ci: publish the vpc-controller image and kustomize bundle

.github/workflows/publish.yaml

@@ -0,0 +1,40 @@
+name: Publish
+
+on:
+  push:
+  release:
+    types: ["published"]
+
+jobs:
+  validate-kustomize:
+    uses: datum-cloud/actions/.github/workflows/validate-kustomize.yaml@v1.21.0
+
+  publish-container-image:
+    needs:
+      - validate-kustomize
+    permissions:
+      id-token: write
+      contents: read
+      packages: write
+      attestations: write
+    uses: datum-cloud/actions/.github/workflows/publish-docker.yaml@v1.21.0
+    with:
+      image-name: vpc-controller
+      platforms: linux/amd64,linux/arm64
+    secrets: inherit
+
+  # The bundle pins the image to this build, so it publishes after the image.
+  publish-kustomize-bundles:
+    needs:
+      - publish-container-image
+    permissions:
+      id-token: write
+      contents: read
+      packages: write
+    uses: datum-cloud/actions/.github/workflows/publish-kustomize-bundle.yaml@v1.21.0
+    with:
+      bundle-name: ghcr.io/datum-cloud/vpc-controller-kustomize
+      bundle-path: config
+      image-name: ghcr.io/datum-cloud/vpc-controller
+      image-overlays: config/manager
+    secrets: inherit

README.md

@@ -60,6 +60,13 @@ kubectl apply -k config/crd      # types only
 kubectl apply -k config/default  # types, RBAC and the controller
 ```
 
+`config/default` requires cert-manager: it issues the webhook serving certificate and injects the CA into the `MutatingWebhookConfiguration`. The attachment mode is carried as the `ATTACHMENT_MODE` environment variable on the manager container (default `Hypervisor`), so a cell selects its mode with a small overlay patch rather than rewriting the args list.
+
+CI publishes on every push and release:
+
+- image: `ghcr.io/datum-cloud/vpc-controller`
+- kustomize bundle (OCI, for Flux `OCIRepository`): `ghcr.io/datum-cloud/vpc-controller-kustomize`, with the image pinned to the matching tag
+
 ## Development
 
 ```bash

cmd/main.go

@@ -137,7 +137,10 @@ func main() {
 		setupLog.Error(err, "unable to set up health check")
 		os.Exit(1)
 	}
-	if err := mgr.AddReadyzCheck("readyz", healthz.Ping); err != nil {
+	// Gate readiness on the webhook server: with failurePolicy Fail, a Pod that
+	// reports Ready before the webhook serves would take endpoints and reject
+	// every labelled Pod in the cell.
+	if err := mgr.AddReadyzCheck("webhook", mgr.GetWebhookServer().StartedChecker()); err != nil {
 		setupLog.Error(err, "unable to set up ready check")
 		os.Exit(1)
 	}

config/manager/kustomization.yaml

@@ -2,3 +2,7 @@ apiVersion: kustomize.config.k8s.io/v1beta1
 kind: Kustomization
 resources:
   - manager.yaml
+images:
+  - name: ghcr.io/datum-cloud/vpc-controller
+    newName: ghcr.io/datum-cloud/vpc-controller
+    newTag: latest

config/manager/manager.yaml

@@ -27,13 +27,16 @@ spec:
           type: RuntimeDefault
       containers:
         - name: manager
-          image: ghcr.io/datum-cloud/vpc-controller:latest
+          image: ghcr.io/datum-cloud/vpc-controller
           args:
             - --leader-elect
-            # Required. Hypervisor for microVM cells, Netns for container cells.
-            - --attachment-mode=Hypervisor
+            - --attachment-mode=$(ATTACHMENT_MODE)
             - --health-probe-bind-address=:8081
             - --metrics-bind-address=:8080
+          env:
+            # Required flag, carried as env so an overlay can retarget a cell by name.
+            - name: ATTACHMENT_MODE
+              value: Hypervisor
           ports:
             - name: metrics
               containerPort: 8080
@@ -49,12 +52,15 @@ spec:
               port: 8081
             initialDelaySeconds: 15
             periodSeconds: 20
+          # /readyz includes the webhook server's started check, so the Service
+          # only gets an endpoint once the webhook can actually admit Pods.
           readinessProbe:
             httpGet:
               path: /readyz
               port: 8081
             initialDelaySeconds: 5
-            periodSeconds: 10
+            periodSeconds: 5
+            failureThreshold: 3
           securityContext:
             allowPrivilegeEscalation: false
             capabilities: