jedcunningham commented on pull request #14085: URL: https://github.com/apache/airflow/pull/14085#issuecomment-794498196
I noticed different behavior this morning around whether the parent process alone is signaled (e.g. kill PID), or the whole process group (bash ctrl-c or dumb-init). With this change when the whole process group is signaled, `LocalExecutor.end()` blows up with a `BrokenPipeError`. However, even without this change (`LocalExecutor.end()` isn't called) you can get a `EOFError` on the task_queue if a task is running. I'm not sure this change makes it "worse", just different, and in both cases the task process is left behind. This makes me think there are some dark corners in LocalExecutor and I feel I've drifted pretty far from the original "KubernetesExecutor doesn't exit on SIGTERM" issue. Thoughts? Input? ---------------------------------------------------------------- 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: [email protected]
