feat: adding in v4overlay windows tests

.pipelines/singletenancy/azure-cni-overlay/azure-cni-overlay-e2e-step-template.yaml

@@ -172,3 +172,41 @@ steps:
     name: "Cleanupartifactdir"
     displayName: "Cleanup artifact dir"
     condition: always()
+
+  - task: AzureCLI@2
+    inputs:
+      azureSubscription: $(BUILD_VALIDATIONS_SERVICE_CONNECTION)
+      scriptLocation: "inlineScript"
+      scriptType: "bash"
+      addSpnToEnvironment: true
+      inlineScript: |
+        set -e
+        make -C ./hack/aks windows-nodepool-up AZCLI=az SUB=$(SUB_AZURE_NETWORK_AGENT_BUILD_VALIDATIONS) CLUSTER=${{ parameters.clusterName }}
+        echo "Windows node are successfully added to v4 Overlay Cluster"
+        kubectl cluster-info
+        kubectl get node -owide
+        kubectl get po -owide -A
+    name: "Add_Windows_Node"
+    displayName: "Add windows node on v4 overlay cluster"
+
+  - script: |
+      nodeList=`kubectl get node -owide | grep Windows | awk '{print $1}'`
+      for node in $nodeList; do
+          taint=`kubectl describe node $node | grep Taints | awk '{print $2}'`
+          if [ $taint == "node.cloudprovider.kubernetes.io/uninitialized=true:NoSchedule" ]; then
+              kubectl taint nodes $node node.cloudprovider.kubernetes.io/uninitialized=true:NoSchedule-
+          fi
+      done
+      CNS=$(make cns-version) DROPGZ=$(make cni-dropgz-version)
+      sudo -E env "PATH=$PATH" make test-load SCALE_UP=32 OS_TYPE=windows CNI_TYPE=cniv2 VALIDATE_STATEFILE=true INSTALL_CNS=true INSTALL_AZURE_CNI_OVERLAY=true VALIDATE_V4OVERLAY=true CNS_VERSION=${CNS} CNI_DROPGZ_VERSION=${DROPGZ} CLEANUP=true
+    name: "WindowsOverlayControlPlaneScaleTests"
+    displayName: "Windows v4Overlay ControlPlane Scale Tests"
+    retryCountOnTaskFailure: 3
+
+  - script: |
+      echo "IPv4 Overlay DataPath Test"
+      cd test/integration/datapath
+      sudo -E env "PATH=$PATH" go test -count=1 datapath_windows_test.go -timeout 3m -tags connection -initbyocni true -run ^TestDatapathWin$
+    name: "WindowsV4OverlayDatapathTests"
+    displayName: "Windows v4Overlay Datapath Tests"
+    retryCountOnTaskFailure: 3

hack/aks/Makefile

@@ -90,7 +90,7 @@ overlay-byocni-up: rg-up overlay-net-up ## Brings up an Overlay BYO CNI cluster
 	$(AZCLI) aks create -n $(CLUSTER) -g $(GROUP) -l $(REGION) \
 		--node-count $(NODE_COUNT) \
 		--node-vm-size $(VM_SIZE) \
-		--load-balancer-sku basic \
+		--load-balancer-sku standard \
 		--network-plugin none \
 		--network-plugin-mode overlay \
 		--pod-cidr 192.168.0.0/16 \

hack/aks/README.md

@@ -35,6 +35,7 @@ AKS Clusters
   windows-cniv1-up              Bring up a Windows AzCNIv1 cluster
   linux-cniv1-up                Bring up a Linux AzCNIv1 cluster
   dualstack-overlay-byocni-up   Bring up an dualstack overlay cluster without CNS and CNI installed
+  windows-nodepool-up           Add windows node pool
   down                          Delete the cluster
   vmss-restart                  Restart the nodes of the cluster
 ```

test/integration/datapath/datapath_windows_test.go

@@ -10,21 +10,26 @@ import (
 
 	"github.com/Azure/azure-container-networking/test/internal/datapath"
 	"github.com/Azure/azure-container-networking/test/internal/kubernetes"
