Merge pull request 'mosquitto: auto-update the broker image (image-automation) + guard the rollout' (#130) from feat/mosquitto-pin-image into master

Reviewed-on: gitea_admin/homelab#130

kubernetes/apps/gatus/configmap.yaml

@@ -124,5 +124,17 @@ data:
           - "[DNS_RCODE] == NOERROR"
           - "[RESPONSE_TIME] < 500"
 
+      # Mosquitto MQTT broker (ClusterIP, home-assistant ns). A TCP-connect
+      # check on :1883: the broker's image tag is auto-bumped by Flux
+      # image-automation straight to master (no CI gate), and the deployment
+      # is a single Recreate/rolling pod, so a bad bump that fails to come up
+      # would otherwise be silent. This turns that into an alert.
+      - name: Mosquitto (MQTT broker)
+        group: infrastructure
+        url: tcp://mosquitto.home-assistant.svc.cluster.local:1883
+        interval: 60s
+        conditions:
+          - "[CONNECTED] == true"
+
       # Removed 2026-05-21: PostgreSQL HA (VIP) tcp://10.2.40.135:5432 —
       # VIP no longer responds (no TCP, no ICMP). Re-add if it returns.

kubernetes/apps/home-assistant/mosquitto-deployment.yaml

@@ -8,7 +8,16 @@ metadata:
 spec:
   replicas: 1
   strategy:
-    type: Recreate
+    # No PVC anymore, so nothing to fence: the new broker can come up and pass
+    # its readiness probe before the old one is torn down. maxUnavailable:0 keeps
+    # a broker serving throughout an image bump / reschedule (clients on the old
+    # pod just reconnect to the new one); maxSurge:1 permits the brief two-pod
+    # overlap — two independent stateless brokers behind the LB, fine since there
+    # is no shared state and discovery self-heals.
+    type: RollingUpdate
+    rollingUpdate:
+      maxUnavailable: 0
+      maxSurge: 1
   selector:
     matchLabels:
       app.kubernetes.io/name: mosquitto
@@ -21,10 +30,10 @@ spec:
       # IoT-VLAN traffic to the broker wherever it lands.
       # Availability: home-critical priority (never preempted, reschedules
       # first) + short fault tolerations so a node failure reschedules the
-      # broker in ~30s instead of the default ~5min. Single replica + Recreate
-      # stays (mosquitto can't run two brokers at once), but with persistence
-      # off there is no PVC to fence/detach/reattach, so it can restart
-      # anywhere instantly and no longer depends on Ceph being healthy.
+      # broker in ~30s instead of the default ~5min. With persistence off there
+      # is no PVC to fence/detach/reattach, so the broker can (re)start anywhere
+      # instantly, no longer depends on Ceph being healthy, and rolls without a
+      # tear-down-first gap (see the RollingUpdate strategy above).
       priorityClassName: home-critical
       terminationGracePeriodSeconds: 10
       tolerations:
@@ -40,7 +49,11 @@ spec:
         fsGroup: 1883
       containers:
         - name: mosquitto
-          image: eclipse-mosquitto:2
+          # Pinned tag, auto-bumped within the 2.x line by Flux image-automation
+          # (infrastructure/image-automation/mosquitto.yaml). The marker tells
+          # Flux where to write; patch/minor updates land as git commits and
+          # roll the broker (brief restart; discovery self-heals via HA).
+          image: eclipse-mosquitto:2.1.2 # {"$imagepolicy": "flux-system:mosquitto"}
           # The shim runs as root (image sets no USER) and prepares the
           # auth files for the broker, which drops to user mosquitto
           # (UID 1883) before reading them — so files must be owned by

kubernetes/infrastructure/image-automation/kustomization.yaml

@@ -17,6 +17,7 @@ resources:
   - paperless-ngx.yaml
   - protonmail-bridge.yaml
   - pve-exporter.yaml
+  - mosquitto.yaml
   - redis.yaml
   - shelly-manager.yaml
   - tika.yaml

kubernetes/infrastructure/image-automation/mosquitto.yaml

@@ -0,0 +1,26 @@
+apiVersion: image.toolkit.fluxcd.io/v1
+kind: ImageRepository
+metadata:
+  name: mosquitto
+  namespace: flux-system
+spec:
+  image: eclipse-mosquitto
+  interval: 1h
+---
+# Stay on the mosquitto 2.x line — auto-bump minor+patch releases; a 3.0 major
+# would be a deliberate migration. Plain-semver tags only, which excludes the
+# `-openssl`, `latest`, and bare `2` tags.
+apiVersion: image.toolkit.fluxcd.io/v1
+kind: ImagePolicy
+metadata:
+  name: mosquitto
+  namespace: flux-system
+spec:
+  imageRepositoryRef:
+    name: mosquitto
+  filterTags:
+    pattern: '^(?P<ver>\d+\.\d+\.\d+)$'
+    extract: '$ver'
+  policy:
+    semver:
+      range: "2.x"

kubernetes/infrastructure/network-policies/default-deny.yaml

@@ -34,6 +34,17 @@ spec:
         - ports:
             - port: "53"
               protocol: UDP
+    # gatus health-checks the mosquitto broker (home-assistant ns) on :1883.
+    # The broker's own ingress policy scopes this to the gatus pod, so this
+    # namespace-wide egress allow is gatus-only in effect.
+    - toEndpoints:
+        - matchLabels:
+            "k8s:io.kubernetes.pod.namespace": home-assistant
+            app.kubernetes.io/name: mosquitto
+      toPorts:
+        - ports:
+            - port: "1883"
+              protocol: TCP
     - toEntities:
         - host
         - remote-node

kubernetes/infrastructure/network-policies/home-assistant-ns.yaml

@@ -41,6 +41,18 @@ spec:
         - ports:
             - port: "1883"
               protocol: TCP
+    # gatus (apps ns) TCP-checks that the broker is listening on :1883 — an
+    # availability probe that fires after image-automation bumps the
+    # eclipse-mosquitto tag, so an image that fails to come up alerts instead
+    # of failing silently. Scoped to the gatus pod on :1883.
+    - fromEndpoints:
+        - matchLabels:
+            "k8s:io.kubernetes.pod.namespace": apps
+            app: gatus
+      toPorts:
+        - ports:
+            - port: "1883"
+              protocol: TCP
     # Kubelet probes (host) and hostNetwork siblings (remote-node).
     - fromEntities:
         - host