+{{- end }}
diff --git a/packages/apps/kubernetes/templates/csi/deploy.yaml b/packages/apps/kubernetes/templates/csi/deploy.yaml
index c1af7f13..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:
@@ -238,3 +239,4 @@ spec:
             secretName: {{ .Release.Name }}-admin-kubeconfig
             optional: true
           name: kubeconfig
+{{- end }}
diff --git a/packages/apps/kubernetes/templates/kccm/manager.yaml b/packages/apps/kubernetes/templates/kccm/manager.yaml
index 20ac0a1e..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:
@@ -60,3 +61,4 @@ spec:
           optional: true
         name: kubeconfig
       serviceAccountName: {{ .Release.Name }}-kccm
+{{- end }}

From 1ddd2aaea97024f6803c60d29a946273b55e4185 Mon Sep 17 00:00:00 2001
From: Aleksei Sviridkin <f@lex.la>
Date: Thu, 16 Apr 2026 21:28:11 +0300
Subject: [PATCH 18/26] fix(hack): detect remediation via status.history, not
 transient counters

Flux helm-controller's ClearFailures() zeroes installFailures and
upgradeFailures on every successful reconciliation (see the upstream
HelmReleaseStatus method). The previous guard ran after the HelmRelease
was Ready, at which point the counters were always 0 - the assertion
was vacuous and would have passed against a reverted fix.

Switch to .status.history, which retains per-revision release
Snapshots that survive a subsequent successful reconciliation. A
remediation cycle leaves behind a Snapshot with status=uninstalled
(the install-remediation code path) or status=failed (Helm release
failure that remediation then uninstalled). Either one signals the
race actually fired.

Rewrite the bats unit tests to cover: empty history, deployed-only,
deployed+superseded (happy path - not detected), single failed,
single uninstalled, uninstalled-then-deployed, and deployed-then-failed
(all detected). The pinned-shape test feeds a realistic HR status
snippet through yq the same way run-kubernetes.sh does via kubectl
-o jsonpath.

Assisted-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
---
 hack/e2e-apps/remediation-guard.sh |  45 +++++---
 hack/e2e-apps/run-kubernetes.sh    |  24 +++--
 hack/remediation-guard.bats        | 158 ++++++++++++++---------------
 3 files changed, 120 insertions(+), 107 deletions(-)