+	"github.com/Azure/azure-container-networking/test/validate"
 	"github.com/stretchr/testify/require"
 	apiv1 "k8s.io/api/core/v1"
 )
 
 const (
-	WindowsDeployYamlPath = "../manifests/datapath/windows-deployment.yaml"
-	podLabelKey           = "app"
-	podCount              = 2
-	nodepoolKey           = "agentpool"
+	WindowsDeployYamlPath         = "../manifests/datapath/windows-deployment.yaml"
+	IPMasqAgentYamlPath           = "../manifests/ip-masq-agent/ip-masq-agent.yaml"
+	IPMasqAgentCustomConfigMap    = "../manifests/ip-masq-agent/config-custom.yaml"
+	IPMasqAgentReconcileConfigMap = "../manifests/ip-masq-agent/config-reconcile.yaml"
+	podLabelKey                   = "app"
+	podCount                      = 2
+	nodepoolKey                   = "agentpool"
 )
 
 var (
 	podPrefix        = flag.String("podName", "datapod", "Prefix for test pods")
 	podNamespace     = flag.String("namespace", "windows-datapath-test", "Namespace for test pods")
 	nodepoolSelector = flag.String("nodepoolSelector", "npwin", "Provides nodepool as a windows Node-Selector for pods")
+	initBYOCNI       = flag.Bool("initbyocni", false, "installs ipmas agent, configmap, and restarts kubeproxy")
 )
 
 /*
@@ -48,12 +53,35 @@ Timeout context is controled by the -timeout flag.
 func setupWindowsEnvironment(t *testing.T) {
 	ctx := context.Background()
 
+	t.Log("Get REST config")
+	restConfig := kubernetes.MustGetRestConfig(t)
+
 	t.Log("Create Clientset")
 	clientset, err := kubernetes.MustGetClientset()
 	if err != nil {
 		t.Fatal(err)
 	}
 
+	if *initBYOCNI {
+		// create ipmasq configmaps
+		err := kubernetes.MustSetupConfigMap(ctx, clientset, IPMasqAgentCustomConfigMap)
+		require.NoError(t, err, "failed to setup ip masq agent custom configmap")
+		err = kubernetes.MustSetupConfigMap(ctx, clientset, IPMasqAgentReconcileConfigMap)
+		require.NoError(t, err, "failed to setup ip masq agent reconcile configmap")
+
+		// create ipmasq agent daemonset
+		ds, err := kubernetes.MustParseDaemonSet(IPMasqAgentYamlPath)
+		require.NoError(t, err, "failed to parse ipmasq agent")
+		err = kubernetes.MustCreateDaemonset(ctx, clientset.AppsV1().DaemonSets(ds.Namespace), ds)
+		require.NoError(t, err, "failed to create ipmasq daemonset")
+
+		// restart kubeproxy
+		validator, err := validate.CreateValidator(ctx, clientset, restConfig, *podNamespace, "cniv2", false, "windows")
+		require.NoError(t, err)
+		err = validator.RestartKubeProxyService(ctx)
+		require.NoError(t, err)
+	}
+
 	t.Log("Create Label Selectors")
 	podLabelSelector := kubernetes.CreateLabelSelector(podLabelKey, podPrefix)
 	nodeLabelSelector := kubernetes.CreateLabelSelector(nodepoolKey, nodepoolSelector)

test/integration/load/load_test.go

@@ -21,6 +21,7 @@ type TestConfig struct {
 	Replicas          int    `env:"REPLICAS" default:"1"`
 	ValidateStateFile bool   `env:"VALIDATE_STATEFILE" default:"false"`
 	ValidateDualStack bool   `env:"VALIDATE_DUALSTACK" default:"false"`
+	ValidateV4Overlay bool   `env:"VALIDATE_V4OVERLAY" default:"false"`
 	SkipWait          bool   `env:"SKIP_WAIT" default:"false"`
 	RestartCase       bool   `env:"RESTART_CASE" default:"false"`
 	Cleanup           bool   `env:"CLEANUP" default:"false"`
@@ -107,13 +108,19 @@ func TestLoad(t *testing.T) {
