hkc-8010 opened a new issue, #70368:
URL: https://github.com/apache/airflow/issues/70368
### Apache Airflow version
3.3.0 and later
### What happened
Clearing a failed DAG run or task instance returns HTTP 500 ("Internal
Service Error") in the UI when the DAG defines a deadline that references the
DAG run's `queued_at` time (`DeadlineReference.DAGRUN_QUEUED_AT`). The API
server logs:
```
TypeError: unsupported type for timedelta seconds component: dict
```
### What you think should happen instead
The clear should succeed, and the queued-at deadline should be recalculated
against the new `queued_at`.
### Root cause
In 3.3.0, migration `0117_3_3_0_change_deadline_interval_to_json.py` (from
#64751, which added `VariableInterval` support) changed the
`deadline_alert.interval` column from FLOAT to JSON. Existing values were
rewritten into the serde envelope, e.g. `{"__classname__":
"datetime.timedelta", "__version__": 2, "__data__": 3600.0}`.
`_recalculate_dagrun_queued_at_deadlines()` in
`airflow-core/src/airflow/models/taskinstance.py` (called from
`clear_task_instances`) still reads the raw column and does
`timedelta(seconds=deadline_alert.interval)`, passing the JSON dict straight
into `timedelta`. Every other consumer (DagRun creation in
`serialization/definitions/dag.py`, the model `__repr__`) decodes the JSON
first. Only the clear path was missed.
### How to reproduce
1. Define a DAG with a deadline referencing
`DeadlineReference.DAGRUN_QUEUED_AT`.
2. Let a run queue so a `Deadline` row is created.
3. Clear the DAG run or one of its task instances.
The API server returns 500 with the `TypeError` above.
### Workarounds
- Delete the task instance in the UI instead of clearing it.
- Temporarily remove the `deadline` parameter from the DAG.
### Are you willing to submit a PR?
Yes.
--
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]
