Bind 99% problem still occurs with 9.2.3 and 9.3.0

[phn at icke-reklam.ipsec.nu]

Michael Lang <michi+bind-users at relay3.jackal-net.at> wrote:
> Hi List,
> i still can reproduce 99% CPU Usage for named process on a Intel(R)
> Pentium(R) 4 CPU 2.26GHz, with 1 GB Ram, doing nothing else than
> named.

> The Machine is setup for Resolving and Zoneholding, it has 2 views
> configured. One for customers and one for internal resolving, with
> limiting the max-cache value for the customer view to 200M and the
> internal to 8M.

> The Machine works fine after startup. Ive seen the problem reproduceable
> after one week of production (it sure happens after one week).
> on Monday 15 Nov 2004 i start the named process, with 4533 Zone
> configured localy. It works find response times are as the should be.
> on Monday 22 Nov 2004 i try to reproduce the problem as seen with 9.3.0
> on the same machine without changing anything else except the Bind and
> its librarys.
> Therefore i do a cache cleanup with rndc flush.
> The Machine starts using all the CPU, and dont get back to normal life.
> I´ve attached two traces of the process and the MEMusage and the
> named.stats. The kernel was already upgraded (for both version 9.3 and
> 9.2) to the latest Redhat Enterprise 3.0 2.4.21-20.EL. 

> Does anyone know howto solve this ?
> (my next debugging step is to do a flush every day)


Is there enough free memory ? Is any part of named swapped ?
is the swapdevice "active" ?  Any messages in syslog that may be
related to resource shortage ? Network errors ?



> -- Binary/unsupported file stripped by Ecartis --
> -- Type: application/x-bzip
> -- File: debug.bz2




-- 
Peter Håkanson         
        IPSec  Sverige      ( At Gothenburg Riverside )
           Sorry about my e-mail address, but i'm trying to keep spam out,
	   remove "icke-reklam" if you feel for mailing me. Thanx.