 		t.Run("Validate state file", TestValidateState)
 	}
 
+	if testConfig.ValidateV4Overlay {
+		t.Run("Validate v4overlay", TestV4OverlayProperties)
+	}
+
 	if testConfig.ValidateDualStack {
 		t.Run("Validate dualstack overlay", TestDualStackProperties)
 	}
 
 	if testConfig.Cleanup {
 		err = kubernetes.MustDeleteDeployment(ctx, deploymentsClient, deployment)
 		require.NoError(t, err, "error deleteing load deployment")
+		err = kubernetes.WaitForPodsDelete(ctx, clientset, namespace, podLabelSelector)
+		require.NoError(t, err, "error waiting for pods to delete")
 	}
 }
 
@@ -171,9 +178,94 @@ func TestScaleDeployment(t *testing.T) {
 	if testConfig.Cleanup {
 		err = kubernetes.MustDeleteDeployment(ctx, deploymentsClient, deployment)
 		require.NoError(t, err, "error deleteing load deployment")
+		err = kubernetes.WaitForPodsDelete(ctx, clientset, namespace, podLabelSelector)
+		require.NoError(t, err, "error waiting for pods to delete")
 	}
 }
 
+func TestValidCNSStateDuringScaleAndCNSRestart(t *testing.T) {
+	clientset, err := kubernetes.MustGetClientset()
+	require.NoError(t, err)
+
+	config := kubernetes.MustGetRestConfig(t)
+	ctx, cancel := context.WithTimeout(context.Background(), 30*time.Minute)
+	defer cancel()
+
+	validator, err := validate.CreateValidator(ctx, clientset, config, namespace, testConfig.CNIType, testConfig.RestartCase, testConfig.OSType)
+	require.NoError(t, err)
+
+	err = validator.Validate(ctx)
+	require.NoError(t, err)
+
+	deployment, err := kubernetes.MustParseDeployment(noopDeploymentMap[testConfig.OSType])
+	require.NoError(t, err)
+	deploymentsClient := clientset.AppsV1().Deployments(namespace)
+
+	// Create a deployment
+	err = kubernetes.MustCreateDeployment(ctx, deploymentsClient, deployment)
+	require.NoError(t, err)
+
+	// Scale it up and "skipWait", so CNS restart can happen immediately after scale call is made (while pods are still creating)
+	skipWait := true
+	err = kubernetes.MustScaleDeployment(ctx, deploymentsClient, deployment, clientset, namespace, podLabelSelector, testConfig.ScaleUpReplicas, skipWait)
+	require.NoError(t, err)
+
+	// restart linux CNS (linux, windows)
+	err = kubernetes.RestartCNSDaemonset(ctx, clientset)
+	require.NoError(t, err)
+
+	// wait for pods to settle before checking cns state (otherwise, race between getting pods in creating state, and getting CNS state file)
+	err = kubernetes.WaitForPodDeployment(ctx, clientset, namespace, deployment.Name, podLabelSelector, testConfig.ScaleUpReplicas)
+	require.NoError(t, err)
+
+	// Validate the CNS state
+	err = validator.Validate(ctx)
+	require.NoError(t, err)
+
+	// Scale it down
+	err = kubernetes.MustScaleDeployment(ctx, deploymentsClient, deployment, clientset, namespace, podLabelSelector, testConfig.ScaleDownReplicas, skipWait)
+	require.NoError(t, err)
+
+	// restart linux CNS (linux, windows)
+	err = kubernetes.RestartCNSDaemonset(ctx, clientset)
+	require.NoError(t, err)
+
+	// wait for pods to settle before checking cns state (otherwise, race between getting pods in terminating state, and getting CNS state file)
+	err = kubernetes.WaitForPodDeployment(ctx, clientset, namespace, deployment.Name, podLabelSelector, testConfig.ScaleDownReplicas)
