FelipeLahti commented on issue #51569: URL: https://github.com/apache/airflow/issues/51569#issuecomment-4063079672
Hit this on 3.0.1 with LocalExecutor + DockerOperator. Tasks were stuck in "queued" for 10+ minutes even though parallelism=32 and pool had 128 slots. Traced it to `_check_workers()` — the condition `len(self.workers) < num_outstanding` doesn't account for busy workers, so it never spawns new ones when existing workers are all occupied. Upgrading to 3.1.8 fixed it. PR #58365 changed the condition and also pre-spawns all workers at startup on Linux. Went from 4 stuck workers to 32 pre-spawned and idle at boot. -- 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]
