Bind9 Crazy-high CPU on Linux

[Tony Earnshaw]

Schlosser, Matt D. wrote:

> Specs:
> P3 1400mhz
> 2GB RAM
> Debian Sarge (3.1), also tried Etch (4.0 beta)
> Bind 9 9.2.4-1sarge1 (9.2.4 with debian patches) also tried 9.3.2-P1.0.1
> (9.3.2 with debian patches)
> Kernel 2.4.25 and 2.6.8-2 (2.6.8 with Debian patches)
> 
> There are two of these machines with the following issue:
> 
> The machines run between 800 and 1,000 queries/second for both
> authoritative and recursive zones.  After 12-24 hours, the CPU will
> spike to 100% and sit there while the machine times out any more
> queries.  The only resolution is to restart bind.
> 
> After exhausting efforts to localize the problem (kernel, nic drivers,
> Bind version, configs, etc.), we installed Windows 2003 and Bind 9.3.3
> for Windows on one of the servers and copied the configs and zone files.
> The server has been up for 4 days with no problems.
> 
> We prefer not to run on Windows mainly because our entire DNS
> infrastructure is Linux-based and we would have to redevelop our
> provisioning and management systems which is a rather large undertaking.
> We won't get budget help to throw more horsepower at it since Windows is
> a solution.  Besides, we are not convinced that will help.
> 
> As anyone seen/resolved this before?

Not seen with Red Hat - even with an enterprise DNS service with servers 
in 2 countries, so not resolved ;) - but could you perhaps consider 
trying an other Linux vendor than Debian?

--Tonni

-- 
Tony Earnshaw
Email: tonni at hetnet.nl