Merge pull request #12554 from LingyanCao/dev/lingyancao/timeout

🐛  Fix timeout handling in GetAPIServerCertificateExpiry and DialContext

Author: k8s-ci-robot

controlplane/kubeadm/internal/proxy/dial.go

@@ -83,16 +83,30 @@ func (d *Dialer) DialContextWithAddr(ctx context.Context, addr string) (net.Conn
 }
 
 // DialContext creates proxied port-forwarded connections.
-// ctx is currently unused, but fulfils the type signature used by GRPC.
-func (d *Dialer) DialContext(_ context.Context, _ string, addr string) (net.Conn, error) {
+func (d *Dialer) DialContext(ctx context.Context, _ string, addr string) (net.Conn, error) {
+	// Check if context is already cancelled or timed out
+	select {
+	case <-ctx.Done():
+		return nil, errors.Wrap(ctx.Err(), "context cancelled before establishing connection")
+	default:
+	}
+
 	req := d.clientset.CoreV1().RESTClient().
 		Post().
 		Resource(d.proxy.Kind).
 		Namespace(d.proxy.Namespace).
 		Name(addr).
 		SubResource("portforward")
 
-	dialer := spdy.NewDialer(d.upgrader, &http.Client{Transport: d.proxyTransport}, "POST", req.URL())
+	httpClient := &http.Client{
+		Transport: d.proxyTransport,
+	}
+
+	if deadline, ok := ctx.Deadline(); ok {
+		httpClient.Timeout = time.Until(deadline)
+	}
+
+	dialer := spdy.NewDialer(d.upgrader, httpClient, "POST", req.URL())
 
 	// Create a new connection from the dialer.
 	//

controlplane/kubeadm/internal/workload_cluster.go

@@ -318,6 +318,10 @@ func (w *Workload) ClusterStatus(ctx context.Context) (ClusterStatus, error) {
 
 // GetAPIServerCertificateExpiry returns the certificate expiry of the apiserver on the given node.
 func (w *Workload) GetAPIServerCertificateExpiry(ctx context.Context, kubeadmConfig *bootstrapv1.KubeadmConfig, nodeName string) (*time.Time, error) {
+	// Create a context with 15 second timeout
+	ctx, cancel := context.WithTimeoutCause(ctx, 15*time.Second, errors.New("timeout getting API server certificate expiry"))
+	defer cancel()
+
 	// Create a proxy.
 	p := proxy.Proxy{
 		Kind:       "pods",