diff --git a/hack/e2e-apps/remediation-guard.sh b/hack/e2e-apps/remediation-guard.sh
index 6da69f13..38463310 100644
--- a/hack/e2e-apps/remediation-guard.sh
+++ b/hack/e2e-apps/remediation-guard.sh
@@ -1,24 +1,39 @@
 # Helpers for asserting that a Flux HelmRelease did not fall into an
 # install/upgrade remediation cycle during an e2e run.
 #
-# A non-zero installFailures/upgradeFailures counter means flux
-# helm-controller hit its wait timeout, ran remediation (uninstall),
-# and re-installed. That is exactly the race this guard is meant to
-# catch, so the function returns success (0) when a cycle is detected
-# and failure (1) otherwise.
+# 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.
 #
-# Both arguments may be empty strings, the literal "0", or a positive
-# integer. Shell's && and || have equal precedence with left-to-right
-# associativity, so each half of the disjunction is grouped explicitly
-# to avoid (A && B) || C && D parsing that masks the common
-# install_failures=1, upgrade_failures="" case.
+# 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() {
-    install_failures="$1"
-    upgrade_failures="$2"
-    if { [ -n "${install_failures}" ] && [ "${install_failures}" != "0" ]; } || \
-       { [ -n "${upgrade_failures}" ] && [ "${upgrade_failures}" != "0" ]; }; then
-        return 0
+    statuses="$1"
+    if [ -z "${statuses}" ]; then
+        return 1
     fi
+    while IFS= read -r status; do
+        case "${status}" in
+            failed|uninstalled)
+                return 0
+                ;;
+        esac
+    done <<EOF
+${statuses}
+EOF
     return 1
 }
diff --git a/hack/e2e-apps/run-kubernetes.sh b/hack/e2e-apps/run-kubernetes.sh
index e6002b53..64b1ae90 100644
--- a/hack/e2e-apps/run-kubernetes.sh
+++ b/hack/e2e-apps/run-kubernetes.sh
@@ -326,17 +326,19 @@ EOF
   # 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 status shape: .status.installFailures and
-  # .status.upgradeFailures are counters populated by the controller on
-  # every failed install/upgrade. If a future flux release renames them,
-  # kubectl returns the empty string and the guard silently passes. The
-  # shape is pinned by hack/remediation-guard.bats (see that file for
-  # details), and the vendored API types live under
-  # vendor/github.com/fluxcd/helm-controller/api/v2.
-  install_failures=$(kubectl get hr -n tenant-test "kubernetes-${test_name}" -ojsonpath='{.status.installFailures}')
-  upgrade_failures=$(kubectl get hr -n tenant-test "kubernetes-${test_name}" -ojsonpath='{.status.upgradeFailures}')
-  if helmrelease_has_remediation_cycle "${install_failures}" "${upgrade_failures}"; then
-    echo "Parent HelmRelease entered remediation cycle: installFailures=${install_failures:-0}, upgradeFailures=${upgrade_failures:-0}" >&2
+  # 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}')
+  if helmrelease_has_remediation_cycle "${history_statuses}"; then
+    echo "Parent HelmRelease entered remediation cycle. History statuses:" >&2
+    printf '%s\n' "${history_statuses}" >&2
     kubectl -n tenant-test describe hr "kubernetes-${test_name}" >&2
     exit 1
   fi
diff --git a/hack/remediation-guard.bats b/hack/remediation-guard.bats
index 9c33452a..30e4e04c 100644
--- a/hack/remediation-guard.bats
+++ b/hack/remediation-guard.bats
@@ -2,17 +2,15 @@
 # -----------------------------------------------------------------------------
 # Unit tests for hack/e2e-apps/remediation-guard.sh
 #
-# helmrelease_has_remediation_cycle is consumed from e2e tests to assert that
-# the parent HelmRelease did not hit flux helm-controller's wait timeout and
-# enter uninstall remediation. The function accepts two arguments (values of
-# .status.installFailures and .status.upgradeFailures) and returns 0 when a
-# remediation cycle is detected, 1 otherwise.
+# helmrelease_has_remediation_cycle takes a newline-delimited list of
+# HelmRelease history snapshot status values (deployed/superseded/failed/
+# uninstalled/...) and returns 0 when any entry is "failed" or "uninstalled"
+# (meaning flux helm-controller performed install/upgrade remediation).
 #
-# Each argument can be empty (controller never populated the field), "0"
-# (populated but never failed), or a positive integer. Shell's && and ||
-# have equal precedence with left-to-right associativity, which used to
-# break this check on the most common failure mode - install_failures=1
-# and upgrade_failures="". These tests pin the correct behavior.
+# This is used by the e2e script after the HelmRelease reaches Ready. The
+# failure/upgrade counters (.status.installFailures / .status.upgradeFailures)
+# are useless there because flux's ClearFailures zeroes them on successful
+# reconciliation; .status.history retains the snapshot trail.
 #
 # cozytest.sh's awk parser recognizes only @test blocks and a bare `}` on
 # its own line; there is no bats `run` or `$status`. Assertions are
@@ -21,83 +19,76 @@
 # Run with: hack/cozytest.sh hack/remediation-guard.bats
 # -----------------------------------------------------------------------------
 
-@test "no counters set returns not-detected" {
+@test "empty history returns not-detected" {
     . hack/e2e-apps/remediation-guard.sh
-    rc=0
-    helmrelease_has_remediation_cycle "" "" || rc=$?
-    [ "$rc" -eq 1 ]
+    if helmrelease_has_remediation_cycle ""; then
+        echo "expected not-detected for empty history" >&2
+        exit 1
+    fi
 }
 
-@test "both counters zero returns not-detected" {
+@test "single deployed snapshot returns not-detected" {
     . hack/e2e-apps/remediation-guard.sh
-    rc=0
-    helmrelease_has_remediation_cycle "0" "0" || rc=$?
-    [ "$rc" -eq 1 ]
+    if helmrelease_has_remediation_cycle "deployed"; then
+        echo "expected not-detected for deployed-only history" >&2
+        exit 1
+    fi
 }
 
-@test "install zero upgrade empty returns not-detected" {
+@test "deployed then superseded returns not-detected" {
     . hack/e2e-apps/remediation-guard.sh
-    rc=0
-    helmrelease_has_remediation_cycle "0" "" || rc=$?
-    [ "$rc" -eq 1 ]
+    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 "install empty upgrade zero returns not-detected" {
+@test "single failed snapshot returns detected" {
     . hack/e2e-apps/remediation-guard.sh
-    rc=0
-    helmrelease_has_remediation_cycle "" "0" || rc=$?
-    [ "$rc" -eq 1 ]
+    if ! helmrelease_has_remediation_cycle "failed"; then
+        echo "expected detected when history contains failed snapshot" >&2
+        exit 1
+    fi
 }
 
-@test "install one upgrade empty returns detected" {
-    # Canonical race: first install exceeded helm-wait, remediation fired,
-    # no upgrade has happened yet.
+@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
-    rc=0
-    helmrelease_has_remediation_cycle "1" "" || rc=$?
-    [ "$rc" -eq 0 ]
+    if ! helmrelease_has_remediation_cycle "uninstalled"; then
+        echo "expected detected when history contains uninstalled snapshot" >&2
+        exit 1
+    fi
 }
 
