Merge Phase-2 no-secret rollout prep into main

DECISIONS-LOG.md

@@ -91,6 +91,67 @@ primary play runs, so a limit that excludes the primary yields
   self-contained. `ansible-playbook --syntax-check` OK on all playbooks; role lint clean under the CI
   gate (`make -C ci gate` PASS). No cluster change; the running cluster is untouched.
 
+## Phase 2 un-suspend PREP — no-secret rollout (author-first, 2026-08-16)
+
+Operator directive: "Phase 2 un-suspend prep … verify cilium parity + fill REPLACE-MEs for the
+no-secret components … author-first. Does NOT un-suspend/deploy yet." **Nothing was deployed;** the
+running cluster is untouched (still 0 Flux Kustomizations / 0 HelmReleases — only controllers + the
+GitRepository).
+
+### Findings from three parallel readiness audits (Longhorn / monitoring / Kyverno)
+- **Cilium parity VERIFIED (safety-critical):** every `[parity]` field in `platform/cilium/helmrelease.yaml`
+  equals the live CNI — `k8sServiceHost=10.73.0.11` (rendered by the chart as the pod env
+  `KUBERNETES_SERVICE_HOST`, **not** a configmap key — the empty `k8s-service-host` cm value was a red
+  herring), kube-proxy-replacement, gateway-api, egress+bpf masquerade, hubble, l2-announcements,
+  cgroup, v1.16.5. **Flux adoption is a no-op reconcile.**
+- **Topology:** "suspended" = Option A (**no root Kustomization was ever created**), and the single base
+  `platform` Kustomization renders **all 11 components at once** — so a naive un-suspend would drag in
+  the secret-blocked components (cert-manager/externaldns/velero/keycloak/oauth2-proxy/image-automation)
+  and never go healthy (`apps` dependsOn platform, wait:true).
+- **Longhorn:** manifests clean; node prereqs (open-iscsi, /var/lib/longhorn mount, iscsid) are in
+  cloud-init IaC; the REPLACE-ME bulk-NFS SC is isolated/harmless. Ready.
+- **Monitoring:** right-sized (~1.25 GiB) but blocked by (a) needing the `longhorn` SC first and (b) two
+  **by-name secrets** (`grafana-admin-credentials`, `alertmanager-smtp`) that don't exist — the missing
+  mounts (not the REPLACE-ME text) are what wedge Grafana/Alertmanager.
+- **Kyverno:** zero policy violations today (policies scope to `app-namespace=true`, none exist), but the
+  fail-closed webhook + 3 HA replicas + Enforce are unsafe/oversized for a cold cluster.
+
+### DECISION D-P2A — per-component Flux Kustomizations for the dev no-secret rollout (operator-approved)
+Chose **Option A (per-component Kustomizations)**. Authored `clusters/dev/rollout/` = 4 Flux
+Kustomizations (`kyverno` → `cilium` → `longhorn` → `monitoring`, monitoring `dependsOn: longhorn`),
+and the dev overlay now `$patch: delete`s the monolithic base `platform`/`apps` and adds `./rollout`.
+The no-secret tier carries **no SOPS content**, so the missing `sops-age` key does not block it
+(decryption dropped from these Kustomizations). Secret-blocked components are simply **not created**
+until their hand-offs land.
+- **cilium** uses `prune: false` (adopt the out-of-band CNI; never own-delete it).
+- **R-ENV-02 divergence — REVIEW FLAG (architect):** the dev overlay now *adds resources* and diverges
+  its reconciliation graph from prod (base doc says "dev adds NO resources … graph reused verbatim").
+  This is a **transitional bootstrap scaffold**, consistent with the already-inlined dev bootstrap
+  patches (SOPS/URL). **Convergence path is coded in:** delete `clusters/dev/rollout/` + the two
+  `$patch: delete`s to re-adopt the unified base graph once all hand-offs arrive. **prod is untouched**
+  (still builds the monolith; verified). Flag for architect ratification.
+
+### Component hardening (env-gated via cluster-vars so prod keeps the strict posture)
+- New base knobs (strict/prod defaults) + dev overrides: `KYVERNO_VALIDATION_ACTION` (Enforce→**Audit**),
+  `KYVERNO_WEBHOOK_FAILURE_POLICY` (Fail→**Ignore**), `KYVERNO_ADMISSION_REPLICAS` (3→**1**),
+  `MONITORING_STORAGE_CLASS` (longhorn), `GRAFANA_ADMIN_SECRET` (grafana-admin-credentials→**""** in dev
+  → chart-default password, dev-only/NAT-isolated).
+- Kyverno webhook also gains an explicit `namespaceSelector` excluding kube-system/flux-system/kyverno/
+  longhorn-system/monitoring (belt-and-suspenders vs a fail-closed brick).
+- Monitoring Alertmanager: the `alertmanager-smtp` mount + smtp_*/email_configs are **deferred** (commented
+  with verbatim re-enable markers) so Alertmanager starts with NULL receivers — alerts fire/route/show in
+  UI+Grafana, just no email until the Phase-4 SMTP hand-off.
+- Validated: `kubectl kustomize` clean for clusters/dev (exactly the 4 Kustomizations, monolith removed),
+  clusters/prod (monolith intact), and all 4 components; envsubst→YAML-parse of kyverno+monitoring valid;
+  `make -C ci gate` PASS.
+
+### Still required BEFORE an actual deploy (not part of this author-first prep)
+1. **Push current `main` to the host Gitea** — Flux source is pinned at `6c3f9ed` (pre-dates these merges).
+2. **Runtime node check:** `systemctl is-active iscsid` + `mountpoint /var/lib/longhorn` on all 3 nodes.
+3. **Cilium adoption mechanism:** confirm helm-controller can adopt the Ansible-installed release (was it
+   `helm`-installed with a release secret, or CLI/kubectl?) — verify before creating the cilium Kustomization.
+4. Create the dev root Kustomization (the actual un-suspend) — deliberately NOT done here.
+
 ## KNOWN ISSUES to fix before Phase-2 un-suspend (logged, not yet fixed)
 | # | Issue | Impact | Fix |
 |---|---|---|---|

