DanielLeens commented on issue #12164: URL: https://github.com/apache/seatunnel/issues/12164#issuecomment-5584883373
Thanks @Rangsh. I rechecked current `dev` (`f6ed9579c5e`) and found no matching open PR or later `TaskExecutionService` change. The post-publication failure boundary remains real: `executionContexts` is populated before `onContextPublished` and task submission, while the catch path only invokes the supplied failure cleanup before context publication. A failure after publication can therefore leave the location looking active and suppress a later redeploy. Your claim is accepted. GitHub does not recognize `Rangsh` as an assignable `apache/seatunnel` assignee, so I have deliberately left the formal assignee and labels unchanged rather than pretending the assignment succeeded. Please make the first PR a narrow rollback fix with deterministic regression coverage that injects a failure after publication (including the callback/submission boundary), proves that the published context and associated cancellation state do not leak, and proves that a second deployment of the same `TaskGroupLocation` actually executes. Preserve the existing master-failover behavior in which an already active, healthy task returns success rather than being redeployed. Do not broaden this into unrelated executor or classloader lifecycle changes. -- 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]