-@test "install empty upgrade one returns detected" {
+@test "uninstalled then deployed still returns detected" {
     . hack/e2e-apps/remediation-guard.sh
-    rc=0
-    helmrelease_has_remediation_cycle "" "1" || rc=$?
-    [ "$rc" -eq 0 ]
+    statuses=$(printf 'uninstalled\ndeployed\n')
+    if ! helmrelease_has_remediation_cycle "${statuses}"; then
+        echo "expected detected despite later successful deploy" >&2
+        exit 1
+    fi
 }
 
-@test "install two upgrade zero returns detected" {
+@test "deployed then failed still returns detected" {
     . hack/e2e-apps/remediation-guard.sh
-    rc=0
-    helmrelease_has_remediation_cycle "2" "0" || rc=$?
-    [ "$rc" -eq 0 ]
+    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 "install zero upgrade two returns detected" {
-    . hack/e2e-apps/remediation-guard.sh
-    rc=0
-    helmrelease_has_remediation_cycle "0" "2" || rc=$?
-    [ "$rc" -eq 0 ]
-}
-
-@test "both counters positive returns detected" {
-    . hack/e2e-apps/remediation-guard.sh
-    rc=0
-    helmrelease_has_remediation_cycle "3" "5" || rc=$?
-    [ "$rc" -eq 0 ]
-}
-
-@test "installFailures and upgradeFailures extraction pins HR v2 status shape" {
-    # Pins the Flux HelmRelease v2 status shape that run-kubernetes.sh relies
-    # on. If a future Flux version renames .status.installFailures (or
-    # .status.upgradeFailures), kubectl get -o jsonpath returns an empty
-    # string, the guard quietly says "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. yq
-    # evaluates the same json-ish jsonpath against a pinned HR snippet, so
-    # the test fails loudly if the field ever disappears or moves. Cross
-    # reference: vendor/github.com/fluxcd/helm-controller/api/v2/ status
-    # struct field tags.
+@test "installFailures 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
 
@@ -107,25 +98,30 @@ kind: HelmRelease
 metadata:
   name: kubernetes-test
   namespace: tenant-test
-spec:
-  interval: 5m
 status:
-  installFailures: 2
-  upgradeFailures: 0
-  conditions:
-    - type: Ready
-      status: "False"
-      reason: UninstallSucceeded
+  history:
+    - name: kubernetes-test
+      namespace: tenant-test
+      version: 1
+      status: uninstalled
+    - name: kubernetes-test
+      namespace: tenant-test
+      version: 2
+      status: deployed
 YAML
 
-    install_failures=$(yq '.status.installFailures' "$tmp/hr.yaml")
-    upgrade_failures=$(yq '.status.upgradeFailures' "$tmp/hr.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")
 
-    [ "$install_failures" = "2" ]
-    [ "$upgrade_failures" = "0" ]
+    [ -n "$statuses" ]
+    echo "$statuses" | grep --quiet '^uninstalled$'
 
     . hack/e2e-apps/remediation-guard.sh
-    rc=0
-    helmrelease_has_remediation_cycle "$install_failures" "$upgrade_failures" || rc=$?
-    [ "$rc" -eq 0 ]
+    if ! helmrelease_has_remediation_cycle "$statuses"; then
+        echo "expected detected for pinned HR snippet with uninstalled + deployed history" >&2
+        exit 1
+    fi
 }

From 48312cc3696904235e87fcf2bc33d05e21a77b72 Mon Sep 17 00:00:00 2001
From: Aleksei Sviridkin <f@lex.la>
Date: Thu, 16 Apr 2026 21:29:49 +0300
Subject: [PATCH 19/26] build: wire go-unit-tests into make unit-tests

CI runs make unit-tests on every PR, which already covers helm
unittests and bats (hack/admin-kubeconfig-invariant.bats and
hack/remediation-guard.bats are both picked up by the existing
hack/*.bats glob). What was missing was any go test invocation.

Add a go-unit-tests target scoped to pkg/registry, pkg/config, and
pkg/cmd/server - the cozystack-api surface this repo actually owns
and tests in-tree. Running go test ./... pulls in generated-code
round-trip suites whose behavior is governed by generator tool
versions outside this repo's control; those are better exercised
from their own generator workflows.

Assisted-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
---
 Makefile | 10 +++++++++-
 1 file changed, 9 insertions(+), 1 deletion(-)

diff --git a/Makefile b/Makefile
index 0e94c5e6..10a55314 100644
--- a/Makefile
+++ b/Makefile
@@ -82,11 +82,19 @@ test:
 	make -C packages/core/testing apply
 	make -C packages/core/testing test
 
-unit-tests: helm-unit-tests bats-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.

