I am still experimenting with PHP-FPM but it seems like a pretty good way to go as far as handling the PHP-CGI processes.

To do this, you will need to install EPEL & CentALT repositories:

rpm -Uvh http://download.fedora.redhat.com/pub/epel/5/x86_64/epel-release-5-3.noarch.rpm

rpm -Uvh http://centos.alt.ru/repository/centos/5/x86_64/centalt-release-5-3.noarch.rpm

Then do:

yum install php-fpm

to install php-fpm.

Edit the file /etc/php-fpm.conf

and my suggested setting changes:

<value name=”user”>apache</value>
<value name=”group”>apache</value>

The max_children is set to 1 as default, set it to something like 50, this really is the cool part about php-fpm.

<value name=”max_children”>50</value>

PHP-FPM doesn’t allow for root user and group so I set it to apache.  You will also need to set your web directories to apache user and group by doing:

chown apache:apache /mywebdirectory

Start the PHP-FPM by doing:

/sbin/service php-fpm start

If you get some kind of session errors, simply erase all the session files where the error shows.

For CentOS, you can do:

chkconfig php-fpm on

so it will start whenever your webserver is rebooted.

If you have been using fastcgi, make sure to do:

chkconfig fastcgi off

so it doesn’t interfere.

I’ve been researching many ways to optimize large file downloads but probably the best way is to simply rate limit the bandwidth on how fast each user can download.

You can add the following line to your HTTP directive in /etc/nginx/nginx.conf:

limit_rate 100k

This will limit each user to download at maximum of 100kb/second.

You can change this number to see which one fits best for your server.

You can also add limit_rate_after clause to only limit after a portion of the file has been downloaded:

limit_rate_after 1m;
limit_rate 100k

This will limit downloads only after the first megabyte, potentially not slowing down smaller file downloads and your website.

Well, my Nginx kept crashing, actually my PHP-cgi processes started crashing after the server ran for a bit. The weird thing was, the server had no load.

Upon carefully watching my “top” status in linux for awhile, I realized there’s only one PHP-CGI process running, whereas all my other Nginx web servers ran multiple (around 5-6 at any given time).

I compared out the versions of the Spawn-cgi and realized the new Spawn-cgi I had installed only spawned one PHP-CGI process.

I decided to install an older version off Lighttpd and it fixed the problem:

wget http://www.lighttpd.net/download/lighttpd-1.4.18.tar.bz2

tar -xvjf lighttpd-1.4.18.tar.bz2

cd lighttpd-1.4.18

./configure

make

cp src/spawn-fcgi /usr/bin/spawn-fcgi

After that restart your PHP-CGI:

/sbin/service fastcgi stop

/sbin/service fastcgi start

Yey, problem solved and now Nginx server runs multiple PHP-CGI processes and no more weird crashes. Sometimes, older programs work better. Perhaps the new Spawn-cgi version has bugs?