wyattwalter opened a new issue, #37263: URL: https://github.com/apache/airflow/issues/37263
### Apache Airflow Provider(s) cncf-kubernetes ### Versions of Apache Airflow Providers ``` $ pip freeze | grep cncf apache-airflow-providers-cncf-kubernetes==7.8.0 ``` ### Apache Airflow version 2.7.3 ### Operating System Amazon Linux 2 ### Deployment Official Apache Airflow Helm Chart ### Deployment details _No response_ ### What happened We had a need to cancel a number of DAG runs on a specific DAG. Using the UI, we marked tasks as failed to cancel. The task and DAG were then marked as failed, but the pod was left running. Eventually the pod finished and sat in `Completed` state until someone manually deleted it. ### What you think should happen instead I expected that the pod operator would've attempted to terminate the pod before the task was abandoned entirely. Other operators like the Python operator will stop the current operation in this situation in my experience, but the pod wasn't cleaned up. ### How to reproduce Run a DAG that has a task using the KubernetesPodOperator, let it start, and mark the running task as failed. ### Anything else This happens every time with the Kubernetes operator in my experience. I'm not familiar enough with Airflow's operator expectations to know if this belongs with the provider or if there's nothing that core provides for the provider to clean up. ### 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]