From d2e8f7e86cc7157aa83a7130b3f9857f10518218 Mon Sep 17 00:00:00 2001
From: Aleksei Sviridkin <f@lex.la>
Date: Thu, 16 Apr 2026 21:31:29 +0300
Subject: [PATCH 20/26] chore(kubernetes): drop busybox mirror Containerfile,
 pin upstream directly

The image-busybox Makefile target + images/busybox/Dockerfile wrapper
just rebuilt an upstream busybox digest as ghcr.io/cozystack/cozystack/
busybox. Payload is a one-shot sh loop run once per pod; the pinned
upstream digest is already immutable, so maintaining a private mirror
adds churn (rebuild on every release) for no real hardening benefit.
Delete the wrapper and reference docker.io/library/busybox:<digest>
directly from images/busybox.tag. Document the choice in _helpers.tpl.

Also drop the false coupling in the go table test: the "unrelated
kind without configured timeout" case used the real kind name Qdrant,
which tied the test to the Qdrant ApplicationDefinition for no reason.
Switch to a clearly fictional kind so a future Qdrant timeout override
does not break this assertion.

Assisted-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
---
 packages/apps/kubernetes/Makefile                  | 13 +------------
 packages/apps/kubernetes/images/busybox/Dockerfile |  1 -
 packages/apps/kubernetes/templates/_helpers.tpl    |  9 ++++++++-
 pkg/registry/apps/application/rest_timeout_test.go | 10 +++++++---
 4 files changed, 16 insertions(+), 17 deletions(-)
 delete mode 100644 packages/apps/kubernetes/images/busybox/Dockerfile

diff --git a/packages/apps/kubernetes/Makefile b/packages/apps/kubernetes/Makefile
index eb23c2ef..4fea42ef 100644
--- a/packages/apps/kubernetes/Makefile
+++ b/packages/apps/kubernetes/Makefile
@@ -15,7 +15,7 @@ update:
 	hack/update-versions.sh
 	make generate
 
-image: image-ubuntu-container-disk image-kubevirt-cloud-provider image-kubevirt-csi-driver image-cluster-autoscaler image-busybox
+image: image-ubuntu-container-disk image-kubevirt-cloud-provider image-kubevirt-csi-driver image-cluster-autoscaler
 
 image-ubuntu-container-disk:
 	$(foreach ver,$(KUBERNETES_VERSIONS), \
@@ -71,14 +71,3 @@ image-cluster-autoscaler:
 		> images/cluster-autoscaler.tag
 	rm -f images/cluster-autoscaler.json
 
-image-busybox:
-	docker buildx build images/busybox \
-		--tag $(REGISTRY)/busybox:$(call settag,$(KUBERNETES_PKG_TAG)) \
-		--tag $(REGISTRY)/busybox:$(call settag,$(KUBERNETES_PKG_TAG)-$(TAG)) \
-		--cache-from type=registry,ref=$(REGISTRY)/busybox:latest \
-		--cache-to type=inline \
-		--metadata-file images/busybox.json \
-		$(BUILDX_ARGS)
-	echo "$(REGISTRY)/busybox:$(call settag,$(KUBERNETES_PKG_TAG))@$$(yq e '."containerimage.digest"' images/busybox.json -o json -r)" \
-		> images/busybox.tag
-	rm -f images/busybox.json
diff --git a/packages/apps/kubernetes/images/busybox/Dockerfile b/packages/apps/kubernetes/images/busybox/Dockerfile
deleted file mode 100644
index 0c57fd39..00000000
--- a/packages/apps/kubernetes/images/busybox/Dockerfile
+++ /dev/null
@@ -1 +0,0 @@
-FROM docker.io/library/busybox:1.37.0@sha256:1487d0af5f52b4ba31c7e465126ee2123fe3f2305d638e7827681e7cf6c83d5e
diff --git a/packages/apps/kubernetes/templates/_helpers.tpl b/packages/apps/kubernetes/templates/_helpers.tpl
index 89f06934..e6ad9dcf 100644
--- a/packages/apps/kubernetes/templates/_helpers.tpl
+++ b/packages/apps/kubernetes/templates/_helpers.tpl
@@ -60,10 +60,17 @@ 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 scoped to the Kubernetes Application kind so the
+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 pinned busybox image in images/busybox.tag 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, the digest pin makes the
+pull immutable, and the cost of maintaining a private mirror of a tiny
+upstream image that does not move often is not worth it.
+
 Call site owns the surrounding volumes block; the kubeconfig volume
 must exist on the pod and mount at /etc/kubernetes/kubeconfig.
 */}}
diff --git a/pkg/registry/apps/application/rest_timeout_test.go b/pkg/registry/apps/application/rest_timeout_test.go
index 8474d1cb..85ac0b97 100644
--- a/pkg/registry/apps/application/rest_timeout_test.go
+++ b/pkg/registry/apps/application/rest_timeout_test.go
@@ -48,9 +48,13 @@ func TestConvertApplicationToHelmRelease_AppliesReleaseConfigTimeout(t *testing.
 			wantSet:    true,
 		},
 		{
-			name:       "Qdrant kind without configured timeout keeps flux defaults",
-			kind:       "Qdrant",
-			prefix:     "qdrant-",
+			// 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,
 		},

