github-actions[bot] opened a new pull request, #72679: URL: https://github.com/apache/airflow/pull/72679
* Fix Dag scheduling stall after switching to a coarser cron Changing a Dag's schedule to a coarser cron (e.g. hourly to daily) could make the realigned next run collide with the logical_date of the run already on record. The scheduler would then loop on "run already exists; skipping dagrun creation" forever instead of advancing, since every input to the recompute (the DAG's static start_date, the stale reference run) is a fixed constant that reproduces the identical collision every time. A single-step guard for this class of collision already existed for the narrower case of a CronTriggerTimetable-to-CronDataIntervalTimetable switch. A single step turns out to always be enough for both shipped _DataIntervalTimetable subclasses, but only because of invariants that are specific to them and aren't enforced anywhere -- a custom Timetable isn't guaranteed to uphold them. Retry a bounded number of times instead of assuming one step suffices, and fail loudly if that budget is exhausted rather than looping unbounded, which would otherwise hang the scheduler's main loop for every Dag if a future _get_next implementation ever stopped strictly advancing. * Simplify the coarser-schedule-stall guard to a single bounded retry Only one advance past the previous run's start is provably necessary for both shipped _DataIntervalTimetable subclasses; a second retry was unproven margin that only existed to satisfy a synthetic test case, not any real or documented timetable behavior. A bounded while loop implied that margin was load-bearing. Replacing it with a single retry followed by a loud failure keeps the same fail-fast guarantee -- a custom Timetable needing more than one step is still caught immediately rather than silently retried -- while matching what the invariant actually proves. * Use ValueError for timetable alignment guard * Avoid exposing timetable internals in error message * Simplify schedule change test runner helper * Clarify zero-length interval test wording * Clarify coarser schedule regression test comment (cherry picked from commit 31117014f3c5a8cd4d4acd4223309eb344e21981) Co-authored-by: Jason(Zhe-You) Liu <[email protected]> -- 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]
