bind and glibc issue?

[Darron Froese]

Howdy,

I've got a couple machines that have been giving me grief lately and had a
couple of questions to ask:

If you compile bind using one version of glibc and then update to a newer
version of glibc - could that make (seemingly) random queries fail? If that
could - could recompiling bind (with the new glibc) fix those errors?

Here's the Problem:

Updated my 2 dns servers to 8.2.2p5.
Updated my mailhost and webserver (both running yellowdoglinux-1.0) to a
newer version of glibc so that I could remove ssh and install openssh
(because of the ssh/RSARAF exploits).

Old glibc: glibc-2.1-1a.ppc.rpm
New glibc: glibc-2.1.3-0j.ppc.rpm

Two days after the update, I started receiving email bounces like this:

> The original message was received at Tue, 28 Dec 1999 18:06:23 -0700
> from [209.91.88.219]
> 
> ----- The following addresses had permanent fatal errors -----
> <bill_richard at transcanada.com>
> 
> ----- Transcript of session follows -----
> 550 <bill_richard at transcanada.com>... Host unknown (Name server:
> gate2.tcpl.ca.: host not found)

Here's the error for this particular message in the /var/log/maillog:

Dec 28 18:06:23 mulder sendmail[13145]: SAA13143: SAA13145: DSN: Host
unknown (Name server: gate2.tcpl.ca.: host not found)
Dec 28 18:06:23 mulder sendmail[13145]: SAA13143:
to=<bill_richards at transcanada.com>, ctladdr=<bill at fudgehead.com> (501/502),
delay=00:00:00, xdelay=00:00:00, mailer=esmtp, relay=gate2.tcpl.ca.,
stat=Host unknown (Name server: gate2.tcpl.ca host not found)

It only happened about 5-7 times every other day (some days it would bounce
a few and other days it wouldn't bounce any) - but these have been rock
stable boxes since they were installed in July. The host *can* be found
without any problems - it's just during various emails that it fails.

Attempted Solutions:

1. I reverted my dns servers back to 8.1.2 - the last working version - that
didn't change anything.

2. Finally I had to downgrade to an older version of glibc (I had to use the
glibc-2.1.1-6h.ppc.rpm from ydl-1.1 as I couldn't find the 1.0 release
available anywhere) - that didn't fix anything either.

3. Now I'm in the process of recompiling bind (with the different and
slightly updated glibc).

Has anyone got any pointers? I've got no idea what to look at next and this
is driving me insane. Could anyone also point me to a site that can help to
explain how linux does its name resolution so that I could maybe trace the
path and the programs that are involved here?

Thanks for any help that anyone can offer.
--
Darron
darron at fudgehead.com

*Please* cc me directly if you respond in any way.