e2e-tests (2/4): Add parallel cluster-based e2e tests via gingko

Dockerfile

@@ -16,6 +16,7 @@ WORKDIR /workspace
 RUN --mount=type=cache,target=/go/pkg/mod \
     --mount=type=bind,source=go.mod,target=go.mod \
     --mount=type=bind,source=go.sum,target=go.sum \
+    --mount=type=bind,source=test/gnmi,target=test/gnmi \
     go mod download -x
 
 RUN --mount=type=bind,target=. \

Makefile

@@ -50,6 +50,9 @@ $(LOCALBIN):
 install-gofumpt: FORCE
 	@if ! hash gofumpt 2>/dev/null; then printf "\e[1;36m>> Installing gofumpt...\e[0m\n"; go install mvdan.cc/gofumpt@latest; fi
 
+install-ginkgo: FORCE
+	@if ! hash ginkgo 2>/dev/null; then printf "\e[1;36m>> Installing ginkgo...\e[0m\n"; go install github.com/onsi/ginkgo/v2/ginkgo@latest; fi
+
 install-kubebuilder: FORCE
 	@set -eou pipefail;  if ! hash kubebuilder 2>/dev/null; then printf "\e[1;36m>> Installing kubebuilder...\e[0m\n"; if command -v curl >/dev/null 2>&1; then GET="curl -sLo"; elif command -v wget >/dev/null 2>&1; then GET="wget -O"; else echo "Didn't find curl or wget to download kubebuilder"; exit 2; fi; BIN=$$(go env GOBIN); if [[ -z $$BIN ]]; then BIN=$$(go env GOPATH)/bin; fi; $$GET "$$BIN/kubebuilder" "https://go.kubebuilder.io/dl/latest/$$(go env GOOS)/$$(go env GOARCH)"; chmod +x "$$BIN/kubebuilder"; fi
 
@@ -68,11 +71,15 @@ lint: FORCE bin/golangci-lint-custom ## Run golangci-lint linter
 	@bin/golangci-lint-custom config verify
 	@bin/golangci-lint-custom run
 
+# PROVIDER defines which provider to test (cisco-nxos-gnmi, cisco-iosxr-gnmi, openconfig).
+# Used by test-e2e-cluster and test-e2e-envtest to filter tests.
+PROVIDER ?= cisco-nxos-gnmi
+
 fmt: FORCE install-gofumpt
 	@printf "\e[1;36m>> gofumpt -l -w .\e[0m\n"
 	@gofumpt -l -w $(shell git ls-files '*.go' | grep -v '^internal/provider/openconfig')
 
-# Run the e2e tests against a k8s cluster.
+# Run the scaffolded e2e tests (unchanged from Kubebuilder).
 test-e2e: FORCE
 	@command -v kind >/dev/null 2>&1 || { \
 	  echo "Kind is not installed. Please install Kind manually."; \
@@ -85,6 +92,26 @@ test-e2e: FORCE
 	@printf "\e[1;36m>> go test ./test/e2e/ -v -ginkgo.v\e[0m\n"
 	@KIND_CLUSTER=$(KIND_CLUSTER) go test ./test/e2e/ -v -ginkgo.v
 
+# Run gNMI controller tests in cluster mode (requires Kind cluster).
+# Uses ginkgo for parallel execution.
+GINKGO_PROCS ?= 4
+test-e2e-cluster: FORCE install-ginkgo
+	@command -v kind >/dev/null 2>&1 || { \
+	  echo "Kind is not installed. Please install Kind manually."; \
+	  exit 1; \
+	}
+	@kind get clusters | grep -q $(KIND_CLUSTER) || { \
+	  echo "No Kind cluster is running. Please start a Kind cluster before running the e2e tests."; \
+	  exit 1; \
+	}
+	@printf "\e[1;36m>> ginkgo -procs=$(GINKGO_PROCS) -tags=cluster -timeout=15m -v ./test/e2e/ (PROVIDER=$(PROVIDER))\e[0m\n"
+	@KIND_CLUSTER=$(KIND_CLUSTER) E2E_PROVIDER=$(PROVIDER) ginkgo -procs=$(GINKGO_PROCS) -tags=cluster -timeout=15m -v ./test/e2e/
+
+# Run gNMI controller tests in envtest mode (no cluster required).
+test-e2e-envtest: FORCE install-setup-envtest
+	@printf "\e[1;36m>> go test ./test/e2e/ -tags=envtest -v -ginkgo.v (PROVIDER=$(PROVIDER))\e[0m\n"
+	@KUBEBUILDER_ASSETS=$$(setup-envtest use 1.32 -p path) E2E_PROVIDER=$(PROVIDER) go test ./test/e2e/ -tags=envtest -v -ginkgo.v
+
 docker-build: FORCE
 	@printf "\e[1;36m>> $(CONTAINER_TOOL) build --tag=$(IMG) .\e[0m\n"
 	@$(CONTAINER_TOOL) build --build-arg=BININFO_BUILD_DATE=$(BININFO_BUILD_DATE) --build-arg=BININFO_COMMIT_HASH=$(BININFO_COMMIT_HASH) --build-arg=BININFO_VERSION=$(BININFO_VERSION) --tag=$(IMG) .
@@ -98,15 +125,16 @@ build-installer: FORCE generate install-kustomize
 	@printf "\e[1;36m>> kustomize build config/default > dist/install.yaml\e[0m\n"
 	@mkdir -p dist; kustomize build config/default > dist/install.yaml
 
