bind-9.3.2 / CPU issue.

[Kelsey Cummings]

On Fri, Aug 18, 2006 at 05:33:06PM +0200, bert hubert wrote:
> On Fri, Aug 18, 2006 at 11:23:31AM +0200, Daniel Ryslink wrote:
> > Strange thing is that we have tried to experimentally deploy an 
> > alternative nameserver with dnscache (part of djbdns) instead of BIND. The 
> 
> dnscache needs some tuning for high(er) load, so it may not be related.
> 
> When things go wrong, check the output of vmstat 1 to see of the load is
> 'sy' of 'us' (system, or user).
> 
> If the load is in 'sy', the problem might be kernel related.
> 
> Do you run with iptables enabled?

Just to get back to an old thread, the problem is not related to connection
tracking in Linux in our case (on, off, doesn't matter and as it showed
under both FreeBSD and linux...)  We've been trying to reproduce it in the
lab from query logs but haven't been able to do it yet which at least tends
to suggest it's not dependent on the queries them selves.

-- 
Kelsey Cummings - kgc at corp.sonic.net      sonic.net, inc.
System Architect                          2260 Apollo Way
707.522.1000                              Santa Rosa, CA 95407