clusters/base/cluster-vars.yaml

@@ -27,6 +27,15 @@ data:
   MONITORING_PROMETHEUS_RETENTION: "7d"     # dev short | prod long
   MONITORING_PROMETHEUS_PVC_SIZE: "10Gi"
   MONITORING_LOKI_PVC_SIZE: "10Gi"
+  MONITORING_STORAGE_CLASS: "longhorn"      # PVC tier for Prometheus/Loki (dev may override)
+  GRAFANA_ADMIN_SECRET: "grafana-admin-credentials"  # by-name SOPS secret (prod); dev="" → chart default
+
+  # --- admission enforcement (Kyverno, ADR-0011). Base defaults are the STRICT/prod posture;
+  #     the dev overlay softens them for bootstrap (Audit / Ignore / 1) so a cold cluster can
+  #     come up before policies are proven and admission is HA. ---
+  KYVERNO_VALIDATION_ACTION: "Enforce"      # dev: Audit (observe before enforcing)
+  KYVERNO_WEBHOOK_FAILURE_POLICY: "Fail"    # dev: Ignore (fail-open during bootstrap)
+  KYVERNO_ADMISSION_REPLICAS: "3"           # dev: 1 (no HA needed to iterate)
 
   # --- identity / auth (ADR-0014) ---
   KEYCLOAK_REPLICAS: "1"                     # dev 1 | prod 2 (HA)

clusters/dev/kustomization.yaml

@@ -2,39 +2,38 @@
 # clusters/dev — the ACTIVE build target overlay.
 #
 # `flux bootstrap ... --path=clusters/dev` points the dev cluster's Flux here.
-# This overlay adds NO resources of its own. Its ONLY job is to patch the shared cluster-vars
+# TRANSITIONAL (Phase-2 bootstrap): this overlay reuses the base source + cluster-vars, but
-# ConfigMap with dev-scoped values (R-ENV-02). The base reconciliation graph is reused verbatim.
+# REPLACES the monolithic base `platform`/`apps` Kustomizations with a per-component NO-SECRET
+# rollout (./rollout) — so dev brings up cilium/longhorn/monitoring/kyverno in order WITHOUT
+# dragging in the secret-blocked components. It patches cluster-vars with dev values (R-ENV-02)
+# and softens Kyverno for bootstrap. CONVERGENCE: delete ./rollout + the two $patch:delete
+# below to return to the unified base graph once all operator hand-offs land. See DECISIONS-LOG.
 apiVersion: kustomize.config.k8s.io/v1beta1
 kind: Kustomization
 resources:
   - ../base
+  - rollout
 patches:
-  # --- SOPS decryption + bootstrap source (PM-inlined from bootstrap/flux/patches/, 2026-08-16) ---
+  # --- Drop the monolithic base platform/apps Kustomizations. They reconcile ALL 11 components
-  # Inlined (not cross-dir file refs) to stay within kustomize's default load restrictions.
+  #     as one unit (incl. secret-blocked cert-manager/externaldns/velero/keycloak/oauth2-proxy/
-  # Realises the commented decryption TODOs in clusters/base and sets the real source URL.
+  #     image-automation/gateway/kubevirt); ./rollout replaces them with the per-component
-  # The `sops-age` Secret is created out-of-band (bootstrap/sops/install-age-key.sh) — name only.
+  #     no-secret set for dev bootstrap. Remove these two deletes to re-adopt the unified graph.
+  #     (No decryption patch is needed here: the rollout components carry no SOPS content, so
+  #     the `sops-age` Secret is not required for the dev bootstrap tier.) ---
   - patch: |-
+      $patch: delete
       apiVersion: kustomize.toolkit.fluxcd.io/v1
       kind: Kustomization
       metadata:
         name: platform
         namespace: flux-system
-      spec:
-        decryption:
-          provider: sops
-          secretRef:
