ashb commented on PR #24743: URL: https://github.com/apache/airflow/pull/24743#issuecomment-1173987183
If the docstring in the example dag is still accurate: > Note that even though dag4 depends on the dataset in dag1, it will not be triggered until dag2 runs (and dag2 is Then that is a major change from the AIP, and directly counter to one point of future work section: > In a similar manner to how a Task has trigger_rules for changing when it can run based on different possible states of upstream tasks, we would like to add different "modes" for scheduling a DAG when it has multiple input datasets. The two modes that come to bind off the bat are "all changed" (to only trigger the run once all input datasets have updated) and "any changed" so trigger a run when any one dataset has changed. **(The default behaviour, and the only behaviour in this AIP would be "any changed")** (Emphasis added) Not to mention that if you only work in this mode, you can then include run_id in the event's table and then that's how we sort out the HA locking issue, because we only look for the row matching the source dag run (`self`) which is already locked. -- 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]
