svrajesh21 opened a new issue, #45846:
URL: https://github.com/apache/airflow/issues/45846
### Apache Airflow version
2.10.4
### If "Other Airflow 2 version" selected, which one?
_No response_
### What happened?
When executing DAGs that involve retries, the task retries do not behave as
expected in certain edge cases. Specifically, the task does not retry when the
retry condition should be met, or it retries beyond the expected number of
retries.
### What you think should happen instead?
The task should retry exactly as per the defined retry parameters in the DAG
definition, not exceeding the maximum retries, and retrying when the conditions
trigger.
### How to reproduce
Define a task in a DAG with a retry policy (e.g., max retries and retry
delay).
Induce a failure scenario where the task would fail and should retry.
Observe the task’s behavior in the UI or logs, particularly in cases where
retries should trigger but do not.
Check the retry count to see if it exceeds the allowed maximum retries or
doesn’t retry when it should.
### Operating System
Ubuntu 20.04
### Versions of Apache Airflow Providers
apache-airflow-providers-google==6.0.0
apache-airflow-providers-s3==7.0.0
### Deployment
Official Apache Airflow Helm Chart
### Deployment details
Managed deployment on Kubernetes via Helm
Airflow version deployed via Helm Chart 2.0.0
### Anything else?
This issue occurs every time when the retry conditions are met, but retries
either do not happen or exceed the maximum retry limit set in the DAG.
logs:[2025-01-20 12:34:56,789] {taskinstance.py:1200} INFO - Task failed.
Retrying...
[2025-01-20 12:35:01,002] {taskinstance.py:1230} INFO - Task retry exceeded
max retries.
### 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]
