nielifeng commented on issue #10187: URL: https://github.com/apache/seatunnel/issues/10187#issuecomment-5140416960
I rechecked this report against the current SeaTunnel lifecycle code and the linked DolphinScheduler investigation. This is an upstream scheduler process-tree termination issue rather than a Zeta job-state bug. Evidence: - The reported command uses `--deploy-mode local`, so the SeaTunnel client process owns an embedded Local engine rather than submitting a detached job to a remote Zeta cluster. - The reporter confirmed DolphinScheduler 3.1.7 and confirmed that upgrading was the intended resolution. - DolphinScheduler tracks the same behavior in apache/dolphinscheduler#17786. - apache/dolphinscheduler#17005 (merged) changed cancellation to terminate the process tree correctly. - apache/dolphinscheduler#17320 (merged) added a process-status check after task termination. - The upstream thread contains direct A/B evidence: DS 3.3.1/3.3.2 cancels the SeaTunnel CDC process correctly, while 3.2.2 can leave an orphan process after `kill -9`. SeaTunnel already registers a shutdown hook that closes the Local job and embedded Hazelcast instance when the client receives a normal termination signal. A process killed with SIGKILL cannot run any JVM shutdown hook, so adding SeaTunnel-side cleanup cannot repair that scheduler behavior. I recommend closing this issue as an upstream DolphinScheduler issue and using a DS version containing the process-tree cancellation fixes. No SeaTunnel PR is warranted for the evidence reported here. -- 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]
