Kafkacluster controler tests (#191)

test: add basic object lifecycle tests for KafkaCluster controller

Tests verify:
- KafkaCluster CRs can be created successfully
- KafkaCluster specs can be updated
- Multiple clusters can coexist in same namespace

These are true unit tests that don't require full reconciliation
and don't interfere with other controller tests in the suite.

Author: amuraru

.github/workflows/e2e-test.yaml

@@ -11,7 +11,7 @@ concurrency:
   cancel-in-progress: true
 
 env:
-  REPOSITORY: koperator_e2e_test  
+  REPOSITORY: koperator_e2e_test
 
 jobs:
   build:
@@ -20,7 +20,7 @@ jobs:
     steps:
       - name: Checkout code
         uses: actions/checkout@v5
-        
+
       - name: Clean workspace
         run: |
           # Remove any existing go.work files to ensure clean module resolution
@@ -35,7 +35,7 @@ jobs:
         uses: azure/setup-helm@v4
         with:
           version: '3.19.0'
-          
+
       - name: Clean Go module cache
         run: |
           # Clean any existing Go module cache to avoid conflicts
@@ -47,13 +47,17 @@ jobs:
           # Ensure proper permissions
           chmod -R 755 ~/go/pkg/mod
           chmod -R 755 ~/.cache/go-build
-          
-      # Enable Tmate Session if you'd like to Debut the E2E Kind Cluster
-      # - name: Setup tmate session
-      #   uses: mxschmitt/action-tmate@v3
-      #   with:
-      #     detached: true
-      
+
+      # Enable Tmate Session to debug the E2E Kind Cluster
+      # Only activates when re-running the workflow in debug mode
+      - name: Setup tmate session
+        if: ${{ runner.debug == '1' }}
+        continue-on-error: true
+        uses: mxschmitt/action-tmate@v3
+        with:
+          detached: true
+          limit-access-to-actor: true
+
       - name: Build docker image
         run: |
           IMG=$REPOSITORY:$GITHUB_SHA make docker-build
@@ -75,7 +79,7 @@ jobs:
           cd properties && go mod download && cd ..
           cd tests/e2e && go mod download && cd ../..
 
-          
+
       - name: Run E2E tests
         env:
            KUBECONFIG: ${{ steps.setup-kind.outputs.kubeconfig }}

README.md

@@ -84,6 +84,9 @@ metadata:
     namespace: zookeeper
 spec:
     replicas: 1
+    image:
+        repository: ghcr.io/adobe/zookeeper-operator/zookeeper
+        tag: 3.8.4-0.2.15-adobe-20250923
     persistence:
         reclaimPolicy: Delete
 EOF

api/v1beta1/common_types.go

@@ -31,7 +31,7 @@ operator:
   # the Cert-manager's Custom Resource Namespace must be included in the comma separated list.
   # When it is empty, all namespaces will be watched.
   # -- List of namespaces where Operator watches for custom resources.<br><br>**Note** that the operator still requires to read the cluster-scoped `Node` labels to configure `rack awareness`. Make sure the operator ServiceAccount is granted `get` permissions on this `Node` resource when using limited RBACs.
-  namespaces: ""
+  namespaces: "kafka, cert-manager"
   # -- Enable verbose logging
   verboseLogging: false
   # -- Enable development logging

charts/kafka-operator/values.yaml

@@ -194,6 +194,10 @@ func (r *KafkaClusterReconciler) Reconcile(ctx context.Context, request ctrl.Req
 		if err := k8sutil.UpdateRollingUpgradeState(r.Client, instance, time.Now(), log); err != nil {
 			return requeueWithError(log, err.Error(), err)
 		}
+		// Don't transition to Running state while rolling upgrade is in progress
+		// The state will be updated to Running by the kafka reconciler when the upgrade completes
+		log.Info("Rolling upgrade in progress, keeping current state")
+		return reconciled()
 	}
 
 	if err := k8sutil.UpdateCRStatus(r.Client, instance, v1beta1.KafkaClusterRunning, log); err != nil {
@@ -300,16 +304,15 @@ func (r *KafkaClusterReconciler) checkFinalizers(ctx context.Context, cluster *v
 		// user finalizations are done before it does its final cleanup
 		log.Info("Tearing down any PKI resources for the kafkacluster")
 		if err = pki.GetPKIManager(r.Client, cluster, v1beta1.PKIBackendProvided).FinalizePKI(ctx); err != nil {
-			switch err.(type) {
-			case errorfactory.ResourceNotReady:
+			// Use errors.As for consistency with the rest of the controller
+			if errors.As(err, &errorfactory.ResourceNotReady{}) {
 				log.Info("The PKI is not ready to be torn down")
 				return ctrl.Result{
 					Requeue:      true,
 					RequeueAfter: time.Duration(5) * time.Second,
 				}, nil
-			default:
-				return requeueWithError(log, "failed to finalize PKI", err)
 			}
+			return requeueWithError(log, "failed to finalize PKI", err)
 		}
 	}
 
@@ -356,6 +359,7 @@ func (r *KafkaClusterReconciler) updateAndFetchLatest(ctx context.Context, clust
 	if err != nil {
 		return nil, err
 	}
+
 	cluster.TypeMeta = typeMeta
 	return cluster, nil
 }