zhisheng17/flink-learning · error · RuntimeException

Could not get the rest endpoint of ${clusterId}

Error message

Could not get the rest endpoint of ${clusterId}

What it means

Thrown by KubernetesClusterDescriptor.createClusterClientProvider when the FlinkKubeClient returns no REST endpoint for the given clusterId, meaning the JobManager service exists but did not expose a usable REST endpoint. It wraps a ClusterRetrieveException inside a RuntimeException, aborting retrieval of a client for an existing cluster.

Source

Thrown at flink-learning-k8s/flink-k8s/src/main/java/org/apache/flink/kubernetes/KubernetesClusterDescriptor.java:98

			"ClusterId must be specified!");
	}

	@Override
	public String getClusterDescription() {
		return CLUSTER_DESCRIPTION;
	}

	private ClusterClientProvider<String> createClusterClientProvider(String clusterId) {
		return () -> {
			final Configuration configuration = new Configuration(flinkConfig);

			final Optional<Endpoint> restEndpoint = client.getRestEndpoint(clusterId);

			if (restEndpoint.isPresent()) {
				configuration.setString(RestOptions.ADDRESS, restEndpoint.get().getAddress());
				configuration.setInteger(RestOptions.PORT, restEndpoint.get().getPort());
			} else {
				throw new RuntimeException(
						new ClusterRetrieveException(
								"Could not get the rest endpoint of " + clusterId));
			}

			try {
				// Flink client will always use Kubernetes service to contact with jobmanager. So we have a pre-configured web
				// monitor address. Using StandaloneClientHAServices to create RestClusterClient is reasonable.
				return new RestClusterClient<>(
					configuration,
					clusterId,
					new StandaloneClientHAServices(getWebMonitorAddress(configuration)));
			} catch (Exception e) {
				client.handleException(e);
				throw new RuntimeException(new ClusterRetrieveException("Could not create the RestClusterClient.", e));
			}
		};
	}

View on GitHub (pinned to d731cee761)

Solutions

1. Wait for the JobManager pod and its REST service endpoints to become Ready before calling getRestEndpoint (kubectl get endpoints <clusterId>).
2. Verify the clusterId matches an actually running Flink cluster (kubectl get svc | grep <clusterId>).
3. Check JobManager pod logs/events for startup crashes and fix the underlying deployment failure.
4. Inspect Kubernetes service labels/selectors — if the selector does not match the JobManager pod the service will have no endpoints.
5. Increase timeout/retry logic around cluster retrieval in your launcher.

Example fix

// before: retrieve immediately after deploy
ClusterClient<String> client = descriptor.deploySessionCluster(spec).getClusterClient();
// after: wait for REST service endpoints first
await().atMost(2, MINUTES).until(() ->
    kubeClient.getService(KubernetesService.ServiceType.REST_SERVICE, clusterId).isPresent()
    && kubectlEndpointsReady(clusterId));
ClusterClient<String> client = provider.getClusterClient();

// verify REST service has endpoints before retrieval
Optional<KubernetesService> svc = kubeClient.getService(KubernetesService.ServiceType.REST_SERVICE, clusterId);
if (svc.isEmpty()) throw new IllegalStateException("REST service for " + clusterId + " missing");

try {
    ClusterClient<String> client = provider.getClusterClient();
} catch (RuntimeException e) {
    if (e.getCause() instanceof ClusterRetrieveException) {
        // back off and retry until JobManager endpoints are Ready
    }
}

When it happens

Trigger: Calling clusterClientProvider or deployClusterInternal after deployment when the Kubernetes REST service for clusterId was not found or had no endpoint addresses (e.g. service exists but JobManager pod not ready/no endpoints).

Common situations: Retrieving a cluster immediately after deploying before the JobManager pod becomes Ready; JobManager crashed after service creation; kube-proxy/DNS issues leaving the service with no endpoints; wrong clusterId pointing at a stale service.

Understand the failure class

Background: 'Could not be found', 'does not exist', 'not found in database': the resource-not-found family when an ID, slug, key, or URI lookup comes back empty — this error's family across 20 libraries.

Related errors

• Could not create the RestClusterClient.
• The Flink cluster ${clusterId} already exists.
• Couldn't deploy Kubernetes Application Cluster. Expected dep
• Per-Job Mode not supported by Active Kubernetes deployments.
• Could not create Kubernetes cluster "${clusterId}".

AI-assisted analysis of zhisheng17/flink-learning@d731cee761 (2026-09-06). Data as JSON: /api/errors/53292dc8c4b95f2e. Report an issue: GitHub.