fix: wait for CoreDNS to be fully ready before running e2e tests

The previous check only verified CoreDNS pods reached the Running phase,
but did not wait for readiness probes to pass or for kube-dns endpoints
to be registered. This caused flaky DNS resolution failures for
host.minikube.internal in e2e tests.

Author: ReuDa

go/action_kit_test/e2e/minikube.go

@@ -135,6 +135,58 @@ func (m *Minikube) waitForDefaultServiceAccount() error {
 	}
 }
 
+func (m *Minikube) waitForDNSReady(timeout time.Duration) error {
+	ctx, cancel := context.WithTimeout(context.Background(), timeout)
+	defer cancel()
+
+	start := time.Now()
+
+	// First, wait for CoreDNS pods to be Running
+	pods, err := m.WaitForDeploymentPhase(
+		&metav1.ObjectMeta{Name: "coredns", Namespace: "kube-system"},
+		corev1.PodRunning,
+		"k8s-app=kube-dns",
+		timeout,
+	)
+	if err != nil {
+		return fmt.Errorf("coredns pods not running: %w", err)
+	}
+
+	// Then wait for CoreDNS pods to be Ready (readiness probe passed)
+	remaining := timeout - time.Since(start)
+	if remaining <= 0 {
+		remaining = 10 * time.Second
+	}
+	for _, pod := range pods {
+		if err := m.WaitForPodReady(pod.GetObjectMeta(), remaining); err != nil {
+			return fmt.Errorf("coredns pod %s not ready: %w", pod.Name, err)
+		}
+	}
+
+	// Finally, wait for kube-dns service to have ready endpoints.
+	// This ensures kube-proxy has updated iptables rules to route DNS traffic.
+	for {
+		select {
+		case <-ctx.Done():
+			return fmt.Errorf("kube-dns endpoints not ready within %s", timeout)
+		case <-time.After(500 * time.Millisecond):
+			endpoints, epErr := m.GetClient().CoreV1().Endpoints("kube-system").Get(context.Background(), "kube-dns", metav1.GetOptions{})
+			if epErr != nil {
+				continue
+			}
+			for _, subset := range endpoints.Subsets {
+				if len(subset.Addresses) > 0 {
+					log.Info().
+						Int("endpoints", len(subset.Addresses)).
+						TimeDiff("duration", time.Now(), start).
+						Msg("kube-dns is ready")
+					return nil
+				}
+			}
+		}
+	}
+}
+
 func (m *Minikube) delete() error {
 	globalMinikubeMutex.Lock()
 	defer globalMinikubeMutex.Unlock()
@@ -263,8 +315,8 @@ func WithMinikube(t *testing.T, mOpts MinikubeOpts, extFactory ExtensionFactory,
 				}
 			}
 
-			if _, dnsErr := minikube.WaitForDeploymentPhase(&metav1.ObjectMeta{Name: "coredns", Namespace: "kube-system"}, corev1.PodRunning, "k8s-app=kube-dns", 1*time.Minute); dnsErr != nil {
-				log.Warn().Err(dnsErr).Msg("coredns not started withing 1 minute.")
+			if dnsErr := minikube.waitForDNSReady(2 * time.Minute); dnsErr != nil {
+				t.Fatal("DNS not ready:", dnsErr)
 			}
 
 			wg.Wait()