tests: drop hardcoded kube-system CNI (#50667)

* tests: drop hardcoded kube-system CNI

This was originally set in 2021 to make the tests run on GKE/others.
Long ago, the installation automatically detects the kube-system
requirements and installs appropriately.

As-is, this means if you install the integ test on a running cluster off
GKE, it will end up with 2 CNIs and cause havoc

* fix kube-system

* skip multi

pkg/test/framework/components/istio/kube.go

@@ -620,7 +620,6 @@ func commonInstallArgs(ctx resource.Context, cfg Config, c cluster.Cluster, defa
 
 	// Include all user-specified values and configuration options.
 	if cfg.EnableCNI {
-		args.AppendSet("components.cni.namespace", "kube-system")
 		args.AppendSet("components.cni.enabled", "true")
 	}
 

tests/integration/pilot/cni_race_test.go

@@ -86,7 +86,7 @@ func TestCNIRaceRepair(t *testing.T) {
 
 func getCNIDaemonSet(ctx framework.TestContext, c cluster.Cluster) *appsv1.DaemonSet {
 	cniDaemonSet, err := c.(istioKube.CLIClient).
-		Kube().AppsV1().DaemonSets("kube-system").
+		Kube().AppsV1().DaemonSets(i.Settings().SystemNamespace).
 		Get(context.Background(), "istio-cni-node", metav1.GetOptions{})
 	if err != nil {
 		ctx.Fatalf("failed to get CNI Daemonset %v", err)
@@ -99,15 +99,15 @@ func getCNIDaemonSet(ctx framework.TestContext, c cluster.Cluster) *appsv1.Daemo
 
 func deleteCNIDaemonset(ctx framework.TestContext, c cluster.Cluster) {
 	if err := c.(istioKube.CLIClient).
-		Kube().AppsV1().DaemonSets("kube-system").
+		Kube().AppsV1().DaemonSets(i.Settings().SystemNamespace).
 		Delete(context.Background(), "istio-cni-node", metav1.DeleteOptions{}); err != nil {
 		ctx.Fatalf("failed to delete CNI Daemonset %v", err)
 	}
 
 	// Wait until the CNI Daemonset pod cannot be fetched anymore
 	retry.UntilSuccessOrFail(ctx, func() error {
 		scopes.Framework.Infof("Checking if CNI Daemonset pods are deleted...")
-		pods, err := c.PodsForSelector(context.TODO(), "kube-system", "k8s-app=istio-cni-node")
+		pods, err := c.PodsForSelector(context.TODO(), i.Settings().SystemNamespace, "k8s-app=istio-cni-node")
 		if err != nil {
 			return err
 		}
@@ -127,7 +127,7 @@ func deployCNIDaemonset(ctx framework.TestContext, c cluster.Cluster, cniDaemonS
 		Labels:      cniDaemonSet.ObjectMeta.Labels,
 		Annotations: cniDaemonSet.ObjectMeta.Annotations,
 	}
-	_, err := c.(istioKube.CLIClient).Kube().AppsV1().DaemonSets("kube-system").
+	_, err := c.(istioKube.CLIClient).Kube().AppsV1().DaemonSets(i.Settings().SystemNamespace).
 		Create(context.Background(), &deployDaemonSet, metav1.CreateOptions{})
 	if err != nil {
 		ctx.Fatalf("failed to deploy CNI Daemonset %v", err)

tests/integration/pilot/multiplecontrolplanes/main_test.go

@@ -58,6 +58,12 @@ func TestMain(m *testing.M) {
 		RequireMultiPrimary().
 		// Requires two CPs with specific names to be configured.
 		Label(label.CustomSetup).
+		// We are deploying two isolated environments, which CNI doesn't support.
+		// We could deploy one of the usergroups as the CNI owner, but for now we skip
+		SkipIf("CNI is not suppored", func(ctx resource.Context) bool {
+			c, _ := istio.DefaultConfig(ctx)
+			return c.EnableCNI
+		}).
 		SetupParallel(
 			namespace.Setup(&userGroup1NS, namespace.Config{Prefix: "usergroup-1", Labels: map[string]string{"usergroup": "usergroup-1"}}),
 			namespace.Setup(&userGroup2NS, namespace.Config{Prefix: "usergroup-2", Labels: map[string]string{"usergroup": "usergroup-2"}})).