[lug] Memory problem

[Nate Duehr]

This is farily normal in my experience with big BIND servers.

BIND will use up all the physical RAM available unless you limit
it's cache size.  It doesn't seem to hurt anything.

As long as swap isn't filling up, you're fine.  The kernel will play
traffic cop and swap out unused BIND information when you run other
apps.  If you're worried about the performance of those apps or you're
seeing a lot of paging,  you can limit BIND's use of RAM with the 
"datasize" and "stacksize" directives in the "options" section.

(plug) Evi's book has a nice discussion of RAM/resource utilization that
talks about how to look at this stuff.  I had similar concerns until I
dumped some stats and graphed them.  My machines weren't paging too bad,
and some of them are dedicated nameservers anyway, so I gave BIND the
"datasize unlimited" and "stacksize unlimited" directives to make sure
BIND wouldn't limit itself on what it asked the kernel for... and I let
the kernel make all the RAM utilization decisions.  

Another interesting thing I've seen is that somehow BIND appears to be
written (not being a C coder -- I don't know how) to avoid swap.  The
machines only seem to swap when other apps besides BIND run.

Hopefully that's helpful.

On Tue, Nov 28, 2000 at 12:12:29PM -0700, Justin wrote:
> I'm running into a memory issue on my main server machine. I have never had
> this problem before and it seemed to start within the last 2 weeks. Using
> procinfo I am seeing the memory usage gradually increasing until it reaches
> the max of 128megs. Last week I started killing off process's randomly to see
> if that free'd memory but no luck. I tried restarting all my main server apps
> including apache, postfix, identd, syslog, inetd, and some others. The memory
> usage still showed the same, but when doing a ps aux none of the process's
> seemed to be hogging any memory. I ended up shutting down and booting back up
> to clear the memory problem. After the reboot the mem usage, according to
> procinfo, was about 50megs. After my users had started up all their stuff
> again the memory was up to about 90megs, which is normal. I ran my
> counterstrike server for a while saturday night, which consumes about 28megs,
> then shut it down later on. The memory dropped back down to around 90megs
> after that but has been increasing daily for no particular reason. It's now at
> about 122megs of the max 128, and the only thing I've done since the weekend
> was update my BIND to P7. Any way to trouble shoot this or any ideas of what
> might be going on? The box is Redhat 6.1 (updated) w/ kernel 2.2.15. Thanks in
> advance for any help.
> 
> Justin

-- 
Nate Duehr <nate at natetech.com>

GPG Key fingerprint = DCAF 2B9D CC9B 96FA 7A6D AAF4 2D61 77C5 7ECE C1D2
Public Key available upon request, or at wwwkeys.pgp.net and others.