g-dailey opened a new issue, #66729: URL: https://github.com/apache/airflow/issues/66729
### Description Airflow's scheduler deliberately deprioritizes backfill DAG runs relative to regular scheduled and manual runs within each scheduling cycle. This behavior is intentional and prevents backfills from starving production pipelines, but it is not documented anywhere in the official docs, leading to user confusion. What's happening under the hood: The scheduler uses `max_dagruns_per_loop_to_schedule` (default: 20) to limit how many DAG runs it examines per cycle. Within that batch, DAG runs are ordered by: `ORDER BY backfill_dag_run.sort_ordinal ASC NULLS FIRST` Non-backfill runs have no row in `backfill_dag_run`, so the `LEFT JOIN` produces `NULL` for `sort_ordinal`. `NULLS FIRST` places them at the top of the list. Backfill runs, which have actual `sort_ordinal` values, always sort after all regular runs. ### Use case/motivation On the backfill dagrun page, it will be helpful to have a short blurb letting user know that the backfill run will have lower priority over the scheduled and manual runs, which can lead dagrun remaining in scheduled state for an extended period of time. <img width="2180" height="206" alt="Image" src="https://github.com/user-attachments/assets/f13eb5f9-a4f4-43fb-ab7e-44052efc43de"; /> ### Related issues _No response_ ### Are you willing to submit a PR? - [x] Yes I am willing to submit a PR! ### Code of Conduct - [x] I agree to follow this project's [Code of Conduct](https://github.com/apache/airflow/blob/main/CODE_OF_CONDUCT.md) -- 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]