-            name: sops-age
   - patch: |-
+      $patch: delete
       apiVersion: kustomize.toolkit.fluxcd.io/v1
       kind: Kustomization
       metadata:
         name: apps
         namespace: flux-system
-      spec:
-        decryption:
-          provider: sops
-          secretRef:
-            name: sops-age
   - patch: |-
       apiVersion: source.toolkit.fluxcd.io/v1
       kind: GitRepository
@@ -84,3 +83,18 @@ patches:
       - op: replace
         path: /data/VELERO_RETENTION
         value: "72h"
+      # dev-softened Kyverno for cold-start (base defaults Enforce/Fail/3 = prod posture).
+      - op: replace
+        path: /data/KYVERNO_VALIDATION_ACTION
+        value: "Audit"
+      - op: replace
+        path: /data/KYVERNO_WEBHOOK_FAILURE_POLICY
+        value: "Ignore"
+      - op: replace
+        path: /data/KYVERNO_ADMISSION_REPLICAS
+        value: "1"
+      # dev: no Grafana admin Secret yet → empty makes the chart use its default admin password
+      # (dev-only, NAT-isolated). Prod keeps the by-name SOPS secret from the base default.
+      - op: replace
+        path: /data/GRAFANA_ADMIN_SECRET
+        value: ""

clusters/dev/rollout/cilium.yaml

@@ -0,0 +1,25 @@
+---
+# Cilium — ADOPT the running Ansible-installed CNI. Parity between platform/cilium and the
+# live config was verified (k8sServiceHost/port via pod env, kube-proxy-replacement, gateway-api,
+# egress+bpf masquerade, hubble, l2, cgroup, v1.16.5) → this reconcile is a NO-OP.
+# prune: false — Cilium was bootstrapped out-of-band; adopt its config, never own-delete the CNI.
+# No decryption (no SOPS content). No dependsOn (it is the datapath everything else rides on).
+apiVersion: kustomize.toolkit.fluxcd.io/v1
+kind: Kustomization
+metadata:
+  name: cilium
+  namespace: flux-system
+spec:
+  interval: 30m
+  retryInterval: 2m
+  timeout: 10m
+  prune: false
+  wait: true
+  sourceRef:
+    kind: GitRepository
+    name: flux-system
+  path: ./platform/cilium
+  postBuild:
+    substituteFrom:
+      - kind: ConfigMap
+        name: cluster-vars

clusters/dev/rollout/kustomization.yaml

@@ -0,0 +1,20 @@
+---
+# Phase-2 NO-SECRET incremental rollout — DEV bootstrap scaffold.
+#
+# Per-component Flux Kustomizations for the components that need NO operator secret:
+#   kyverno (admission, dev-softened) · cilium (adopt running CNI) · longhorn (storage)
+#   · monitoring (dependsOn longhorn).
+# Secret-blocked components (cert-manager, externaldns, gateway, velero, keycloak,
+# oauth2-proxy, image-automation, kubevirt) are intentionally NOT created until their
+# operator hand-offs arrive (DECISIONS-LOG A2 / hand-off table).
+#
+# TRANSITIONAL: this is the dev bootstrap graph. Once every hand-off lands, dev drops
+# ./rollout and re-adopts the unified base monolith (clusters/base/platform.yaml) so dev
+# and prod share one reconciliation graph again (R-ENV-02 convergence). See DECISIONS-LOG.
+apiVersion: kustomize.config.k8s.io/v1beta1
+kind: Kustomization
+resources:
+  - kyverno.yaml       # first: admission present (Audit/Ignore in dev — non-blocking) before workloads
+  - cilium.yaml        # adopt the running CNI (parity verified — no-op reconcile)
+  - longhorn.yaml      # storage; monitoring depends on its StorageClass
+  - monitoring.yaml    # dependsOn longhorn

