LakeSail PlatformPREVIEW

Appearance

The managed Apache Flink service on the LakeSail platform is deprecated.

LakeSail is building Sail, an open-source computation framework in Rust to seamlessly integrate stream-processing, batch-processing, and compute-intensive (AI) workloads. The LakeSail platform will offer the managed solution for Sail. Existing PySpark and Flink SQL workloads can be migrated with ease. Please stay tuned and contact us if you are interested!

Troubleshooting ​

The Flink application is stuck in the PENDING state.

Run kubectl get pods to check the status of the Flink pods.

  1. If you can find the corresponding Flink application pod with name prefix lakesail-flinkapplication-, it means the application has been deployed but has errors. The lowercased application ID will be part of the pod name if the application pod exists.
    1. Run kubectl describe pod/<pod-name> to check the pod events.
    2. If you see errors in the init containers, run kubectl logs pod/<pod-name> -c artifact-loader to check the logs of the init containers. It typically indicates that the artifact (JAR files or plugins) is not found in the specified location, or there is an issue with the file system configuration.
    3. Otherwise, run kubectl logs pod/<pod-name> to check the pod logs. You should find errors in the logs when starting the Flink application. This will cause the pod to restart repeatedly.
  2. If you cannot find the corresponding Flink application pod, run kubectl logs pod/<operator-pod-name> to check the logs of the Flink Kubernetes operator pod. The pod name has the prefix flink-kubernetes-operator-. You may find reconciliation errors in the logs, which typically indicate configuration errors of the Flink application.

    INFO

    The Flink Kubernetes operator pod log may be very long. You can pipe the log to less -R to view the logs (with colored output), or use tail to only show the last few lines.

The Flink session is stuck in the PENDING state.

Follow the steps similar to the Flink application troubleshooting guide above. Look for the Flink session pod with name prefix lakesail-flinksession- and check the pod events and logs. The lowercased session ID will be part of the pod name.

The Flink session job is stuck in the PENDING state.

Follow the steps similar to the Flink application troubleshooting guide above. Look for the Flink session job pod with name prefix lakesail-flinksessionjob- and check the pod events and logs. The lowercased session job ID will be part of the pod name.

The Flink application or session job is in the FAILED state.

This happens when the Flink job started successfully but failed during execution. You can find the error message in the pod logs, or in the Flink UI.

INFO

Please refer to Accessing the Flink Web UI for how to access the Flink UI for Flink applications and sessions.

The Flink session cannot be deleted.

Please make sure to delete all Flink session jobs before deleting the Flink session.