YAshhh29 commented on PR #75916: URL: https://github.com/apache/airflow/pull/75916#issuecomment-6482225329
No, not from a deployment, After #71575 this area stuck with me, so I kept going through the common.ai durable code to understand it better. And while comparing the two journal backends I noticed they don't behave the same way: the task state store one does what `DurableStorageProtocol` says (skip a write it can't make, without failing the step), and the ObjectStorage one doesn't. The case that made me want to fix it is cleanup: a transient S3/GCS error there can mark a task failed after its agent work already succeeded. I reproduced each case in tests on a fresh Linux setup rather than in a real deployment, so I don't have a user report to point to. If you'd rather keep the < 3.3 path minimal, I'm happy to cut this down, e.g. keep the write guard and drop the read-side changes. -- 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]
