Hey Chris, On Nov 20, 2006, at 7:57 PM, Chris wrote:
Hi I use apache 1.3 as well as 2.2.3 so this is off topic since this is for 2.2.3 but I have just noticed my processes use in excess of 100meg per child, I run eaccelerator which I know accounts for some of this 32meg so this would leave around 80 meg per process without eacellerator, I always thought this was normal and would explain why on a box with 2 gig of ram I cant get anywhere near the amount of child processes others reach without using swap.
This happens because you are serving run-on sentences. You might try breaking up your message into multiple, shorter sentences, so Apache doesn't have to keep your entire message in memory all at once.
I can't speak to eaccelerator because I don't know it or use it. It is possible that Apache allocates that much pool memory to serve requests and, after the request, the pool or its parent have grown by that much and will not shrink until they are destroyed. This can be an outright bug, or a memory leak, or a combination of both. I don't really use 1.3 anymore, but I don't think the core would do this so perhaps you might look at third-party modules. Are you perhaps serving very large files?
Also, please keep in mind that Apache allocates shared memory for the scoreboard, and depending on your implementation of top the text pages of the code and libraries it loads might also show up as shared memory. This memory is really only allocated once.
What can you do about this? Try to find out which requests or modules cause this behaviour, and from there try to find out what actually happens.
S. -- sctemme@xxxxxxxxxx http://www.temme.net/sander/ Open Source Software Consultant PGP FP: 51B4 8727 466A 0BC3 69F4 B7B8 B2BE BC40 1529 24AF
Attachment:
smime.p7s
Description: S/MIME cryptographic signature
|