Server load issues

[wjdennen]

Hi, we're using version 11.25, running apache 2.2 prefork.

Since putting this server into production last week, we've had many server load spikes. The server will be performing fine, and then out of the blue, the load spikes dramatically. The access and error logs do not indicate a spike in traffic, though.

I've checked for slow mysql queries (nothing). Memory is not a problem.

When this happens, there are many many apache processes running.

The load has gotten as high as 200+ at which point we have to restart the server (or kill -9 htttpd) to get it to respond.

Then, upon restart, things are fine... until the next time.

I've read about others having problems with 11.25 and the log processing.

Could that problem be causing the problems we're seeing?

Thanks for any advice

[Spiral]

It sounds to me that your Apache isn't properly optimized (and perhaps your MySQL) though I couldn't say for certain to what degree without looking at your server but based on what you said, I suspect that you either have requests coming in faster than your system has been configured to handle or have too many handlers setup wasting resources. Either way does sound like you aren't using your resources efficiently. As for your question about "log procession", yes that can be an issue but is more a case when dealing with the logs of "high traffic" sites where the logs tend to grow more quickly in a short amount of time.

Out of curiosity, do you know your current Apache settings that you are running against the MPM prefork that you just mentioned? Or at least your server specifications (cpu, memory, number of sites, etc) ....

With that, I might be able to tell you something a little more targeted.

[wjdennen]

VPS (1.5GB RAM, burstable to 3.0GB)

Apache 2.2
StartServers 5
MinSpareServers 10
MaxSpareServers 20
MaxClients 200
MaxRequestsPerChild 4000
KeepAlive OFF

Load is generally low - about 0.5. However, when the spikes occur, the load skyrockets to 150-200. At those times, we do not see an increase in the apache logs, though.

Memory use is relatively low, too.

The server is currently hosting one domain. A vast majority of the web pages it is hosting are static pages. We are running WordPressMU with WP Super Cache enabled.

This site was recently hosted on a similarly-sized VPS for a few years without any problems. The old site is running apache 1.3.

We moved to the new site a week+ ago and have had problem since.

I turned off the pre-release eaccelerator last night and we have not had any spikes since. Time will tell.

[johnburk]

Do you have a lot of commands run by nobody?

The command being /usr/local/apache/bin/httpd -k start -DSSL ?

I know some people have had problems with this since upgrading to 11.25, what happens is that the command runs indefinitely till you read 200+ loads within a few minutes!

[wjdennen]

Yes, when we had (have) the problem, there are lots and lots of those processes. However, since disabling Eaccelerator yesterday, we've been good. (Knock on wood)

[wjdennen]

I found this bug report over at the Eaccelerator site:

#368 (hung httpd threads with high page request rates) - eAccelerator - Trac