SEZ9 commented on issue #11437: URL: https://github.com/apache/seatunnel/issues/11437#issuecomment-5140969638
The 'Steps to reproduce' section appears to have been cut off. Could you post the exact reproduction steps, plus the relevant hazelcast/seatunnel config files (seatunnel.yaml, hazelcast.yaml) and the FakeSource job config, so we can reproduce on dev? Please attach the new active Master's logs around the restore window (ResourceManager / SlotService / CoordinatorService entries) and both Workers' SlotService logs. This would confirm whether the new Master's ResourceManager ever learned about the previously assigned slots (worker profile re-registration) after failover, which the 'Slot request future not created' error suggests it did not. The symptom that Workers still report slots for a cancelled job (requiring Worker restart to free them) points to a slot-release path that is not invoked when the owning Master changed. This looks distinct from #10568/#9807 since it blocks before deployTask(); likely the restored job's ResourceProfile/slot ownership state is not rebuilt on the replacement Master. Marking this as a Zeta resource-manager failover bug pending log confirmation. <!-- streview-comment:16 --> -- 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]