clusters/dev/rollout/kyverno.yaml

@@ -0,0 +1,25 @@
+---
+# Kyverno — in-cluster admission enforcer. DEV-SOFTENED via cluster-vars:
+#   KYVERNO_VALIDATION_ACTION=Audit, KYVERNO_WEBHOOK_FAILURE_POLICY=Ignore,
+#   KYVERNO_ADMISSION_REPLICAS=1 (base defaults Enforce/Fail/3 for prod).
+# Non-blocking on today's cluster (policies match only `app-namespace=true` namespaces,
+# none exist yet). No SOPS content. Listed first so admission is present before workloads.
+apiVersion: kustomize.toolkit.fluxcd.io/v1
+kind: Kustomization
+metadata:
+  name: kyverno
+  namespace: flux-system
+spec:
+  interval: 1h
+  retryInterval: 2m
+  timeout: 10m
+  prune: true
+  wait: true
+  sourceRef:
+    kind: GitRepository
+    name: flux-system
+  path: ./platform/kyverno
+  postBuild:
+    substituteFrom:
+      - kind: ConfigMap
+        name: cluster-vars

clusters/dev/rollout/longhorn.yaml

@@ -0,0 +1,25 @@
+---
+# Longhorn — block storage / CSI (default `longhorn` StorageClass). No SOPS content.
+# No dependsOn. The bulk-NFS StorageClass in this component carries REPLACE-ME (OQ-2) but is
+# an isolated, unreferenced object — it does NOT block Longhorn's HelmRelease reconcile.
+# PRE-DEPLOY CHECK (runtime, out of git): confirm `iscsid` active + /var/lib/longhorn mounted
+# on all 3 nodes (cloud-init provides both; verify after the T-1.5 rebuild).
+apiVersion: kustomize.toolkit.fluxcd.io/v1
+kind: Kustomization
+metadata:
+  name: longhorn
+  namespace: flux-system
+spec:
+  interval: 10m
+  retryInterval: 2m
+  timeout: 10m
+  prune: true
+  wait: true
+  sourceRef:
+    kind: GitRepository
+    name: flux-system
+  path: ./platform/longhorn
+  postBuild:
+    substituteFrom:
+      - kind: ConfigMap
+        name: cluster-vars

clusters/dev/rollout/monitoring.yaml

@@ -0,0 +1,26 @@
+---
+# Monitoring — kube-prometheus-stack + Loki + Alloy. Prometheus/Loki PVCs bind the
+# `longhorn` StorageClass, so this dependsOn longhorn (Flux waits for it Ready first).
+# Grafana admin password (chart default in dev) and Alertmanager SMTP (deferred to the
+# Phase-4 hand-off) are handled WITHOUT an operator secret in dev → NO decryption block.
+apiVersion: kustomize.toolkit.fluxcd.io/v1
+kind: Kustomization
+metadata:
+  name: monitoring
+  namespace: flux-system
+spec:
+  interval: 10m
+  retryInterval: 2m
+  timeout: 15m
+  prune: true
+  wait: true
+  dependsOn:
+    - name: longhorn
+  sourceRef:
+    kind: GitRepository
+    name: flux-system
+  path: ./platform/monitoring
+  postBuild:
+    substituteFrom:
+      - kind: ConfigMap
+        name: cluster-vars

