nathadfield commented on PR #71074: URL: https://github.com/apache/airflow/pull/71074#issuecomment-6438313105
@kaxil I’ve re-run the PostgreSQL reproduction and checked MySQL and the upgrade path too. These were against merged main at `5243901c81`, rather than the earlier head you linked. - **PostgreSQL**: all five live runs passed. The two producer tasks finished less than 0.5 ms apart each time, both events landed on one APDR, and the consumer successfully consumed both. No pending rows remained. - **MySQL 8.4.11**: 13 focused tests passed with no skips, including the forced two-session distinct-assets race and savepoint recovery. This was database-level testing rather than a full producer/consumer run. - **3.3 → 3.4 upgrade on PostgreSQL**: I seeded split pending rows in an isolated 3.3 schema and ran the upgrade. Both events and their links survived unchanged, only the newest APDR received the pending marker, and the uniqueness constraint worked. The old split remained, as intended by the final migration. That confirms the original reproduction is working with the merged fix. I haven’t validated the separate producer/scheduler overlap follow-up yet though. -- 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]
