Too many apache processes

apache-2.2memorymemory usage

I'm wondering why my server has so many free apache slots open. The server only has a couple of busy and idle slots. Please look at the attached images for the amount of process used by the server versus our traffic.

Questions

Should I change my configurations so that there's less free slots?
If I do reduce the amount of free slots will it reduce the amount of memory the server is using?

Stats

Apache process: http://i.imgur.com/cMthh.png
Google Analytics: http://i.imgur.com/yl8Ui.png

Best Answer

You did not provide information on how idle slot/free slot and busy slot are computed on your graph. So I'll assume the following :

Busy slots are process/threads actually serving requests
Idle slots are process/threads created but not serving request
Free slots are process/threads that can be created before reaching max client.

Question 1 :

You should reduce max client only if you don't have enough RAM to handle all the HTTP requests you allow. For now, you can wait because "busy slot" never got above 100 so far.

Question 2 :

No. Apache uses more memory only when it spawns more process/thread.

Related Solutions

Vps – Apache crashing VPS – need to reduce memory usage

The best method for reducing apache's memory usage is to move away from mod_php and to something like fastcgi. Each of your apache processes is 15mb or more because of mod_php overhead (most likely). Having php requests handed off to fastcgi will reduce the average apache process size to approx 1mb or so depending on the apache configuration.

Since php is now centralized using fastcgi it's memory usage is more efficient and the total amount of memory used by the system should decrease slightly.

Another approach would be to place an http server that's more memory efficient in front of apache and have it server static content directly and proxy non-static requests to apache. Nginx would be great for this.

As a temporary fix you can also look into decreasing the MaxRequestsPerChild to 1000 or something more aggressive. Since apache processes tend to grow in size as they serve requests this will limit their size by killing them off and spawning new ones.

Apache process count spike bringing vm to its knees

semaphores were maxed out. i dont know why apache has allocated itself ~115 semaphores but upping the value fixed all the problems - apache no longer has to IOWAIT for another semaphore, so the system doesnt hang.

echo "kernel.sem = 512 32000 100 512" >> /etc/sysctl.conf; sysctl -p;

Best Answer

Related Solutions

Vps – Apache crashing VPS – need to reduce memory usage

Apache process count spike bringing vm to its knees

Related Topic