dsuhinin opened a new pull request, #69336: URL: https://github.com/apache/airflow/pull/69336
---
When a task instance gets dispatched twice, only one worker actually runs
it — the duplicate loses the race and dies. The problem was that the scheduler
treated the leftover event from that dead duplicate as a failure and killed the
live run that was happily executing.
This fixes both sides of the race:
- Scheduler: before failing a queued task, it now checks whether the task
is actually alive — RUNNING with a recent heartbeat. If a worker is clearly
still on it, the stale event is ignored. Tasks that genuinely died are still
caught later by heartbeat-timeout detection.
- Supervisor: when the server tells a duplicate worker to back off
(TaskAlreadyRunningError), it now exits quietly instead of reporting a crash.
The duplicate never did any real work, so there's nothing to fail.
possible fix for: #57041
---
--
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]
