houqp edited a comment on issue #14422: URL: https://github.com/apache/airflow/issues/14422#issuecomment-821296305
Interesting, I was expecting the second SIGTERM would have resulted in the task subprocess to set its own state in through `handle_failure` because `self.on_kill` calls `self.task_runner.terminate()`, which is supposed to wait for the subprocess to exit: https://github.com/apache/airflow/blob/e7c642ba2a79ea13d6ef84b78242f6c313cd3457/airflow/task/task_runner/standard_task_runner.py#L108-L117 > We should probably add self.task_instance.state=State.FAILED in handle_task_exit if exit_code != 1. WDYT @houqp @ephraimbuddy ? I think we should do this as an extra safe guard because in rare cases, the task sub process could crash any time after it received SIGTERM and before it updates its own task state. However, I think the state update logic should be set with an extra condition: ```python if self.task_instance.state not in State.finished(): self.task_instance.state = State.FAILED ``` This is to handle the case where the task could have exited successfully right after the pod is getting killed but before local task job executes `handle_task_exit`. -- 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. For queries about this service, please contact Infrastructure at: us...@infra.apache.org