platform/kyverno/admission/helmrelease.yaml

@@ -34,18 +34,32 @@ spec:
     remediation:
       retries: 3
   values:
-    # HA admission controller so a single pod restart cannot open an enforcement gap.
+    # Admission controller replicas — cluster-vars knob (prod 3 / dev 1). HA so a single pod
+    # restart cannot open an enforcement gap; dev drops to 1 to save RAM during bootstrap.
     admissionController:
-      replicas: 3
+      replicas: ${KYVERNO_ADMISSION_REPLICAS}
     backgroundController:
       enabled: true               # needed for generate (default-deny baseline) + sync
     reportsController:
       enabled: true
     cleanupController:
       enabled: true
-    # Fail-CLOSED: if the webhook is unreachable, reject rather than silently admit
+    # Webhook failure policy — cluster-vars knob (prod Fail / dev Ignore). Fail is fail-CLOSED
-    # non-conforming resources (coordination-protocol.md #8). Tune per-resource in
+    # (reject rather than silently admit when the webhook is unreachable, coordination-protocol.md
-    # the policies' webhook config if a bootstrap deadlock appears.
+    # #8); dev uses Ignore (fail-open) so a cold/unhealthy Kyverno cannot brick admission during
+    # bootstrap. The namespaceSelector ALSO excludes system + platform namespaces at the webhook
+    # layer — belt-and-suspenders: policies already scope to app-namespace=true, but this
+    # guarantees kube-system/flux-system/CNI/storage/monitoring admission never routes to Kyverno.
     config:
       webhooks:
-        - failurePolicy: Fail
+        - failurePolicy: ${KYVERNO_WEBHOOK_FAILURE_POLICY}
+          namespaceSelector:
+            matchExpressions:
+              - key: kubernetes.io/metadata.name
+                operator: NotIn
+                values:
+                  - kube-system
+                  - flux-system
+                  - kyverno
+                  - longhorn-system
+                  - monitoring

platform/kyverno/admission/secret-provenance.yaml

@@ -24,7 +24,7 @@ metadata:
     homelab.keithjackson.dev/requirements: R-SEC-01
     homelab.keithjackson.dev/scope: admission-only
 spec:
-  validationFailureAction: Enforce
+  validationFailureAction: ${KYVERNO_VALIDATION_ACTION}
   background: false
   rules:
     - name: secrets-must-be-flux-managed

platform/kyverno/policies/require-auth-mode-label.yaml

@@ -25,7 +25,7 @@ metadata:
       Pods in application namespaces must declare
       homelab.keithjackson.dev/auth-mode = oidc | forward-auth | native | none.
 spec:
-  validationFailureAction: Enforce
+  validationFailureAction: ${KYVERNO_VALIDATION_ACTION}
   background: true
   rules:
     - name: require-auth-mode

platform/kyverno/policies/require-egress-class-label.yaml

@@ -25,7 +25,7 @@ metadata:
       Pods in application namespaces must declare
       homelab.keithjackson.dev/egress-class = internet | vpn.
 spec:
-  validationFailureAction: Enforce
+  validationFailureAction: ${KYVERNO_VALIDATION_ACTION}
   background: true
   rules:
     - name: require-egress-class

platform/kyverno/policies/restrict-host-namespaces.yaml

@@ -24,7 +24,7 @@ metadata:
       Application-namespace pods must not use hostNetwork/hostPID/hostIPC (which
       bypass default-deny) or run privileged containers.
 spec:
-  validationFailureAction: Enforce
+  validationFailureAction: ${KYVERNO_VALIDATION_ACTION}
   background: true
   rules:
     - name: disallow-host-namespaces

platform/kyverno/policies/validate-default-deny-shape.yaml

@@ -20,7 +20,7 @@ metadata:
       to include both Ingress and Egress. Rejects a baseline that silently leaves
       a direction open.
 spec:
-  validationFailureAction: Enforce
+  validationFailureAction: ${KYVERNO_VALIDATION_ACTION}
   background: true
   rules:
     - name: baseline-must-deny-both-directions