From 64b216e10a2392394add9eb7fbf7e908f4c8864a Mon Sep 17 00:00:00 2001
From: Aleksei Sviridkin <f@lex.la>
Date: Thu, 16 Apr 2026 21:46:01 +0300
Subject: [PATCH 21/26] fix(kubernetes): gate child HelmReleases on tenant etcd
 DataStore

17 child HelmReleases (cilium, coredns, csi, cert-manager,
metrics-server, ...) referenced *-admin-kubeconfig via
kubeConfig.secretRef and rendered even when _namespace.etcd was empty.
On an etcd-less tenant each one sat in NotReady forever because the
admin-kubeconfig Secret only exists after a KamajiControlPlane
reconciles, and KamajiControlPlane now only renders when etcd is set.
The outcome contradicted the "beacon only" contract claimed in the
soft-skip commit.

Extend the existing addon guards to also require _namespace.etcd, and
wrap the four unconditional HelmReleases (csi, metrics-server,
prometheus-operator-crds, volumesnapshot-crd) plus the always-on
cilium/coredns HR resources in the same gate. Add an invariant bats
test that renders the whole chart with etcd empty and asserts zero
HelmReleases reference *-admin-kubeconfig.

Assisted-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
---
 hack/admin-kubeconfig-invariant.bats          | 40 +++++++++++++++++++
 .../helmreleases/cert-manager-crds.yaml       |  2 +-
 .../templates/helmreleases/cert-manager.yaml  |  2 +-
 .../templates/helmreleases/cilium.yaml        |  2 +
 .../templates/helmreleases/coredns.yaml       |  2 +
 .../templates/helmreleases/csi.yaml           |  2 +
 .../templates/helmreleases/fluxcd.yaml        |  2 +-
 .../helmreleases/gateway-api-crds.yaml        |  2 +-
 .../templates/helmreleases/gpu-operator.yaml  |  2 +-
 .../templates/helmreleases/ingress-nginx.yaml |  2 +-
 .../helmreleases/metrics-server.yaml          |  2 +
 .../helmreleases/monitoring-agents.yaml       |  2 +-
 .../prometheus-operator-crds.yaml             |  2 +
 .../templates/helmreleases/velero.yaml        |  2 +-
 .../vertical-pod-autoscaler-crds.yaml         |  2 +-
 .../helmreleases/vertical-pod-autoscaler.yaml |  2 +-
 .../victoria-metrics-operator.yaml            |  2 +-
 .../helmreleases/volumesnapshot-crd.yaml      |  2 +
 18 files changed, 63 insertions(+), 11 deletions(-)

diff --git a/hack/admin-kubeconfig-invariant.bats b/hack/admin-kubeconfig-invariant.bats
index c027e2c4..4c809c41 100644
--- a/hack/admin-kubeconfig-invariant.bats
+++ b/hack/admin-kubeconfig-invariant.bats
@@ -115,3 +115,43 @@
 
   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 \
+    --set _namespace.etcd="" \
+    --set _namespace.monitoring="" \
+    --set _namespace.ingress="" \
+    --set _namespace.seaweedfs="" \
+    --set _namespace.host="" \
+    --set _cluster.cluster-domain=cozy.local \
+    --set 'nodeGroups=null' \
+    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/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..e243bfad 100644
--- a/packages/apps/kubernetes/templates/helmreleases/gpu-operator.yaml
+++ b/packages/apps/kubernetes/templates/helmreleases/gpu-operator.yaml
@@ -1,4 +1,4 @@
-{{- if .Values.addons.gpuOperator.enabled }}
+{{- if and .Values.addons.gpuOperator.enabled .Values._namespace.etcd }}
 apiVersion: helm.toolkit.fluxcd.io/v2
 kind: HelmRelease
 metadata:
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 }}

From 6072723e1e03c0072fe3bd83e2a5723c50d6ebd8 Mon Sep 17 00:00:00 2001
From: Aleksei Sviridkin <f@lex.la>
Date: Thu, 16 Apr 2026 21:47:19 +0300
Subject: [PATCH 22/26] feat(config): extract + test annotation-timeout parser
 with Flux unit gate
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Pulls the release.cozystack.io/helm-install-timeout parsing out of
start.go into ParseHelmInstallTimeoutAnnotation in pkg/config. The
helper rejects units that time.ParseDuration accepts but Flux
helm-controller rejects (ns, us, µs): feeding one of those through
would cause the HelmRelease admission webhook to reject the object at
install time, giving a silent drop to flux defaults that is hard to
debug. Fail loudly at cozystack-api startup instead.

Adds a table-driven unit test covering: unset (empty), accepted units
ms/s/m/h, compound 2h30m, decimal 1.5m, and the rejected shapes
(bare digits, garbage, negative, ns/us/µs). The test lives in
pkg/config so it runs under the existing go-unit-tests make target.

