Abolfazlrwm commented on issue #63374: URL: https://github.com/apache/airflow/issues/63374#issuecomment-6693686087
Hi all, I'd like to work on this. Plan: open a fresh PR within 48h building on the approach from the (stale-closed) #63692 — credit to @Sathvik-Chowdary-Veerapaneni for the original root-cause analysis. From @kaxil's review guidance on #69991, the fix belongs in the callback-production path used by `DagRun.update_state()`: - `DagRunContext._sanitize_consumed_asset_events` currently recovers only from `DetachedInstanceError`. An invalidated transaction or closed session escapes that narrow catch (e.g. `PendingRollbackError`), so the `DagCallbackRequest` is never produced and the scheduler logs `callback is empty`. - I'll widen the recovery to `SQLAlchemyError`, read the DagRun PK via `inspect(...).identity` so the recovery itself cannot re-raise on an expired instance, and add unit tests that simulate the poisoned-session failure mode directly. Since @ieCecchetti could not reproduce this in clean environments, I'll note in the PR that it appears to be a session/timing-dependent race — happy to adjust scope based on maintainer feedback. -- 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]
