Subject: Re: Excessive swapping / Memory problems
To: Vincent van Scherpenseel <firstname.lastname@example.org>
From: Chuck Swiger <email@example.com>
Date: 09/06/2006 14:17:02
On Sep 6, 2006, at 1:49 PM, Vincent van Scherpenseel wrote:
> 11518 www -4 0 24M 4556K semwait 0:08 0.00% 0.00%
> 27634 www -4 0 3104K 2972K semwait 0:00 0.00% 0.00%
> 27635 www -4 0 3100K 2948K semwait 0:00 0.00% 0.00%
> 27817 www -4 0 2936K 3592K semwait 0:00 0.00% 0.00%
> 355 root 2 0 2704K 408K select 0:12 0.00% 0.00%
> Resident has increased quite a lot since the last snapshot (about
> 35 minutes ago). There are indeed some cronjobs (mostly a perl
> script fetching weather information for me) but I don't think
> they're causing this, since they have nothing to do with httpd.
An initial size of httpd at 3MB seems entirely reasonable; if they
continue to grow over time, that suggests a memory leak within one of
the modules you are using, or maybe within NetBSD's libraries. See
whether setting "MaxRequestsPerChild 100" in your httpd.conf helps
control the growth of httpd process size.
> The problem of processes getting killed happens just every now and
> then when the system is out of swap. Just when I'm doing nothing on
> the system (not even logged in!) but also when I'm compiling etc.
> So is there no chance this is hardware related? Because somehow I'm
> really looking into that corner.
No, there is no chance this is hardware related. You're running
Apache, MySQL, random Perl scripts, and apparently trying to compile
or do other stuff, on a machine that simply doesn't have enough RAM
or swapspace to handle so many tasks at once....