Bind 9.2.3 very, very slow starup

Eric Dynowski bindlist at ip-hotel.com
Tue Jul 13 01:37:56 UTC 2004


Hello,

We are having a major problem with our bind 9 server. When restarting our
server it can take hours for it to load all the zones
and start answering queries. We have discovered that killing the named
processes, and restarting it will eventually result in a fast
load of all zones (less then 1 second). However, sometimes we only need to
kill and restart 2 or 3 times, while 40 or 500 times other times.

When restarting the slowness begins at random points during the load for
each restart, sometimes after only loading one or two zones,
other times after loading 50 zones..

The only common thread we could discover is that if we restart the server
late at night.. 12am, 1am, the restart is almost always successful
the first time. This led us to think it was related to network traffic, so
we tried shutting down the network interface and restarting bind, but the
load was still slow.
Rebooting the machine does nothing.

I was beginning to think it could be related to zone data, so today we
also tried removing all the zones but one, with still no luck.

The server is master for about 35 zones, and slave for about 65 zones with
a total of about 14,000 records, running on a Compaq Proliant
DL380 733Mhz with 384MB of RAM under Debian 3.0, kernel 2.4.18.

I have tried different kernels up to 2.6, I have recompiled bind with and
with out thread support, run it with debugging, etc, and cannot
find any cause for the hang up. Given a restart that is slow, after
several hours of zone loading, named will eventually begin answering
queries and responding normally.

Please help!

Thanks

Eric Dynowski



More information about the bind-users mailing list