Check kube-version inside the build recipe

Render charts with --helm-kube-version from ci/kube-version so kustomize
no longer inherits the local helm binary's default Kubernetes version.

AGENTS.md

@@ -22,10 +22,10 @@ make kubeconform
 make clean
 
 # Quick build + inspect without persisting output
-kustomize build --enable-helm apps/overlays/au-syd1/<app-name>
+kustomize build --enable-helm --helm-kube-version "$(cat ci/kube-version)" apps/overlays/au-syd1/<app-name>
 
 # Check all resource kinds produced by an overlay
-kustomize build --enable-helm apps/overlays/au-syd1/<app-name> | grep "^kind:" | sort | uniq -c
+kustomize build --enable-helm --helm-kube-version "$(cat ci/kube-version)" apps/overlays/au-syd1/<app-name> | grep "^kind:" | sort | uniq -c
 
 # Run pre-commit checks against all files
 uvx pre-commit run --all-files
@@ -118,7 +118,7 @@ In `argocd/projects/platform.yaml` (or `storage.yaml`):
 
 ### 5. Validate
 ```bash
-kustomize build --enable-helm apps/overlays/au-syd1/<app-name>
+kustomize build --enable-helm --helm-kube-version "$(cat ci/kube-version)" apps/overlays/au-syd1/<app-name>
 make kubeconform
 ```
 
@@ -215,7 +215,7 @@ Examples:
 
 Some overlays vendor Helm charts locally under `apps/overlays/au-syd1/<app-name>/charts/<chart-name>/`. When a chart is vendored, the overlay's `kustomization.yaml` references the local path. When not vendored, it references the OCI or HTTP repo directly.
 
-Current Kubernetes target version: **1.33.7** (used by kubeconform in CI).
+The Kubernetes target version lives in `ci/kube-version`; the Makefile and CI scripts read it for both the Helm capabilities version and kubeconform.
 
 ---
 

Makefile

@@ -3,8 +3,11 @@
 # Build a kustomization path to manifests directory
 # Usage: make build clusters/au-syd1/bootstrap
 build:
-	@mkdir -p manifests/$(filter-out $@,$(MAKECMDGOALS))
-	@kustomize build --enable-helm $(filter-out $@,$(MAKECMDGOALS)) --output manifests/$(filter-out $@,$(MAKECMDGOALS))
+	@kube_version="$$(cat ci/kube-version)"; \
+	if [ -z "$$kube_version" ]; then echo "cannot read ci/kube-version" >&2; exit 1; fi; \
+	target="$(filter-out $@,$(MAKECMDGOALS))"; \
+	mkdir -p "manifests/$$target"; \
+	kustomize build --enable-helm --helm-kube-version "$$kube_version" "$$target" --output "manifests/$$target"
 
 # Generate JSON schemas from CRDs and Kubernetes swagger spec (run manually, results committed)
 schemas:

apps/base/authentik/cnpg_cluster.yaml

@@ -64,12 +64,8 @@ spec:
       archive_mode: "on"
       archive_timeout: 5min
       dynamic_shared_memory_type: posix
-      effective_cache_size: 1536MB
+      effective_cache_size: 256MB
       full_page_writes: "on"
-      # Replicas report their oldest xmin to the primary, so multi-second reads on
-      # a hot standby stop exhausting max_standby_streaming_delay and being
-      # cancelled. Retained-dead-tuple cost is negligible on a ~155MB database.
-      hot_standby_feedback: "on"
       log_destination: csvlog
       log_directory: /controller/log
       log_filename: postgres
@@ -81,12 +77,7 @@ spec:
       max_parallel_workers: "16"
       max_replication_slots: "16"
       max_worker_processes: "16"
-      # A pg_stat_statements.* parameter is what makes CNPG treat the extension as
-      # managed and run CREATE EXTENSION in every database; preloading alone does
-      # not create it.
-      pg_stat_statements.max: "10000"
-      pg_stat_statements.track: top
-      shared_buffers: 512MB
+      shared_buffers: 128MB
       shared_memory_type: mmap
       ssl_max_protocol_version: TLSv1.3
       ssl_min_protocol_version: TLSv1.3
@@ -95,9 +86,6 @@ spec:
       wal_log_hints: "on"
       wal_receiver_timeout: 5s
       wal_sender_timeout: 5s
-    # CNPG merges this with the libraries it manages itself.
-    shared_preload_libraries:
-      - pg_stat_statements
     syncReplicaElectionConstraint:
       enabled: false
   primaryUpdateMethod: restart
@@ -117,16 +105,13 @@ spec:
     updateInterval: 30
   resources:
     limits:
-      # 500m is a 50ms CFS quota per 100ms period, exhausted by bursts even at
-      # ~0.01 cores average, so every query pays throttle latency.
-      cpu: "2"
+      cpu: 500m
       # 512Mi OOMKilled replicas under load (shared_buffers 128MB +
       # max_connections 200 leave no headroom) — see incident 2026-07-28.
-      # shared_buffers 512MB needs the same headroom multiple, hence 2Gi.
-      memory: 2Gi
-    requests:
-      cpu: 500m
       memory: 1Gi
+    requests:
+      cpu: 50m
+      memory: 512Mi
   smartShutdownTimeout: 180
   startDelay: 3600
   stopDelay: 1800

