jgoedeke opened a new issue, #70779:
URL: https://github.com/apache/airflow/issues/70779
### Under which category would you file this issue?
Providers
### Apache Airflow version
3.3.0
### What happened and how to reproduce it?
`GET /ui/dags/{dag_id}/deadlineAlerts` loads every historical
`serialized_dag` row for the DAG.
In
[`airflow.api_fastapi.core_api.routes.ui.deadlines.get_dag_deadline_alerts`](https://github.com/apache/airflow/blob/main/airflow-core/src/airflow/api_fastapi/core_api/routes/ui/deadlines.py#L176-L180),
the lookup is ordered descending but has no SQL `LIMIT`:
```python
serialized_dag = session.scalar(
select(SerializedDagModel)
.where(SerializedDagModel.dag_id == dag_id)
.order_by(SerializedDagModel.id.desc())
)
```
For a DAG with 6,725 historical serialized DAG versions totaling about 4
GiB, PostgreSQL returned all rows and the API process exceeded its memory limit
and was OOM-killed.
Add historical rows for one `dag_id`, then request:
```text
/ui/dags/<dag_id>/deadlineAlerts
```
The query should retrieve only the newest serialized DAG row.
### What you think should happen instead?
The query should include `.limit(1)` before calling `session.scalar()`:
```python
.order_by(SerializedDagModel.id.desc())
.limit(1)
```
This returns only the latest serialized DAG and prevents memory usage from
scaling with historical DAG-version count.
### Operating System
_No response_
### Deployment
None
### 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?
_No response_
### Are you willing to submit PR?
- [ ] 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]
