Seems that by following your tutorial on performance tuning, I mistakenly changed the instead of the one you suggested. Now whenever I hit restart in debian squeeze for apache2 and mysql, everything seems ok.

Except when I goto my site on the net where there is justy a blank screen.

NOW what do I do?

So you wrote a blog post about something that seemed fairly innocuous, but for whatever reason, it caught the attention of one of the major sites and now your server load is at 110 and climbing, the ssh command line session is taking thirty seconds to …

I have to disagree here. The control method for that scenario is KeepAliveTimeout. I would suggest changing the default 15 to much lower, such as 2 seconds. This prevents the server from keeping a process reserved for 15 seconds.

Forking processes can put a serious load on your server. It’s dog slow, and really bogs down your server especially if it has to fork dozens of processes fast, like when you’re slashdotted. You only want to tear down processes when they’re leaking memory, which you mentioned. Therefore, I have found the following pattern works very well.

Set MinSpareServers to around 20 so you have enough processes started and waiting to serve. Set MaxSpareServers a bit higher, such as 30. Set MaxRequestsPerChild to 1000 or something high, so you only kill processes when they done plenty of work. MaxClients should be set to the amount of processes your apache can run without swapping, and definitely not too high. You can calculate this number by dividing the amount of RAM available to Apache by the maximum memory used by any Apache process. ServerLimit is set automatically and can be removed from the config.

Also, do not underestimate the effect of a byte-code cache such as APC, or Zend Optimizer+. It will avoid the PHP engine having to compile your PHP application for each and every request and will speed up each request quite a lot, at the expense of some memory. It wil help a lot.

The rest of your tips are great, so thanks for the article.

Some more information about caching on wp can be found over there @ ask apache.

So you wrote a blog post about something that seemed fairly innocuous, but for whatever reason, it caught the attention of one of the major sites and now your server load is at 110 and climbing, the ssh command line session is taking thirty seconds to …

Yup Mats, it’s the swap that kills you every time! That’s the idea behind ServerLimit – don’t let Apache run too many processes and trigger that first swap, because if it swaps once, odds are that that’s the first stumble and a few minutes later the whole server faceplants at speed.

Varnish is something I’ve been talking about with Conor (the sysadmin for boards.ie) actually, but it’s not really a three-minute first aid thing. Even memcached is a bit of a stretch for the first aid mindset 😀