manage.py runfcgi dies in response to SIGHUP/SIGUSR1, disallows log rotation
Thanks to #6687, daemonized fastcgi processes can now log their stderr and stdout. In our environment we would like to capture these logs, and rotate them (via the standard method of using logrotate). However, from looking in the code, it appears that runfcgi does not support closing/reopening its log file (which would allow logrotate to rotate the log out from under it). Sending a SIGUSR1 or SIGHUP to the master fcgi PID simply kills it instead of causing it to do this. Unless I'm missing something, it looks like the only way to do this at the time is to stop and then restart the processes, which is unacceptable in a production environment.
A necessary patch would handle a SIGUSR1 and/or SIGHUP to close() and then re open() the files specified in errlog and outlog.
I can write a patch if necessary using the python signal module.
Change History
(10)
| Cc: |
Alexander Koshelev added
|
| Triage Stage: |
Unreviewed → Design decision needed
|
| Severity: |
→ Normal
|
| Type: |
→ New feature
|
| Component: |
Core (Other) → Core (Management commands)
|
| Resolution: |
→ wontfix
|
| Status: |
new → closed
|
Change UI/UX from NULL to False.