Vamsi-klu commented on issue #65818: URL: https://github.com/apache/airflow/issues/65818#issuecomment-5435506029
Working on the lock-order / ownership split. Not a second retry. @uplsh580 #71391 (retry on submit_event/submit_failure) is complementary. I will not duplicate that retry. If 71391 merges first, I will rebase and keep @provide_session outside @retry_db_transaction. If 71391 is growing into this ownership split, say so and I will stop. Adjacent open work I will not compete with: - #72062 MySQL clean_unused CASCADE race — absorb the DELETE re-check predicates - #70961 batch Trigger deletes — that PR keeps the TI UPDATE this one drops; we should not fight over the DELETE shape This PR: - stop clean_unused from UPDATEing task_instance - NULL trigger_id on every DEFERRED-exit, including clear_task_instances (today that leftover FK is what the bulk UPDATE papers over) - scheduler timeout becomes a bounded PK-ordered fallback for unassigned/dead triggerers only (healthy assigned triggerer + past trigger_timeout must not flip the TI) - submit_event/submit_failure SELECT FOR UPDATE SKIP LOCKED; if another writer already took the row, no-op Issue body swaps the writers: the unbounded timeout UPDATE is scheduler check_trigger_timeouts; the state!=deferred trigger_id NULL UPDATE is triggerer clean_unused. --- Drafted-by: Cursor Grok 4.6 (no human review before posting) -- 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]
