SEZ9 commented on issue #11679: URL: https://github.com/apache/seatunnel/issues/11679#issuecomment-5230720436
Thanks for the thorough attribution work — confirming this reproduces on pristine dev is very helpful. Could you attach the full coordinator log for the affected job id plus a jstack thread dump of the hung seatunnel.sh client process? That would confirm whether the client is blocked in ClientJobProxy.waitForJobComplete waiting on a job status that is never published. The symptom (pipeline loops RUNNING -> CANCELING on checkpoint error, then all job counters drop to 0 without the job ever reaching FINISHED/FAILED/CANCELED) suggests the job's terminal-state transition is lost during pipeline restore when the failure repeats. Worth inspecting SubPlan.handleCheckpointError and the pipeline restart/cancel path in JobMaster to see if the job future is completed when the pipeline can never recover. As a mitigation, we should also consider a client-side safeguard: if the master no longer tracks the job id (counters report zero outstanding jobs), the client's wait loop should fail fast with a clear error instead of blocking indefinitely. <!-- streview-comment:100 --> -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