Assisted-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
---
 pkg/cmd/server/start.go   |  23 ++++----
 pkg/config/config.go      |  42 +++++++++++++-
 pkg/config/config_test.go | 119 ++++++++++++++++++++++++++++++++++++++
 3 files changed, 173 insertions(+), 11 deletions(-)
 create mode 100644 pkg/config/config_test.go

diff --git a/pkg/cmd/server/start.go b/pkg/cmd/server/start.go
index 3e905735..7ebbbc89 100644
--- a/pkg/cmd/server/start.go
+++ b/pkg/cmd/server/start.go
@@ -174,17 +174,20 @@ func (o *CozyServerOptions) Complete() error {
 		// 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.
-		if raw, ok := crd.Annotations["release.cozystack.io/helm-install-timeout"]; ok && raw != "" {
-			d, err := time.ParseDuration(raw)
-			if err != nil {
-				return fmt.Errorf(
-					"ApplicationDefinition %q has invalid release.cozystack.io/helm-install-timeout %q: %w",
-					crd.Name, raw, err,
-				)
-			}
-			release.HelmInstallTimeout = d
+		// 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,
diff --git a/pkg/config/config.go b/pkg/config/config.go
index 16cf4f0c..21de27b2 100644
--- a/pkg/config/config.go
+++ b/pkg/config/config.go
@@ -16,7 +16,47 @@ limitations under the License.
 
 package config
 
-import "time"
+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 {
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)
+			}
+		})
+	}
+}

From b8aec9a9731885cae9217271381680f3e274f46f Mon Sep 17 00:00:00 2001
From: Aleksei Sviridkin <f@lex.la>
Date: Thu, 16 Apr 2026 21:50:10 +0300
Subject: [PATCH 23/26] fix(kubernetes): history guard non-empty check + nits
 from review

- Log .status.history regardless of content so a silently empty result
  (Flux field rename) shows up in CI logs, and treat empty history on
  a Ready HelmRelease as a distinct failure. A Ready HR by definition
  has at least one snapshot; anything else is a shape-drift signal.
- Replace the unquoted heredoc in remediation-guard.sh with a printf |
  grep pipeline. printf %s treats statuses as literal payload (no $
  expansion surprises for future callers), grep --quiet --extended-regexp
  returns exit status the caller can forward directly.
- Share the etcd-absent values file between both invariant tests
  (packages/apps/kubernetes/tests/values-ci-no-etcd.yaml) instead of
  duplicating the --set block.
- Fix typo "override applied" -> "override is applied" in the
  Kubernetes ApplicationDefinition.
- Add a coupling comment in the ApplicationDefinition annotation that
  points at the wait-for-kubeconfig init deadline in _helpers.tpl, so
  a future operator raising the HR timeout updates the init deadline
  too.
- Clarify the per-annotation timeout comment in rest.go so it stops
  implying the feature is Kubernetes-only (it is not - only today's
  one user is).

Assisted-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
---
 hack/admin-kubeconfig-invariant.bats          | 16 ++------------
 hack/e2e-apps/remediation-guard.sh            | 18 ++++++++--------
 hack/e2e-apps/run-kubernetes.sh               | 12 +++++++++--
 .../kubernetes/tests/values-ci-no-etcd.yaml   |  9 ++++++++
 .../kubernetes-rd/cozyrds/kubernetes.yaml     |  8 ++++++-
 pkg/registry/apps/application/rest.go         | 21 +++++++++++--------
 6 files changed, 49 insertions(+), 35 deletions(-)
 create mode 100644 packages/apps/kubernetes/tests/values-ci-no-etcd.yaml

diff --git a/hack/admin-kubeconfig-invariant.bats b/hack/admin-kubeconfig-invariant.bats
index 4c809c41..a4b98b00 100644
--- a/hack/admin-kubeconfig-invariant.bats
+++ b/hack/admin-kubeconfig-invariant.bats
@@ -85,13 +85,7 @@
 
   helm template invariant packages/apps/kubernetes \
     --namespace tenant-root \
-    --set _namespace.etcd="" \
-    --set _namespace.monitoring="" \
-    --set _namespace.ingress="" \
-    --set _namespace.seaweedfs="" \
-    --set _namespace.host="" \
-    --set _cluster.cluster-domain=cozy.local \
-    --set 'nodeGroups=null' \
+    --values packages/apps/kubernetes/tests/values-ci-no-etcd.yaml \
     2>/dev/null > "$tmp/rendered.yaml"
 
   matched=$(
@@ -128,13 +122,7 @@
 
   helm template invariant packages/apps/kubernetes \
     --namespace tenant-root \
-    --set _namespace.etcd="" \
-    --set _namespace.monitoring="" \
-    --set _namespace.ingress="" \
-    --set _namespace.seaweedfs="" \
-    --set _namespace.host="" \
-    --set _cluster.cluster-domain=cozy.local \
-    --set 'nodeGroups=null' \
+    --values packages/apps/kubernetes/tests/values-ci-no-etcd.yaml \
     2>/dev/null > "$tmp/rendered.yaml"
 
