Asthenia0412 commented on PR #12494: URL: https://github.com/apache/seatunnel/pull/12494#issuecomment-6147079839
@SEZ9 Thanks for the thorough trace. You're right that `deployTask` checking only `activeCancellationFuture.isCancelled()` doesn't guarantee the old task actually finished — the runtime chain you walked (`NotifyTaskStatusOperation` carrying no executionId → vertex matched by location alone → slot released while task still running) is a real race. I read the exact lines you cited and I agree that's where the issue lives. The fix path I'm thinking about: `TaskExecutionState` should include the generation id, and the coordinator matching in `updateTaskExecutionState` should filter out stale CANCELED notifications from prior generations. That's a larger change than I can safely push blind — want me to sketch it in a draft comment here before I touch more code? -- 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]
