test/e2e: fix ClusterClass changes flake

Signed-off-by: Stefan Büringer [email protected]

Author: sbueringer

test/e2e/clusterclass_changes.go

@@ -426,7 +426,12 @@ func rebaseClusterClassAndWait(ctx context.Context, input rebaseClusterClassAndW
 	patchHelper, err := patch.NewHelper(input.Cluster, mgmtClient)
 	Expect(err).ToNot(HaveOccurred())
 	input.Cluster.Spec.Topology.Class = newClusterClassName
-	Expect(patchHelper.Patch(ctx, input.Cluster)).To(Succeed())
+	// We have to retry the patch. The ClusterClass was just created so the client cache in the
+	// controller/webhook might not be aware of it yet. If the webhook is not aware of the ClusterClass
+	// we get a "Cluster ... can't be validated. ClusterClass ... can not be retrieved" error.
+	Eventually(func() error {
+		return patchHelper.Patch(ctx, input.Cluster)
+	}, "1m", "5s").Should(Succeed(), "Failed to patch Cluster")
 
 	log.Logf("Waiting for MachineDeployment rollout to complete.")
 	for _, mdTopology := range input.Cluster.Spec.Topology.Workers.MachineDeployments {