DanielLeens commented on issue #12117: URL: https://github.com/apache/seatunnel/issues/12117#issuecomment-5618440416
Thanks for doing the extra call-path audit. I rechecked current dev: CoordinatorService still submits jobMaster.run() to the shared coordinator executor, while JobMaster.run() waits on jobMasterCompleteFuture. I found no open PR that already implements the proposed lifetime/admission separation; #12165 remains a separate checkpoint-trigger scheduling concern. Your scope is appropriate, with one sequencing constraint. Please make the first PR an execution-ownership change only: separate the job-lifetime wait and the completion/control work it depends on from admission, while preserving cleanup, running-master removal, exceptional completion, cancellation, savepoint/checkpoint, restore, and master-step-down semantics. It must retain the existing configuration defaults and keys in that first slice. Only after that is merged and its saturation/recovery contract is proven should a second, independently reviewable PR choose a finite admission bound, queue/rejection policy, configuration documentation, and compatibility guidance. The first PR needs deterministic tests that saturate admission while completion, cancellation, checkpoint/savepoint, restore, and active-master reactivation still make progress. Please keep #12165 out of scope and link the first PR here when ready. No labels or formal assignee were changed. -- 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]