apps/base/authentik/httproute.yaml

@@ -37,22 +37,6 @@ spec:
       name: authentik
       sectionName: https
   rules:
-    - backendRefs:
-        - group: ""
-          kind: Service
-          name: authentik-server
-          port: 80
-          weight: 1
-      filters:
-        - type: URLRewrite
-          urlRewrite:
-            path:
-              type: ReplaceFullPath
-              replaceFullPath: /application/o/token/
-      matches:
-        - path:
-            type: Exact
-            value: /application/o/token
     - backendRefs:
         - group: ""
           kind: Service
@@ -102,22 +86,6 @@ spec:
       name: authentik-internal
       sectionName: https
   rules:
-    - backendRefs:
-        - group: ""
-          kind: Service
-          name: authentik-server
-          port: 80
-          weight: 1
-      filters:
-        - type: URLRewrite
-          urlRewrite:
-            path:
-              type: ReplaceFullPath
-              replaceFullPath: /application/o/token/
-      matches:
-        - path:
-            type: Exact
-            value: /application/o/token
     - backendRefs:
         - group: ""
           kind: Service

apps/base/authentik/kustomization.yaml

@@ -19,7 +19,6 @@ resources:
   - redis-deployment.yaml
   - redis-pvc.yaml
   - redis-service.yaml
-  - server-vmpodscrape.yaml
   - vaultauth.yaml
   - vaultstaticsecret.yaml
   - vmpodscrape.yaml

apps/base/authentik/server-vmpodscrape.yaml

@@ -1,16 +0,0 @@
----
-# Scrape the authentik server's django_prometheus endpoint (:9300). Picked up
-# by the observability VMAgent (selectAllByDefault).
-apiVersion: operator.victoriametrics.com/v1beta1
-kind: VMPodScrape
-metadata:
-  name: authentik-server
-  namespace: authentik
-spec:
-  selector:
-    matchLabels:
-      app.kubernetes.io/name: authentik
-      app.kubernetes.io/component: server
-  podMetricsEndpoints:
-    - port: metrics
-      path: /metrics

apps/base/bind-system/deployment.yaml

@@ -21,7 +21,7 @@ spec:
         runAsNonRoot: true
       containers:
         - name: operator
-          image: artifactapi.k8s.syd1.au.unkin.net/docker-internal/bind-operator:v0.2.7
+          image: artifactapi.k8s.syd1.au.unkin.net/docker-internal/bind-operator:v0.2.6
           args:
             - --metrics-bind-address=:8080
             - --health-probe-bind-address=:8081

apps/base/bind-system/kustomization.yaml

@@ -6,7 +6,7 @@ resources:
   - namespace.yaml
   # CRDs are pulled from the bind-operator repo at the matching tag rather than
   # vendored here, so they never drift from the operator.
-  - https://git.unkin.net/unkin/bind-operator/raw/tag/v0.2.7/config/crd/install.yaml
+  - https://git.unkin.net/unkin/bind-operator/raw/tag/v0.2.6/config/crd/install.yaml
   - rbac.yaml
   - agent-dns-rbac.yaml
   - deployment.yaml

apps/base/cheeztv/service.yaml

@@ -13,4 +13,6 @@ spec:
       targetPort: http
   selector:
     app: cheeztv
+  # Pin each client to one replica to reduce transcode-session churn/takeover.
+  sessionAffinity: ClientIP
   type: ClusterIP

apps/base/cheeztv/statefulset.yaml

@@ -4,8 +4,6 @@ kind: StatefulSet
 metadata:
   name: cheeztv
   namespace: cheeztv
-  annotations:
-    configmap.reloader.stakater.com/auto: "true"
 spec:
   # HA: two replicas coordinate transcode session ownership through Valkey and
   # resume each other's HLS segments off the shared RWX transcode PVC. Stable
@@ -164,7 +162,7 @@ spec:
               readOnly: true
       containers:
         - name: cheeztv
-          image: artifactapi.k8s.syd1.au.unkin.net/docker-internal/jellyfin-ha:v0.3.3
+          image: artifactapi.k8s.syd1.au.unkin.net/docker-internal/jellyfin-ha:v0.3.2
           imagePullPolicy: IfNotPresent
           ports:
             - name: http

apps/base/fafflix/service.yaml

@@ -13,4 +13,6 @@ spec:
       targetPort: http
   selector:
     app: fafflix
+  # Pin each client to one replica to reduce transcode-session churn/takeover.
+  sessionAffinity: ClientIP
   type: ClusterIP

