Jeff Lightner wrote:
> I saw this on my Linux system when I had configured Bind ACLs to use the
> internal NIC instead of the external NIC even though most queries were
> to come through the external. Changing to the external NIC resolved it.
>
> -----Original Message-----
> From: bind-users-bounce@isc.org [mailto:bind-users-bounce@isc.org] On
> Behalf Of Mike O'Hara
> Sent: Tuesday, June 27, 2006 1:11 PM
> To: comp-protocols-dns-bind@isc.org
> Subject: Bind CPU ussage issue
>
> I am having a helleva time managing Bind on my server.
>
> Server Specs are
>
> Dual Xeon 2.66 GHZ
> 1 Gb ram
> Windows 2003 Server
> ISC Bind 9.2.2
>
> Now Bind will regularly spike up to 100% CPU usage and stay there
> untill the named service is restarted. then it will work its way back
> up there again after a non descript period of time.
>
> Is there a configuration glitch that is causing it to eat CPU cycles
> like they are going out of style? or will upgrading to the latest
> release solve my issues?
>
>


I used to run Bind 9.2 and Bind 9.3 on

cat /proc/cpuinfo

cpu family : 4
model : 0
bogomips : 18.94

cat /proc/meminfo

total: used: free: shared: buffers: cached:
Mem: 14737408 13434880 1302528 3059712 1413120 6639616
Swap: 134176768 2469888 131706880
MemTotal: 14392 kB
MemFree: 1272 kB
MemShared: 2988 kB
Buffers: 1380 kB
Cached: 6484 kB
SwapTotal: 131032 kB
SwapFree: 128620 kB

Linux version 2.2.19 (root@fm)
(gcc version 2.95.3 20010315 (SuSE))
#15 Sun Nov 24 22:34:59 CET 2002

I had similar experiences.

Now I am running BIND 9.4.0a6

on

cat /proc/cpuinfo

cpu : 486
model : 486 DX-25/33
vendor_id : GenuineIntel
bogomips : 18.94

cat /proc/meminfo

total: used: free: shared: buffers: cached:
Mem: 14540800 12181504 2359296 2285568 1343488 4046848
Swap: 133885952 12746752 121139200
MemTotal: 14200 kB
MemFree: 2304 kB
MemShared: 2232 kB
Buffers: 1312 kB
Cached: 3952 kB
SwapTotal: 130748 kB
SwapFree: 118300 kB

Linux version 2.0.36 (root@lomi)
(gcc version 2.7.2.1) {Unifix 2.0}
#13 Die Jun 20 18:59:19 CED 2006


It is the same kind of hardware. I downgraded back to linux 2.0
The problem seems to be the ethernet driver on linux 2.2 and 2.4
I tried 2.6 on a different hardware (windows + CoLinux), same problem.

I guess it is about duplex full or duplex half.

I have do have fewer problems on a switch than on a hub.

The problem is not only with Bind but with other programmes too, like
telnet and ssh, even ftp.

The problem is not Bind.

Nevertheless there are moments when Bind takes a nap:

When bind gets an updated zone file from a master (authoritative only)
or when it does a garbadge collection on memory (cache only).

Having only 16 MB really and 128 MB virtual, that is not much of a problem
on my small machine.

I am not running a big company, but only my small home network with a
workstation and three general purpose servers, all linux.

The machine running Bind is doing other work too. Im doing a lot of
DNS queries on both machines. I dont see any more cpu 100% on the
linux 2.0 but on the 2.2 it is unchanged.

OT: djbdns shows the same behavior. That is why I am shure the problem
is not Bind.


Regards
Peter and Karin

--
Peter and Karin Dambier
Cesidian Root - Radice Cesidiana
Graeffstrasse 14
D-64646 Heppenheim
+49(6252)671-788 (Telekom)
+49(179)108-3978 (O2 Genion)
+49(6252)750-308 (VoIP: sipgate.de)
mail: peter@peter-dambier.de
mail: peter@echnaton.serveftp.com
http://iason.site.voila.fr/
https://sourceforge.net/projects/iason/