Conversation
|
Thank you! This may have been fixed (or partly fixed) by f4fe925 which was released as part of version 2.5.2. Please review and report :) |
Pidfile is removed only when interpreter exists. While restarting using SIGHUP signal, obviously, interpreter doesn't exit, so pidfile is not removed. We're fixing this by deleting pidfile in WorkController._shutdown().
|
Thanks for the answer. I've tried 2.5.2, the problem still arises. As I understand ask@f4fe925 is about corrupted pidfiles. Also, I have rewritten my patch, making PIDFile class unchanged. I moved creation/deletion of a pidfile into the WorkController class. |
|
Aha, so I guess the process id is still alive, so that the stale pidfile detection doesn't work, |
|
I have to move it to the end of the stop() method though, because not it could create a race |
|
The process id is still alive because it is the same process :) |
|
I think |
There was an error, saying "Pidfile is already exists" while restarting
using SIGHUP. The problem arises only when "--pidfile" option is used.
Related to: https://github.com/ask/celery/issues/26