platform/monitoring/kube-prometheus-stack.yaml

@@ -62,7 +62,7 @@ spec:
         storageSpec:
           volumeClaimTemplate:
             spec:
-              storageClassName: longhorn                  # replicated tier (ADR-0003)
+              storageClassName: ${MONITORING_STORAGE_CLASS}   # replicated tier (ADR-0003); cluster-vars knob
               accessModes: ["ReadWriteOnce"]
               resources:
                 requests:
@@ -80,22 +80,24 @@ spec:
             memory: 32Mi
           limits:
             memory: 96Mi
-        # SMTP credentials are an OPERATOR HAND-OFF: a SOPS-encrypted Secret, referenced BY NAME
+        # SMTP is an OPERATOR HAND-OFF DEFERRED to Phase 4 — no creds yet, and the by-name
-        # only (docs/07). Mounted at /etc/alertmanager/secrets/<name>/ so the config below can read
+        # `alertmanager-smtp` Secret does NOT exist. Mounting a missing Secret wedges Alertmanager
-        # the password from a file (never inline plaintext).
+        # in ContainerCreating, so the mount is DISABLED until the hand-off. Re-enable verbatim
-        secrets:
+        # (uncomment) when the SOPS `alertmanager-smtp` Secret is delivered. TODO(Phase-4 SMTP).
-          - alertmanager-smtp
+        # secrets:
-      # Routing/receivers. Non-secret SMTP endpoint fields (smarthost/from/to/username) are
+        #   - alertmanager-smtp
-      # operator-supplied placeholders (REPLACE-ME, same convention as velero); the PASSWORD is
+      # Routing stays live; RECEIVERS ARE NULL until the SMTP hand-off (alerts still fire, route,
-      # read from the mounted SOPS secret via *_file. Both severities email; subject/route differ.
+      # and show in the Alertmanager UI + Grafana — they just aren't emailed). The smtp_* globals
+      # and email_configs are retained below, commented, so re-enabling is a pure uncomment.
       config:
         global:
           resolve_timeout: 5m
-          smtp_smarthost: "REPLACE-ME:587"          # operator hand-off (e.g. smtp.gmail.com:587)
+          # TODO(Phase-4 SMTP hand-off): uncomment together with the alertmanager-smtp Secret mount.
-          smtp_from: "REPLACE-ME"                    # operator hand-off (From: address)
+          # smtp_smarthost: "smtp.example.com:587"
-          smtp_auth_username: "REPLACE-ME"           # operator hand-off (SMTP login)
+          # smtp_from: "[email protected]"
-          smtp_auth_password_file: /etc/alertmanager/secrets/alertmanager-smtp/smtp-password
+          # smtp_auth_username: "<smtp-login>"
-          smtp_require_tls: true
+          # smtp_auth_password_file: /etc/alertmanager/secrets/alertmanager-smtp/smtp-password
+          # smtp_require_tls: true
         route:
           group_by: ["alertname", "namespace"]
           group_wait: 30s
@@ -109,25 +111,29 @@ spec:
               group_wait: 10s
               repeat_interval: 1h
         receivers:
+          # NULL receivers until Phase-4 SMTP. email_configs retained (commented) for verbatim re-enable.
           - name: email-warning
-            email_configs:
+            # email_configs:
-              - to: "REPLACE-ME"                     # operator hand-off (alert inbox)
+            #   - to: "<alert-inbox>"
-                send_resolved: true
+            #     send_resolved: true
-                headers:
+            #     headers:
-                  subject: '[WARNING] {{ .CommonLabels.alertname }} ({{ .CommonLabels.namespace }})'
+            #       subject: '[WARNING] {{ .CommonLabels.alertname }} ({{ .CommonLabels.namespace }})'
           - name: email-critical
-            email_configs:
+            # email_configs:
-              - to: "REPLACE-ME"                     # operator hand-off (alert inbox)
+            #   - to: "<alert-inbox>"
-                send_resolved: true
+            #     send_resolved: true
-                headers:
+            #     headers:
-                  subject: '[CRITICAL] {{ .CommonLabels.alertname }} ({{ .CommonLabels.namespace }})'
+            #       subject: '[CRITICAL] {{ .CommonLabels.alertname }} ({{ .CommonLabels.namespace }})'
 
     # --- Grafana: dashboards (R-OBS-03), surfaced in the C2 UI (docs/12) behind Keycloak SSO ---
     grafana:
       enabled: true
       # Admin bootstrap + (later) OIDC client are SOPS secrets referenced BY NAME (WS-SEC/ADR-0014).
+      # cluster-vars knob: prod = the SOPS `grafana-admin-credentials` Secret; dev = "" (empty),
+      # which makes the chart fall back to its default admin password (dev-only, NAT-isolated) so
+      # Grafana starts without an operator secret. TODO: supply a dev Secret if wider dev access.
       admin:
-        existingSecret: grafana-admin-credentials
+        existingSecret: ${GRAFANA_ADMIN_SECRET}
         userKey: admin-user
         passwordKey: admin-password
       # DEV TRIM: emptyDir (default). Dashboards are provisioned from ConfigMaps (sidecar), so no

platform/monitoring/loki.yaml

@@ -60,7 +60,7 @@ spec:
       replicas: 1
       persistence:
         enabled: true
-        storageClass: longhorn          # replicated tier (ADR-0003)
+        storageClass: ${MONITORING_STORAGE_CLASS}   # replicated tier (ADR-0003); cluster-vars knob
         size: ${MONITORING_LOKI_PVC_SIZE}   # overlay knob (dev 5Gi / prod large)
       resources:
         requests: