seruman opened a new issue, #65307:
URL: https://github.com/apache/airflow/issues/65307
### Under which category would you file this issue?
Airflow Core
### Apache Airflow version
3.2.0
### What happened and how to reproduce it?
We set `max_active_runs_per_dag = 1` in `airflow.cfg` under `[core]` so that
DAGs default to sequential execution. DAGs that need concurrency set
`max_active_runs` explicitly in their definition.
The problem: if a DAG sets `max_active_runs=16`, the effective value ends up
being `1` - the config value - in the database. It looks like `16` is being
treated as "not explicitly set" because it matches Airflow's built-in default
for `max_active_runs`.
Setting `max_active_runs=17` or any other value that isn't 16 works fine,
Airflow respects it.
1. Set `max_active_runs_per_dag = 1` in `airflow.cfg`.
2. Create a DAG with `max_active_runs=16`:
```python
@dag(
dag_id="test-concurrent",
schedule=None,
max_active_runs=16,
)
def my_dag():
...
```
3. Let the scheduler pick it up. Check the effective `max_active_runs` in
the metadata database or the DAG details page in the UI. It will show `1`
instead of `16`.
4. Change it to `max_active_runs=17`. Now it correctly shows `17`.
### What you think should happen instead?
If a DAG explicitly sets `max_active_runs=16`, that should take precedence
over `max_active_runs_per_dag` in `airflow.cfg`, even though `16` happens to
match Airflow's built-in default.
### Operating System
Debian GNU/Linux 12 (bookworm) — official apache/airflow:3.2.0 container
image
### Deployment
Other Docker-based deployment
### Apache Airflow Provider(s)
_No response_
### Versions of Apache Airflow Providers
_No response_
### Official Helm Chart version
Not Applicable
### Kubernetes Version
_No response_
### Helm Chart configuration
_No response_
### Docker Image customizations
_No response_
### Anything else?
Our workaround is setting `max_active_runs` to the desired value + 1 for any
DAG that needs concurrency. Not ideal but it works.
### Are you willing to submit 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]
