SUMMARY: Apache Process taking up 3 GB of ram

Thanks to all those who helped, especially John Eisenschmidt. I spoke with
our resident Solaris guru (who runs a system here with ~25,000 user
accounts) and he suggested lowering the MaxRequestsPerChild and
MaxKeepAliveRequests directives to around 100 each (from my previous
values of 300 and 200). 

The server is cycling through lots of http processes but the memory usage
is nice and steady with no spikes. The apache processes don't live long
enough to cause a problem anymore. 




Here's my original message:

I'm probably overreacting, but one of my apache processes which peaked at
25% cpu usage (per top) still has as its "size" and "res" memory (per top)
3.6GB and 3.1GB respectively and vmstat is shoing intermittent high values
(like 12976) in the "sr" field which according to the cockroft book is
bad. Is this a memory leak in apache? I went deep into swap (3 GB) until I
just killed that apache process, at which point all of its memory was
freed up. Everything is happy again.

This is on a 420r with 4gb of ram, 4x450 mhz running
apache/mod_ssl/mod_perl/weblogic. This app does some very bulky things,
like copy a 250MB directory on the filesystem to another directory at the
will of the user, repeatedly, all day long. The apache process got out of
control at about the same time that someone was doing this copy.

One problem is that I'm just using the internal 36gb drives, mirrored with
disksuite. I'm sure things would be faster with an A1000. At the rate that
disk space is disappearing, I'm going to need something very soon.

Anyway, what could have caused that apache process to take over all of the
ram and a lot of CPU? Is there a workaround?


Dave Lowenstein
Programmer/Analyst
Instructional Technology Services
San Diego State University
(619)594-0270
http://www-rohan.sdsu.edu/dept/its
_______________________________________________
sunmanagers mailing list
sunmanagers@sunmanagers.org
http://www.sunmanagers.org/mailman/listinfo/sunmanagers