   matched=$(
diff --git a/hack/e2e-apps/remediation-guard.sh b/hack/e2e-apps/remediation-guard.sh
index 38463310..b45561b9 100644
--- a/hack/e2e-apps/remediation-guard.sh
+++ b/hack/e2e-apps/remediation-guard.sh
@@ -26,14 +26,14 @@ helmrelease_has_remediation_cycle() {
     if [ -z "${statuses}" ]; then
         return 1
     fi
-    while IFS= read -r status; do
-        case "${status}" in
-            failed|uninstalled)
-                return 0
-                ;;
-        esac
-    done <<EOF
-${statuses}
-EOF
+    # 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 64b1ae90..761ee57d 100644
--- a/hack/e2e-apps/run-kubernetes.sh
+++ b/hack/e2e-apps/run-kubernetes.sh
@@ -336,9 +336,17 @@ EOF
   # 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. History statuses:" >&2
-    printf '%s\n' "${history_statuses}" >&2
+    echo "Parent HelmRelease entered remediation cycle." >&2
     kubectl -n tenant-test describe hr "kubernetes-${test_name}" >&2
     exit 1
   fi
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/system/kubernetes-rd/cozyrds/kubernetes.yaml b/packages/system/kubernetes-rd/cozyrds/kubernetes.yaml
index 1128abe8..aaa01247 100644
--- a/packages/system/kubernetes-rd/cozyrds/kubernetes.yaml
+++ b/packages/system/kubernetes-rd/cozyrds/kubernetes.yaml
@@ -7,8 +7,14 @@ metadata:
     # 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 applied by cozystack-api
+    # 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:
diff --git a/pkg/registry/apps/application/rest.go b/pkg/registry/apps/application/rest.go
index 406d3738..d233ecc3 100644
--- a/pkg/registry/apps/application/rest.go
+++ b/pkg/registry/apps/application/rest.go
@@ -1528,15 +1528,18 @@ func (r *REST) convertApplicationToHelmRelease(app *appsv1alpha1.Application) (*
 		},
 	}
 
-	// Per-Application HelmRelease wait budget. When an ApplicationDefinition
-	// sets release.cozystack.io/helm-install-timeout, the annotation is
-	// parsed at startup into ReleaseConfig.HelmInstallTimeout and applied
-	// to both Install and Upgrade here. Applications that leave it unset
-	// (the common case) keep flux defaults, so their failed installs
-	// remediate on the normal cadence. Needed for the Kubernetes kind
-	// because its parent chart contains CAPI/Kamaji resources whose
-	// admin-kubeconfig Secret is provisioned asynchronously and Kamaji
-	// cold-start routinely exceeds flux's default wait budget.
+	// 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

From 39b8f0252b6fb07d510c83aacf48fc20d7fdd96c Mon Sep 17 00:00:00 2001
From: Aleksei Sviridkin <f@lex.la>
Date: Thu, 16 Apr 2026 22:32:24 +0300
Subject: [PATCH 24/26] test(hack): rename remediation-guard bats test to match
 what it pins

The test body asserts .status.history[].status extraction, but the
test name still referenced the old installFailures counter (leftover
from when the guard used that field before switching to status.history
to avoid ClearFailures zeroing the counters on successful reconcile).

Address review feedback from coderabbitai on
hack/remediation-guard.bats:84: rename so grep for what the test
actually pins matches the test name.

Assisted-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
---
 hack/remediation-guard.bats | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/hack/remediation-guard.bats b/hack/remediation-guard.bats
index 30e4e04c..092fe06d 100644
--- a/hack/remediation-guard.bats
+++ b/hack/remediation-guard.bats
@@ -81,7 +81,7 @@
     fi
 }
 
-@test "installFailures extraction pins HR v2 status.history shape" {
+@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,

From 37ecd7b3af9f96ef8cbcb097b8b88a4ffbcef548 Mon Sep 17 00:00:00 2001
From: Aleksei Sviridkin <f@lex.la>
Date: Mon, 27 Apr 2026 13:50:08 +0300
Subject: [PATCH 25/26] feat(kubernetes): make wait-for-kubeconfig image
 overridable for air-gapped registries

Operators in air-gapped or rate-limited environments cannot reach
docker.io and the bundled busybox digest pin gives them no escape
hatch. Add an optional images.waitForKubeconfig chart value that, when
set, replaces the helper's image reference with any registry path
kubelet can pull. Empty value falls back to images/busybox.tag, so the
prior digest-pinned default is preserved.

