potiuk commented on PR #41329: URL: https://github.com/apache/airflow/pull/41329#issuecomment-2506576044
Also (together with escalation path) we should document (alongside the escalation path) whether we send signal to the process or to process group and whether we have separate process group. This makes a lot of difference. For example we have whole section of it and handling it differently for celery workers because signal propagation for Celery should work differently than for most other cases https://airflow.apache.org/docs/docker-stack/entrypoint.html#signal-propagation (in celery we cannot send signals in docker to the whole process group for example if we want celery to handle putting the workers in offline state, but for all other containers we want to send signals to the whole process group in order to make sure we kill everything regardless if signals are properly propagated - for example to handle bash that does not propagate signals to it's children). So we need to make deliberate decision on that one as well. @ashb - what's your thinking on process groups and signal propagation? I personally think we should create a new process group when we create "task" process and send the signal to the whole process group, not just to the process. WDYT? -- 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]