+	require.NoError(t, err)
+
+	// Validate the CNS state
+	err = validator.Validate(ctx)
+	require.NoError(t, err)
+
+	if testConfig.Cleanup {
+		err = kubernetes.MustDeleteDeployment(ctx, deploymentsClient, deployment)
+		require.NoError(t, err, "error deleteing load deployment")
+		err = kubernetes.WaitForPodsDelete(ctx, clientset, namespace, podLabelSelector)
+		require.NoError(t, err, "error waiting for pods to delete")
+	}
+}
+
+func TestV4OverlayProperties(t *testing.T) {
+	if !testConfig.ValidateV4Overlay {
+		return
+	}
+	clientset, err := kubernetes.MustGetClientset()
+	require.NoError(t, err)
+
+	config := kubernetes.MustGetRestConfig(t)
+	ctx, cancel := context.WithTimeout(context.Background(), 30*time.Minute)
+	defer cancel()
+
+	validator, err := validate.CreateValidator(ctx, clientset, config, namespace, testConfig.CNIType, testConfig.RestartCase, testConfig.OSType)
+	require.NoError(t, err)
+
+	// validate IPv4 overlay scenarios
+	t.Log("Validating v4Overlay node labels")
+	err = validator.ValidateV4OverlayControlPlane(ctx)
+	require.NoError(t, err)
+}
+
 func TestDualStackProperties(t *testing.T) {
 	if !testConfig.ValidateDualStack {
 		return

test/internal/kubernetes/utils.go

@@ -233,6 +233,24 @@ func WaitForPodsRunning(ctx context.Context, clientset *kubernetes.Clientset, na
 	return errors.Wrap(retrier.Do(ctx, checkPodIPsFn), "failed to check if pods were running")
 }
 
+func WaitForPodsDelete(ctx context.Context, clientset *kubernetes.Clientset, namespace, labelselector string) error {
+	podsClient := clientset.CoreV1().Pods(namespace)
+
+	checkPodsDeleted := func() error {
+		podList, err := podsClient.List(ctx, metav1.ListOptions{LabelSelector: labelselector})
+		if err != nil {
+			return errors.Wrapf(err, "could not list pods with label selector %s", labelselector)
+		}
+		if len(podList.Items) != 0 {
+			return errors.Errorf("pods still present: %+v", podList.Items)
+		}
+		return nil
+	}
+
+	retrier := retry.Retrier{Attempts: RetryAttempts, Delay: RetryDelay}
+	return errors.Wrap(retrier.Do(ctx, checkPodsDeleted), "failed to wait for pods to delete")
+}
+
 func WaitForPodDeployment(ctx context.Context, clientset *kubernetes.Clientset, namespace, deploymentName, podLabelSelector string, replicas int) error {
 	podsClient := clientset.CoreV1().Pods(namespace)
 	deploymentsClient := clientset.AppsV1().Deployments(namespace)
@@ -431,3 +449,19 @@ func HasWindowsNodes(ctx context.Context, clientset *kubernetes.Clientset) (bool
 	}
 	return false, nil
 }
+
+func MustRestartDaemonset(ctx context.Context, clientset *kubernetes.Clientset, namespace, daemonsetName string) error {
+	ds, err := clientset.AppsV1().DaemonSets(namespace).Get(ctx, daemonsetName, metav1.GetOptions{})
+	if err != nil {
+		return errors.Wrapf(err, "failed to get daemonset %s", daemonsetName)
+	}
+
+	if ds.Spec.Template.ObjectMeta.Annotations == nil {
+		ds.Spec.Template.ObjectMeta.Annotations = make(map[string]string)
+	}
+
+	ds.Spec.Template.ObjectMeta.Annotations["kubectl.kubernetes.io/restartedAt"] = time.Now().Format(time.RFC3339)
+
+	_, err = clientset.AppsV1().DaemonSets(namespace).Update(ctx, ds, metav1.UpdateOptions{})
+	return errors.Wrapf(err, "failed to update ds %s", daemonsetName)
+}