hutiefang76 commented on PR #12510: URL: https://github.com/apache/seatunnel/pull/12510#issuecomment-5874680725
I checked the failed Master-failover tests and found two regressions in my previous update. Fixed in 148dae7d8. The scheduler had already reserved slots for a never-started pipeline, but the takeover path canceled it and tried to reserve them again. That pipeline now uses the existing reservation; partially started tasks still follow the cancellation path. I also recheck cancellation after the recovery wait so a canceled job is not reported as a resource failure. Both real-cluster tests now pass, including repeated Master handoffs and zero retries for the initial dispatch. JDK 8/17 each passed 23 tests and verify. I reran the independent Worker-loss case too: the replacement Worker completed all 400 distinct IDs, with no Unknown slot error. The stale-slot guard is still in place. The PR description and tests are updated; waiting for the new CI run. -- 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]
