PrafullPandey06 commented on issue #70117:
URL: https://github.com/apache/airflow/issues/70117#issuecomment-5047615518

   You're right I shouldn't have stated the dumb-init reaping path as confirmed 
root cause. I don't have strace evidence showing PID 1 actually calling waitpid 
on the task PID, and as you pointed out, that shouldn't normally happen while 
the supervisor is alive. I've updated the issue title and description to 
reflect this.
   
   That said, the hang is reproducible and the impact is real  26–30 stuck 
supervisors accumulating ~8–10 GB of wasted memory per worker pod. Whatever is 
causing wait(timeout=0) to return None, the supervisor doesn't handle that 
case, and the cleanup timeout has no independent fallback.
   
   Your reproduction script confirms the exact code gap  _check_subprocess_exit 
only handles an int return or TimeoutExpired, so None is silently ignored. 
   


-- 
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]

Reply via email to