The one black sheep in our diligent administration is GT3.IN. Despite following the mod_wsgi docs word-for-word, there seems to be some issue when Apache terminates worker processes. I don’t have the time to debug all of that, and as an avid NGINX user, I’d prefer to switch.
Here are the solutions I’ve come up with to fix this issue:
- Deploy gt3.in using a discrete WSGI server (probably Gunicorn/uWSGI) and use mod_proxy to proxy requests to that server.
- Wait till US-VA-01 is upgraded, switch to NGINX and use a standard WSGI proxy deployment.
- Restart Apache weekly.
Option 1 is great, but going through all of that trouble for all of it to be washed away when we upgrade to Trixie isn’t worth it. I’d love to switch to NGINX as in Option 2, but that would break some htaccess configurations used by existing users. Option 3 isn’t really a fix, it’s a Jerry-rigged solution, so I’m not making a cron job for that.
Since there’s basically only two of us sysadmins, and GT3.IN is my thing, I’ll be going for Option 2. I can’t promise a definitive timeline for the fix since I have a lot of other commitments, but I can guarantee that it’ll be done before Debian 12 goes EOL.
I have also created a Statuspage for streamlined communication: https://devinivas.statuspage.io/. Unfortunately, it’s not automated, but it’s certainly better than blog posts on a website for family-related information.
To the few of my friends who use GT3.IN, I apologize for this downtime. I can guarantee >95% uptime with the NGINX method – I’ve used it in multiple deployments which haven’t been touched in years, and they’re holding up just fine!