{code}
export PHP_FCGI_MAX_REQUESTS = 200
export PHP_FCGI_MAX_CHILDREN = 1

php -b localhost:9000 &
php -b localhost:9001 &
php -b localhost:9002 &
php -b localhost:9003 &
php -b localhost:9004 &
{code}

- We now have 5 php instances running, each will last for 200 requests before dying (to prevent memory leaks), and each instance is on its own port.

This is how mod_fastcgi and mod_fcgid (to my knowledge) spawn php instances.
This DOES NOT SUPPORT APC.



{code}
export PHP_FCGI_MAX_REQUESTS = 1000
export PHP_FCGI_MAX_CHILDREN = 5

php -b localhost:9000 &
{code}

- We now have 6 php instances running, 1 master and 5 child processes. Each will last for 200 requests before dying (to prevent memory leaks), but only the master process has a port, it handles sending requests to the children that actually handle the requests.

Because of this master/child setup, APC can work (the master keeps a cache, that all the children can share)

This is how php-fpm and lighttpd's spawn-fcgid work. (usually used with alternative webservers.)




While both of these code blocks will work to start a php fastcgi server without any tools like php-fpm or such, they are NOT for use in production. You need a process manager that will automatically restart PHP when it ends (when it hits the max request limit.)