GitHub user akbarnotopb created a discussion: The mechanism behind Marking task
as Success
# Apache Airflow version
Airflow 2.9.3 (google composer 2)
# What happened?
**tl;dr**
My task’s first attempt ends with `Marking task as SUCCESS` but never shows
`Task exited with return code 0`. Then Airflow retries it two more times. Why?
Which component is actually failing?
---
I have a task that ended up running **3 times**.
### Attempt 1
- The log ends with:
`Marking task as SUCCESS`
but **no** `Task exited with return code 0` (which normally appears).
- After that, there are **no additional logs**, and the scheduler decides to
retry.
### Attempt 2
This attempt looks normal:
```
[2025-11-27T09:08:35.327+0000] {taskinstance.py:1206} INFO - Marking task as
SUCCESS. dag_id=etl_extras, task_id=raw_kafka,
run_id=manual__2025-11-27T08:58:13.686272+00:00,
execution_date=20251127T085813, start_date=20251127T090752,
end_date=20251127T090835
[2025-11-27T09:08:39.026+0000] {local_task_job_runner.py:243} INFO - Task
exited with return code 0
[2025-11-27T09:08:39.746+0000] {taskinstance.py:3506} INFO - 0 downstream tasks
scheduled from follow-on schedule check
[2025-11-27T09:08:39.856+0000] {local_task_job_runner.py:222} INFO -
::endgroup::
```
But despite the normal exit, the scheduler triggered a **third** attempt.
### Attempt 3
It immediately logs:
`Dependencies not met for <TaskInstance: etl_extras.raw_kafka
manual__2025-11-27T08:58:13.686272+00:00 [success]>, dependency 'Task Instance
State' FAILED: Task is in the 'success' state.`
---
### Questions
1. What exactly happens internally after Airflow logs “Marking task as
SUCCESS”?
2. Under what conditions can Airflow mark a task as success but still treat the
attempt as incomplete or failed?
3. How can I prevent this from happening again?
# What you think should happen instead?
The dag should run succesfully
How to reproduce
I have no idea (yet) how to reproduce this problem outside of our existing
environments.
Thanks!
GitHub link: https://github.com/apache/airflow/discussions/59352
----
This is an automatically sent email for [email protected].
To unsubscribe, please send an email to: [email protected]
