jscheffl commented on issue #41894: URL: https://github.com/apache/airflow/issues/41894#issuecomment-2323043882
Thanks for reporting the bug. Whereas retrospective version 2.7 is pretty old and very probably we will not release patches. Do I understand it right that mainly you have problems in your integration tests but your production system is running fine? Or is it a regular (production) use case to run a DAG several times with the same execution date? > We wrote a patch to fix the issue. This patch will get us through the Airflow 2.7.3 upgrade so we can continue upgrading. We don't understand why there aren't other people with the same problem. I assume most users do not re-use the execution date. The execution date is defined ans unique as it was in the past part of the primary key. This had changed within the Ariflow 2 version stream and run_id became the primary key. In Airflow 3 we will remove the execution date as field as it was replaced with logical date. Also we have seen that a unique-ness of this field blocks some specific use cases in regards of data driven even processing. Anyway... question is: Shall we keep this bug report as a kind of documentation of "Known problems"? Do you expect that something on the very old version 2.7 is patched? My direct proposal would be to move-away from re-using the execution date in your tests. In the unit and integration tests we do with Airflow we always clean the DB before adding new test DAG records in order to prevent a clash. As re-using the execution date is not a supported use case I also would believe a patch would be earliest accepted in 2.10 line but I doubt a bit that this is a real bug/problem. -- 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]
