I have a web server set up on Nginx and PHP-FPM (PHP version 8.1). It has worked fine for over a year now. Lately, I've noticed that the site would be down for hours - maybe once or twice a week. After investigating, I found that if I manually restart php-fpm sudo systemctl restart php8.1-fpm, everything starts working again. I suspect that it could be a spike in ddos attacks, because there is some guy that keeps harassing our site. It is also a higher-traffic site as well.
We have multiple sites/pools set up. Those sites continue working fine. It's just this one site where php-fpm seems to freeze up and doesn't work anymore, and won't restart automatically either.
Here are parts of my php8.1-fpm.log for that time period when the site failed:
[03-Aug-2023 22:49:30] NOTICE: [pool mysite] child 3543342 exited with code 0 after 62.878657 seconds from start
[03-Aug-2023 22:49:30] NOTICE: [pool mysite] child 3545476 started
[03-Aug-2023 22:49:33] WARNING: [pool mysite] server reached max_children setting (300), consider raising it
[03-Aug-2023 22:49:35] NOTICE: [pool mysite] child 3543761 exited with code 0 after 57.142360 seconds from start
[03-Aug-2023 22:49:35] NOTICE: [pool mysite] child 3545610 started
Here is the configuration for the site:
[mysite]
user = mysite
group = mysite
listen = /run/php/php8.1-fpm-mysite.sock
listen.owner = www-data
listen.group = www-data
listen.mode = 0660
pm = ondemand
pm.max_children = 300
pm.process_idle_timeout = 10s
pm.max_requests = 20
I few weeks ago, I added the following to the main php-fpm.conf file, in an attempt to restart php-fpm automatically when this happens, but that doesn't seem to be the solution:
emergency_restart_threshold=3
emergency_restart_interval=1m
process_control_timeout=5s
I'm not sure exactly what causes it or why it stops responding for one website only. However, I would like to know an automated solution or some configuration so php-fpm restarts itself when it happens.