-# Deploy controller to the k8s cluster
+# Deploy controller to the k8s cluster.
+# Use PROVIDER to set the provider (default: cisco-nxos-gnmi).
 deploy: FORCE generate install-kustomize
-	@printf "\e[1;36m>> kustomize build config/default | kubectl apply -f -\e[0m\n"
-	@kustomize build config/default | kubectl apply -f -
+	@printf "\e[1;36m>> deploying controller-manager (PROVIDER=$(PROVIDER))\e[0m\n"
+	@kustomize build config/develop | sed 's/--provider=openconfig/--provider=$(PROVIDER)/' | kubectl apply -f -
 
 # Undeploy controller from the k8s cluster
 undeploy: FORCE install-kustomize
-	@printf "\e[1;36m>> kustomize build config/default | kubectl delete -f -\e[0m\n"
-	@kustomize build config/default | kubectl delete --ignore-not-found=true -f -
+	@printf "\e[1;36m>> undeploying controller-manager\e[0m\n"
+	@kustomize build config/develop | kubectl delete --ignore-not-found=true -f -
 
 # Install CRDs into the k8s cluster
 deploy-crds: FORCE generate install-kustomize

config/develop/manager_patch.yaml

@@ -1,6 +1,7 @@
 - op: replace
   path: /spec/template/spec/containers/0/args
   value:
+    - --metrics-bind-address=:8443
     - --leader-elect=false
     - --health-probe-bind-address=:8081
     - --provider=openconfig

test/e2e/cluster_suite_test.go

@@ -0,0 +1,99 @@
+// SPDX-FileCopyrightText: 2025 SAP SE or an SAP affiliate company and IronCore contributors
+// SPDX-License-Identifier: Apache-2.0
+
+//go:build cluster
+
+package e2e
+
+import (
+	"fmt"
+	"testing"
+	"time"
+
+	. "github.com/onsi/ginkgo/v2"
+	. "github.com/onsi/gomega"
+	logf "sigs.k8s.io/controller-runtime/pkg/log"
+	"sigs.k8s.io/controller-runtime/pkg/log/zap"
+
+	"github.com/ironcore-dev/network-operator/test/e2e/testutil"
+)
+
+// TestCluster runs the e2e test suite in cluster mode.
+// Named differently from TestE2E to avoid conflict with scaffolded e2e_suite_test.go.
+func TestCluster(t *testing.T) {
+	RegisterFailHandler(Fail)
+	_, _ = fmt.Fprintf(GinkgoWriter, "Starting network-operator tests in CLUSTER mode\n")
+	RunSpecs(t, "e2e suite (cluster)")
+}
+
+// SynchronizedBeforeSuite enables parallel test execution:
+// - Process 1: Builds images, installs Prometheus/CertManager, deploys manager (runs first, alone)
+// - All processes: Create ClusterEnvironment connection (runs after process 1 completes)
+var _ = SynchronizedBeforeSuite(
+	// First function: runs ONLY on process 1, before other processes start
+	func(ctx SpecContext) []byte {
+		logf.SetLogger(zap.New(zap.WriteTo(GinkgoWriter), zap.UseDevMode(true)))
+		// Note: Timeout is set in the second function for all processes
+
+		By("Ensure that Prometheus is enabled")
+		cwd, err := testutil.GetProjectDir()
+		Expect(err).NotTo(HaveOccurred(), "Failed to get project directory")
+
+		err = testutil.UncommentCode(cwd+"/config/default/kustomization.yaml", "#- ../prometheus", "#")
+		Expect(err).NotTo(HaveOccurred(), "Failed to enable Prometheus")
+
+		// Build and load images to Kind (only process 1)
+		buildAndLoadImages(ctx)
+
+		// Setup Prometheus and CertManager (only process 1)
+		setupClusterDependencies(ctx)
+
+		// Deploy controller-manager (includes CRDs via make deploy)
+		By("deploying controller-manager")
+		tmpEnv := testutil.NewClusterEnvironment()
+		Expect(tmpEnv.Setup(ctx)).To(Succeed())
+		Expect(tmpEnv.DeployManager(ctx)).To(Succeed())
+
+		return nil // No data to pass to other processes
+	},
+	// Second function: runs on ALL processes after the first function completes
+	func(ctx SpecContext, _ []byte) {
+		SetDefaultEventuallyTimeout(testutil.DefaultTimeout)
+		SetDefaultEventuallyPollingInterval(time.Second)
+
+		// All processes create their own ClusterEnvironment connection
+		By("initializing cluster environment")
+		testEnv = testutil.NewClusterEnvironment()
+		Expect(testEnv.Setup(ctx)).To(Succeed())
+	},
+)
+
+// SynchronizedAfterSuite enables parallel test cleanup:
+// - All processes: Local cleanup (runs on all processes)
+// - Process 1: Uninstall shared dependencies (runs last, alone)
+var _ = SynchronizedAfterSuite(
+	// First function: runs on ALL processes
+	func(ctx SpecContext) {
+		// Perform local cleanup (will run only once even if called from signal handler)
+		performCleanup()
+
+		// Wait for all test namespaces to be fully deleted before returning.
+		// This ensures DeferCleanup hooks have finished deleting resources and their
+		// finalizers have been processed by the controller. Without this, the second
+		// function (UndeployManager) may delete the CRDs while resources still exist,
+		// causing finalizers to be stuck forever.
+		if testEnv != nil {
+			_ = testEnv.WaitForTestNamespacesGone(ctx) //nolint:errcheck // best-effort cleanup
+		}
+	},
+	// Second function: runs ONLY on process 1, after all other processes complete
+	func(ctx SpecContext) {
+		// Undeploy the controller-manager
+		tmpEnv := testutil.NewClusterEnvironment()
+		_ = tmpEnv.Setup(ctx)           //nolint:errcheck // best-effort cleanup
+		_ = tmpEnv.UndeployManager(ctx) //nolint:errcheck // best-effort cleanup
+
+		// Uninstall Prometheus and CertManager
+		cleanupClusterDependencies(ctx)
+	},
+)