Linux – apache internal dumthe connection

I'm not saying this is what's happening but based on my own experience as a CentOS admin, it's most likely runaway apache/php processes taking down the server. I've seen this numerous times on CentOS 5. It's frustrating because there's usually not a trace of what happened in the log files. The machine just grinds to a halt due to physical memory and swap being sucked up by apache/php processes. You would think linux memory management or some daemon would jump in and say "hey stop" but it doesn't. It'll let apache grind your system to a halt.

Having said that, to see what's happening you'll need something that can monitor and log resource usage. I like to use a program called atop. Atop is a lot like the top program but it also takes a snapshot of resource usage at defined intervals. It's pretty simple to install.

wget http://www.atcomputing.nl/Tools/atop/packages/atop-1.23.tar.gz 
tar -zxvf atop-1.23.tar.gz
cd atop-1.23 && make install

Open /etc/atop/atop.daily with a text editor and change INTERVAL=600 to INTERVAL=60

Run the command /etc/atop/atop.daily from a command prompt to start it. Wait a few minutes and run atop -r /var/log/atop/atop_20091118 with the correct date of course.

Hit the t key to go forward in time and T to go back. Next time your server crashes do this and check the MEM free and SWP free lines. If you have memory problems these will be in red. Also look for numerous httpd lines under CMD. If apache/php is your problem there'll be a bunch of them.

If this is the case, I recommend looking at you're MaxClients setting in httpd.conf. If set too high, apache will gladly eat all of your memory causing your machine to crash. Apache/php can easily eat 40-50MB/process. If you multiply 40mb x MaxClients you'll get a rough idea of how much memory apache can potentially use. MaxClients usually defaults to 150 on CentOS so apache can potentially use 6GB of memory by default. This doesn't include memory your system needs for itself and other processes to run. Try setting it to a more realistic value based on the amount of memory you have like 40 if you have 2G of memory and see if that helps. Also if you have KeepAlive On, set KeepAliveTimeout to a low number like 2 or 3.

In my opinion CentOS's apache/php compilation is a real pos that should never have seen the light of day. It's buggy and crash prone. If you run a serious site, I highly recommend compiling your own version of apache/php or even using one of the newer high performance webservers like lighttpd or nginx with fgci php.

Apache handles a thundering herd a little differently than you might imagine. When you get a burst of inbound traffic, it spawns a number of child processes, if it determines it needs more, it spawns twice as many in the next interval until it finally has enough processes to serve the requests or hits maxclients.

If you see these, it means that apache is just checking the children, and whatever caused apache to fork that many processes is probably gone. Yes they do take up client connections, but, whatever event caused things to spool up is probably gone.

First thing I would check in your logs would be a bunch of 302s prior to the event.

If you had something like

<?php include("http://www.oursite.com/header.php");?>   

where header.php was missing and

ErrorDocument 404 /404.php 

where 404.php included header.php, you would get a recursive loop and a hit on that page would immediately cause apache to use all available connections.