name server on esx server and ms dns slow recursion

[chris]

I have a very complex issue and I'm needing some help.

Due to the DNS poison cache thing, we need to update our 4 dns
servers. They are currently on SLES 10 and 9. 3 are VMs in ESX
environment. 1 is a physical box (ns1).

Our subscription to Novell has expired and we figured why not tech
refresh to Ubuntu and save the $$. So I built a Ubuntu x64_86 VM on
our dev environment. It worked great.

I migrated it out over time to become NS2, 3, 4. I then switched the
physical box off and made it a VM as well. I was told that they had
originally had NS1 as a VM but had issues with having slow recursion
lookups. So they moved it to physical and it was fine then. NS1 being
the master obviously. Anyway... didn't seem to have that issue here
with Ubuntu.

About 2 hours later after the bring up of the new VMs, we started to
get user complaints of 'slow internet'. Digging into it, looked like
DNS was slow to resolve.

The way our network works, is the following.

Users will send query to a local Microsoft DNS server which will
forward to the root MS DNS server for the domain which will forward to
the Ubuntu boxen. The forward between the MS root servers and the
Ubuntu boxen was where it was slow. Queries taking 2-4 seconds to
return.

If we took users and pointed to the Ubuntu boxes themselves, they were fine.

I backed out and shut down all my VMs and brought the physical NS1 and
the old VMs back online and the users were happy.

I honestly can't see this being an Ubuntu issue or a VM issue. But I
suspect something with the 9.4.2-P1 version of bind and MS DNS. We're
only doing around 300 queries/sec. No where near the 10k saying that
there are issues on isc's site.


I've Googled around and not gotten much back. I'm hoping someone may
have some ideas as to what is the cause or if someone has seen this
before b/c I feel like I'm loosing my mind.

Thanks in advance.