We're running on Amazon Linux (which is a slim version of CentOS 6.X). We're using httpd 2.4 (2.4.12), with event_mpm on defaults settings. We're connecting to php-fpm (5.4.40) with php 5.4 (5.4.40). We're connecting to php-fpm through mod_proxy_fcgid.
These are the relevant apache configs. As you can see, we're been playing around with it. We did not make any change to the default mpm settings
# PHP-FPM FastCGI Proxying
#ProxyPassmatch ^/(.*\.php)$ unix:/var/run/php-fpm/www.sock|fcgi://127.0.0.1:9000/var/www/html/
#ProxyPassmatch ^/(.*\.php)$ fcgi://127.0.0.1:9000/var/www/html/$1
<FilesMatch "\.php$">
SetHandler "proxy:unix:/var/run/php-fpm/www.sock|fcgi://127.0.0.1:9000"
</FilesMatch>
Right now we're using some default settings for php-fpm
;emergency_restart_threshold = 0
;emergency_restart_interval = 5s
We're running on an 8 cores, 15 gb server (AWS c4.2xlarge).. if we use
pm.max_children = 50
we get 90rps, but our load is 12 with no errores. If we set up a higher number (e.g.: 1024) we get a load of 40, no errors as well. If we set it to 10, we get a sane load (6, 7) but we start getting 408 errors, even without traffic.
As a side note, we also included and removed apc without significant changes.
thx!