manipatnam opened a new issue, #70683:
URL: https://github.com/apache/airflow/issues/70683
### Under which category would you file this issue?
Task SDK
### Apache Airflow version
Airflow 3.1.0
### What happened and how to reproduce it?
When `TriggerDagRunOperator` attempts to trigger a non-existent DAG ID on
Airflow 3.x (Task SDK architecture):
1. `on_failure_callback` / `on_retry_callback` are **never executed**.
2. **Retries are completely ignored**. Even if `retries=3` is configured on
the task, the task immediately fails on attempt 1/4 without attempting any
retries.
### Reproduction DAG
```python
from datetime import datetime
from airflow.sdk import dag
from airflow.providers.standard.operators.trigger_dagrun import
TriggerDagRunOperator
def on_failure_cb(context):
print("Failure callback executed!")
@dag(
start_date=datetime(2024, 1, 1),
schedule=None,
catchup=False,
)
def trigger_missing_dag():
TriggerDagRunOperator(
task_id="trigger_nonexistent",
trigger_dag_id="this_dag_does_not_exist",
retries=3,
on_failure_callback=on_failure_cb,
)
trigger_missing_dag()
```
### What you think should happen instead?
### Expected Behavior
1. If `retries > 0`, `on_retry_callback` (if defined) should run, and the
task instance state should transition to `UP_FOR_RETRY`.
2. Once retries are exhausted, `on_failure_callback` should run, and the
task instance state should transition to `FAILED`.
### Actual Behavior
1. The task fails immediately on attempt 1/4 (marking state as `FAILED` in
DB/UI).
2. `on_failure_callback` / `on_retry_callback` are never executed.
### Operating System
_No response_
### Deployment
Astronomer
### Apache Airflow Provider(s)
standard
### 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?
### Technical Analysis & Root Cause
In `airflow/sdk/execution_time/task_runner.py`:
1. `TriggerDagRunOperator.execute()` raises `DagRunTriggerException`.
2. `run()` catches `DagRunTriggerException` and delegates to
`_handle_trigger_dag_run()`.
3. `_handle_trigger_dag_run()` calls
`SUPERVISOR_COMMS.send(TriggerDagRun(...))`.
4. When the API server returns `404 Not Found`, the supervisor returns an
`ErrorResponse(error=API_SERVER_ERROR, ...)`.
5. `CommsDecoder._from_frame()` receives this error and raises
`AirflowRuntimeError`.
6. **Exception Propagation Bug**: Because `AirflowRuntimeError` is raised
inside the `except DagRunTriggerException` handler block, Python propagates the
exception out of `run()` without evaluating sibling `except` clauses in the
same `try` block.
7. As a result:
- `_handle_current_task_failed(ti)` (which checks task retries) is never
called.
- `finalize()` (which fires callbacks) is never called.
- `main()` catches the unhandled exception and exits with code `1`.
8. The Supervisor sees exit code `1`, defaults `final_state` to
`TaskInstanceState.FAILED`, and calls `task_instances.finish(state=FAILED)` on
the API server.
### 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]