Assisted-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
---
 api/apps/v1alpha1/kubernetes/types.go         |  9 ++++
 packages/apps/kubernetes/README.md            | 52 ++++++++++---------
 .../apps/kubernetes/templates/_helpers.tpl    | 12 +++--
 .../tests/admin_kubeconfig_wait_test.yaml     | 23 ++++++++
 packages/apps/kubernetes/values.schema.json   | 12 +++++
 packages/apps/kubernetes/values.yaml          |  7 +++
 .../kubernetes-rd/cozyrds/kubernetes.yaml     |  4 +-
 7 files changed, 87 insertions(+), 32 deletions(-)

diff --git a/api/apps/v1alpha1/kubernetes/types.go b/api/apps/v1alpha1/kubernetes/types.go
index 774fcdc5..ea4a2b64 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 {
@@ -157,6 +160,12 @@ type GatewayAPIAddon struct {
 	Enabled bool `json:"enabled"`
 }
 
+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/packages/apps/kubernetes/README.md b/packages/apps/kubernetes/README.md
index d62d3a39..ecbbc31b 100644
--- a/packages/apps/kubernetes/README.md
+++ b/packages/apps/kubernetes/README.md
@@ -145,31 +145,33 @@ 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/templates/_helpers.tpl b/packages/apps/kubernetes/templates/_helpers.tpl
index e6ad9dcf..40a8ae7a 100644
--- a/packages/apps/kubernetes/templates/_helpers.tpl
+++ b/packages/apps/kubernetes/templates/_helpers.tpl
@@ -65,18 +65,20 @@ release.cozystack.io/helm-install-timeout annotation) so the
 CrashLoopBackOff surfaces before flux remediation fires and uninstalls
 the Cluster CR.
 
-The pinned busybox image in images/busybox.tag points directly at
+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, the digest pin makes the
-pull immutable, and the cost of maintaining a private mirror of a tiny
-upstream image that does not move often is not worth it.
+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: "{{ .Files.Get "images/busybox.tag" | trim }}"
+  image: "{{ default (.Files.Get "images/busybox.tag" | trim) .Values.images.waitForKubeconfig }}"
   command:
   - sh
   - -c
diff --git a/packages/apps/kubernetes/tests/admin_kubeconfig_wait_test.yaml b/packages/apps/kubernetes/tests/admin_kubeconfig_wait_test.yaml
index e7c03bd1..507e00af 100644
--- a/packages/apps/kubernetes/tests/admin_kubeconfig_wait_test.yaml
+++ b/packages/apps/kubernetes/tests/admin_kubeconfig_wait_test.yaml
@@ -95,6 +95,29 @@ tests:
             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:
diff --git a/packages/apps/kubernetes/values.schema.json b/packages/apps/kubernetes/values.schema.json
index 1beeff9c..81e9e2b8 100644
--- a/packages/apps/kubernetes/values.schema.json
+++ b/packages/apps/kubernetes/values.schema.json
@@ -630,6 +630,18 @@
           }
         }
       }
+    },
+    "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..68ba07b0 100644
--- a/packages/apps/kubernetes/values.yaml
+++ b/packages/apps/kubernetes/values.yaml
@@ -197,3 +197,10 @@ 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/system/kubernetes-rd/cozyrds/kubernetes.yaml b/packages/system/kubernetes-rd/cozyrds/kubernetes.yaml
index aaa01247..1631c86e 100644
--- a/packages/system/kubernetes-rd/cozyrds/kubernetes.yaml
+++ b/packages/system/kubernetes-rd/cozyrds/kubernetes.yaml
@@ -22,7 +22,7 @@ spec:
     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","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"]}}}}},"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:
@@ -40,7 +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", "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:

From fb1ef59287114fe9ef6eec2351875139d4d84e97 Mon Sep 17 00:00:00 2001
From: Aleksei Sviridkin <f@lex.la>
Date: Mon, 27 Apr 2026 13:50:18 +0300
Subject: [PATCH 26/26] fix(kubernetes): drop undocumented status-beacon
 annotation from awaiting-etcd ConfigMap

The cozystack.io/status-beacon: "true" annotation had no consumer in
the chart, no documented contract, and no convention defined for other
charts to follow. It would have become accidental precedent for
contributors copying the pattern without understanding it.

The ConfigMap itself is self-explanatory: the name <release>-awaiting-etcd,
data.status: "awaiting-etcd", and the human-readable message in
data.message all surface the same operator signal via kubectl get cm.
Drop the annotation; keep the ConfigMap.

Assisted-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
---
 packages/apps/kubernetes/templates/cluster.yaml | 2 --
 1 file changed, 2 deletions(-)

diff --git a/packages/apps/kubernetes/templates/cluster.yaml b/packages/apps/kubernetes/templates/cluster.yaml
index 12da9818..e1494e64 100644
--- a/packages/apps/kubernetes/templates/cluster.yaml
+++ b/packages/apps/kubernetes/templates/cluster.yaml
@@ -106,8 +106,6 @@ metadata:
     app.kubernetes.io/managed-by: {{ .Release.Service }}
     app.kubernetes.io/name: kubernetes
     app.kubernetes.io/instance: {{ .Release.Name }}
-  annotations:
-    cozystack.io/status-beacon: "true"
 data:
   status: "awaiting-etcd"
   message: |