apps/base/fafflix/statefulset.yaml

@@ -4,8 +4,6 @@ kind: StatefulSet
 metadata:
   name: fafflix
   namespace: fafflix
-  annotations:
-    configmap.reloader.stakater.com/auto: "true"
 spec:
   # HA: two replicas coordinate transcode session ownership through Valkey and
   # resume each other's HLS segments off the shared RWX transcode PVC. Stable
@@ -164,7 +162,7 @@ spec:
               readOnly: true
       containers:
         - name: fafflix
-          image: artifactapi.k8s.syd1.au.unkin.net/docker-internal/jellyfin-ha:v0.3.3
+          image: artifactapi.k8s.syd1.au.unkin.net/docker-internal/jellyfin-ha:v0.3.2
           imagePullPolicy: IfNotPresent
           ports:
             - name: http

apps/base/woodpecker/kustomization.yaml

@@ -15,7 +15,6 @@ resources:
   - serviceaccount_mediamark_ci.yaml
   - serviceaccount_plugin_docker_buildx.yaml
   - serviceaccount_jellyfin_ha_src.yaml
-  - serviceaccount_jellyfin_plugin_sso.yaml
   - serviceaccount_repospawner_ci.yaml
   - serviceaccount_terraform_artifactapi.yaml
   - serviceaccount_terraform_authentik.yaml

apps/base/woodpecker/serviceaccount_jellyfin_plugin_sso.yaml

@@ -1,6 +0,0 @@
----
-apiVersion: v1
-kind: ServiceAccount
-metadata:
-  name: jellyfin-plugin-sso
-  namespace: woodpecker

ci/kube-version

@@ -0,0 +1 @@
+1.33.13

ci/validate-apps.sh

@@ -1,7 +1,7 @@
 #!/usr/bin/env bash
 set -euo pipefail
 
-KUBE_VERSION="1.33.7"
+KUBE_VERSION="$(cat ci/kube-version)"
 
 SCHEMA_DIR="${SCHEMA_DIR:-schemas}"
 
@@ -15,7 +15,7 @@ while IFS= read -r -d "" k; do
   dir="$(dirname "$k")"
   echo "==> kubeconform: $dir" >&2
 
-  kustomize build --enable-helm "$dir" \
+  kustomize build --enable-helm --helm-kube-version "$KUBE_VERSION" "$dir" \
     | kubeconform \
         -kubernetes-version "$KUBE_VERSION" \
         -summary \

ci/validate-clusters.sh

@@ -1,7 +1,7 @@
 #!/usr/bin/env bash
 set -euo pipefail
 
-KUBE_VERSION="1.33.7"
+KUBE_VERSION="$(cat ci/kube-version)"
 
 SCHEMA_DIR="${SCHEMA_DIR:-schemas}"
 
@@ -15,7 +15,7 @@ while IFS= read -r -d "" k; do
   dir="$(dirname "$k")"
   echo "==> kubeconform: $dir" >&2
 
-  kustomize build --enable-helm "$dir" \
+  kustomize build --enable-helm --helm-kube-version "$KUBE_VERSION" "$dir" \
     | kubeconform \
         -kubernetes-version "$KUBE_VERSION" \
         -summary \

clusters/au-syd1/bootstrap/argocd-cm-patch.yaml

@@ -26,10 +26,6 @@ data:
     issuer: https://identity.unkin.net/application/o/argocd/
     clientID: argocd
     clientSecret: $argocd-oidc:client_secret
-    # The Authentik client is public (the iOS app can't hold a secret), so
-    # Authentik no longer enforces clientSecret; PKCE replaces it as the
-    # protection against authorization-code interception.
-    enablePKCEAuthentication: true
     # identity.unkin.net now serves the LetsEncrypt *.unkin.net wildcard, so the
     # stock image trust store validates it; no rootCA pin.
     requestedScopes:

migration.md

@@ -49,8 +49,8 @@
 - Use existing patterns like `apps/overlays/*/csi-*` or `apps/overlays/*/<app-name>`
 
 ### 9. Validation
-- Run `kustomize build --enable-helm apps/overlays/au-syd1/<app-name>` to generate all resources
-- Check resource types: `kustomize build --enable-helm apps/overlays/au-syd1/<app-name> | grep "^kind:" | sort | uniq -c`
+- Run `kustomize build --enable-helm --helm-kube-version "$(cat ci/kube-version)" apps/overlays/au-syd1/<app-name>` to generate all resources
+- Check resource types: `kustomize build --enable-helm --helm-kube-version "$(cat ci/kube-version)" apps/overlays/au-syd1/<app-name> | grep "^kind:" | sort | uniq -c`
 - Verify all resource types are permitted in the target project's `clusterResourceWhitelist` and `namespaceResourceWhitelist`
 - Run `make kubeconform` to validate all resources
 - Fix any validation errors