DanielLeens commented on issue #12442: URL: https://github.com/apache/seatunnel/issues/12442#issuecomment-5794529871
Classification: A / Zeta checkpoint trigger failure handling. Confirmed on current `dev`: `scheduleTriggerPendingCheckpoint` submits `tryTriggerPendingCheckpoint` directly to the scheduler, and `tryTriggerPendingCheckpoint` can throw before either the pending-checkpoint re-arm or the normal next-trigger scheduling path. With the returned scheduled future unobserved, that makes a periodic checkpoint trigger failure operationally invisible and leaves no next trigger armed. This predates the shared-scheduler PR; #12165 must not be used to mask or widen this separate defect. The first fix should be deliberately small but complete: wrap the scheduled body, log the unexpected failure through the engine logger with job id, pipeline id, and checkpoint type, and route it through an explicit coordinator-error policy. Do not blindly reschedule an arbitrary unexpected `Throwable`; that would turn a broken coordinator state into an unbounded silent retry loop. The PR should state the selected failure policy and add deterministic coverage for: an exception before any re-arm becoming visible to the coordinator/job error path, ordinary pending/min-interval re-arms continuing to work, and no loss of task-context diagnostics. Please link one focused PR here when that behavior and its tests are ready. -- 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]
