Upgrade to bind 9.3.1

[grahamcrabb at gmail.com]

Hi,
   we recently upgraded bind on a fairly heavily used dns server (used
as a secondary for thousands of domains) from 9.2.3 to 9.3.1 all seemed
to go well, it pulls all the secondary domains from its primarys with
no problems. It seems to respond to digs no problems. However we get
reports from people that queries to this server are timing out. The
logs dont show any real problems apart from the usual odd axfr's from
primarys that arent working and us getting notifies from non masters.
On looking at what is causing the problem I ran some vmstat statistics
and saw the following:

Bind 9.2.3

procs -----------memory---------- ---swap-- -----io---- --system--
----cpu----
 r  b   swpd   free   buff  cache   si   so    bi    bo   in    cs us
sy id wa
 5  0   2148 140920 169420 1263988    0    0     0     0 19109 50602 42
26 32 0
 2  0   2148 140920 169420 1263988    0    0     0     0 19395 50953 46
24 31 0
 2  0   2148 140916 169428 1263988    0    0     0    84 19283 51063 45
22 33 0

Bind 9.3.1

procs -----------memory---------- ---swap-- -----io---- --system--
----cpu----
 r  b   swpd   free   buff  cache   si   so    bi    bo   in    cs us
sy id wa
 3  0   2148 361540 182296 938144    0    0     0     0 13713 187986 27
27 47 0
 2  0   2148 361544 182296 938144    0    0     0     0 13221 187670 28
27 45 0
 1  0   2148 361552 182296 938140    0    0     0     0 13468 189196 28
25 47 0

As you can see the Context Switching seems to of gone through the roof
(about 4X)

The server is a dual pentium 2.4GHZ with 4gb of Memory and bind is
using approx 2.2 gb of memory.

I compiled bind with --enable-threads option (because the machine is
dual cpu)

However I just noticed:

when 9.2.3 starts it reports:
using 4 CPUs

When 9.3.1 starts it reports
found 4 CPUs, using 4 worker threads

I wasnt at the company when the original bind was compiled so dont know
with what options it was setup.

Was I correct in compiling bind with --enable-threads and if not could
this be causing my problem.

Your help would really be appreciated.

Graham.