Description
Description
Airflow 2.x: The DAG list filters for run-related states behaved like they used DAG runs aggregated per DAG (counts or “has any run in this state”), so you could still find DAGs with failures when the latest run had already succeeded.
Airflow 3.x: The UI still offers filters for Queued, Running, Success, and Failed, but they behave as if they use the latest DAG run’s state only. Example: a DAG with many failed runs and a latest success disappears from a failure-oriented view even though failure runs still exist.
Ask: Change the DAG list state filters so they apply to DAG run totals per DAG by state (same general idea as 2.x), scoped however the backend defines which runs are considered for that list view. If the scope is not obvious in the UI, document it.
Use case/motivation
Use case / motivation
We use the DAG list as the first place to spot trouble: filter to failed (or queued/running), open those DAGs, investigate.
That only works if “failed” means this DAG has failed DAG runs, not the last DAG run failed. Right now 3.x follows the latest run and hides backlog behind a green latest run.
Related issues
Related issues
- #66946 DAG list page: show total run state counts per DAG
- #50624 Bring back “Status of all previous DAG runs” column on the DAGs page
- #60105 UI: task instance summary on DAG cards
Are you willing to submit a PR?
Code of Conduct
Description
Description
Airflow 2.x: The DAG list filters for run-related states behaved like they used DAG runs aggregated per DAG (counts or “has any run in this state”), so you could still find DAGs with failures when the latest run had already succeeded.
Airflow 3.x: The UI still offers filters for Queued, Running, Success, and Failed, but they behave as if they use the latest DAG run’s state only. Example: a DAG with many
failedruns and a latestsuccessdisappears from a failure-oriented view even though failure runs still exist.Ask: Change the DAG list state filters so they apply to DAG run totals per DAG by state (same general idea as 2.x), scoped however the backend defines which runs are considered for that list view. If the scope is not obvious in the UI, document it.
Use case/motivation
Use case / motivation
We use the DAG list as the first place to spot trouble: filter to failed (or queued/running), open those DAGs, investigate.
That only works if “failed” means this DAG has failed DAG runs, not the last DAG run failed. Right now 3.x follows the latest run and hides backlog behind a green latest run.
Related issues
Related issues
Are you willing to submit a PR?
Code of Conduct