fix: add CSI Node verification to avoid race

Signed-off-by: Jakob Möller <jmoller@redhat.com>

cmd/vgmanager/vgmanager.go

@@ -213,7 +213,7 @@ func run(cmd *cobra.Command, _ []string, opts *Options) error {
 		return fmt.Errorf("unable to create controller VGManager: %w", err)
 	}
 
-	if err := mgr.AddReadyzCheck("readyz", readyCheck(mgr, lvmdConfig)); err != nil {
+	if err := mgr.AddReadyzCheck("readyz", readyCheck(mgr)); err != nil {
 		return fmt.Errorf("unable to set up ready check: %w", err)
 	}
 
@@ -334,7 +334,7 @@ func pluginRegistrationSocketPath() string {
 }
 
 // readyCheck returns a healthz.Checker that verifies the operator is ready
-func readyCheck(mgr manager.Manager, config *lvmd.Config) healthz.Checker {
+func readyCheck(mgr manager.Manager) healthz.Checker {
 	return func(req *http.Request) error {
 		// Perform various checks here to determine if the operator is ready
 		if !mgr.GetCache().WaitForCacheSync(req.Context()) {

test/e2e/helper_test.go

@@ -57,6 +57,7 @@ func VerifyLVMSSetup(ctx context.Context, cluster *v1alpha1.LVMCluster) {
 	GinkgoHelper()
 	validateLVMCluster(ctx, cluster)
 	validateCSIDriver(ctx)
+	validateCSINodeInfo(ctx, cluster, true)
 	validateVGManager(ctx)
 	validateLVMVolumeGroup(ctx)
 	validateStorageClass(ctx)

test/e2e/lvm_cluster_test.go

@@ -40,6 +40,7 @@ func lvmClusterTest() {
 			return
 		}
 		DeleteResource(ctx, cluster)
+		validateCSINodeInfo(ctx, cluster, false)
 	})
 
 	Describe("Filesystem Type", Serial, func() {

test/e2e/lvm_pvc_test.go

@@ -77,6 +77,7 @@ func pvcTestThinProvisioning() {
 				return
 			}
 			DeleteResource(ctx, cluster)
+			validateCSINodeInfo(ctx, cluster, false)
 		})
 	})
 
@@ -125,6 +126,7 @@ func pvcTestThickProvisioning() {
 				return
 			}
 			DeleteResource(ctx, cluster)
+			validateCSINodeInfo(ctx, cluster, false)
 		})
 	})
 

test/e2e/validation_test.go

@@ -27,6 +27,7 @@ import (
 	. "github.com/onsi/gomega"
 	"github.com/openshift/lvm-operator/api/v1alpha1"
 	"github.com/openshift/lvm-operator/internal/controllers/lvmcluster/resource"
+	"github.com/openshift/lvm-operator/internal/controllers/lvmcluster/selector"
 	topolvmv1 "github.com/topolvm/topolvm/api/v1"
 	appsv1 "k8s.io/api/apps/v1"
 	k8sv1 "k8s.io/api/core/v1"
@@ -174,6 +175,50 @@ func validatePodData(ctx context.Context, pod *k8sv1.Pod, expectedData string, c
 	return Expect(actualData).To(Equal(expectedData))
 }
 
+func validateCSINodeInfo(ctx context.Context, lvmCluster *v1alpha1.LVMCluster, shouldBePresent bool) bool {
+	GinkgoHelper()
+	By("validating the CSINode(s) for the Cluster to have the driver registered")
+	return Eventually(func(ctx context.Context) error {
+		cluster := lvmCluster
+		if err := crClient.Get(ctx, client.ObjectKeyFromObject(cluster), cluster); err != nil {
+			return fmt.Errorf("failed to get LVMCluster: %v", err)
+		}
+
+		nodeList := &k8sv1.NodeList{}
+		if err := crClient.List(ctx, nodeList); err != nil {
+			return fmt.Errorf("failed to list Nodes: %v", err)
+		}
+
+		nodes, err := selector.ValidNodes(cluster, nodeList)
+		if err != nil {
+			return fmt.Errorf("failed to get valid nodes: %v", err)
+		}
+
+		for _, node := range nodes {
+			var csiNode storagev1.CSINode
+			if err := crClient.Get(ctx, client.ObjectKey{Name: node.Name}, &csiNode); err != nil {
+				return fmt.Errorf("failed to get CSINode %q: %v", node.Name, err)
+			}
+			exists := false
+			for _, d := range csiNode.Spec.Drivers {
+				if d.Name == csiDriverName {
+					exists = true
+					break
+				}
+			}
+			if exists != shouldBePresent {
+				return fmt.Errorf("CSINode %q should %scontain the driver %q",
+					node.Name,
+					map[bool]string{true: "", false: "not "}[shouldBePresent],
+					csiDriverName,
+				)
+			}
+		}
+
+		return nil
+	}, timeout, interval).WithContext(ctx).Should(Succeed())
+}
+
 func SummaryOnFailure(ctx context.Context) {
 	if !CurrentSpecReport().Failed() {
 		GinkgoLogr.Info("skipping test namespace summary due to successful test run")