BIND 8.3.4 / Solaris / Error loading zone

[Erik Beebe]

Hi all;

I have an odd (to me) issue.  I had a nameserver running under Solaris 2.6
with BIND 8.2.2P7, and upgraded to one running BIND 8.3.4 under Solaris 8.
I have ~900 zones (number of zones allocated: 896 according to ndc).  I
copied all of the zones over from the old nameserver and started up 8.3.4.
No apparent problems at first, but.. I have one very large zone, called
(assume it's called 757.org).  This zone has about ~1150 RRs in it.  This
zone loads just fine when the server starts up:

24-Feb-2003 21:32:51.497 load: master zone "757.org" (IN) loaded (serial
2003022413)

Now that's super.  But after a little while, I'll see this:

24-Feb-2003 22:27:58.368 load: master zone "757.org" (IN) rejected due to
errors (serial 2003022413)

What?!  After the server loads the zone properly, and reports that it is
loaded, I can query the nameserver for a RR in that zone, and receive a
correct reply with the AA flag set.  Once the server rejects it, I can no
longer query information in that zone.  Odd, I tell you.  There are _no_
other complaints about that zone file (or the syntax contained inside it)
logged.

Now here is part two (seems likely related):

I compiled BIND with a prefix of /opt/BIND, and then installed everything
into /opt/BIND-8.3.4, and made /opt/BIND a symlink to it.  Therefore, my
ndc control file would end up in /opt/BIND/etc/ndc.d.

I can do a 'ndc reload' with no problems.  However, if I try to 'ndc
restart', after a few seconds I'll get "ndc: error: name server has not
started (yet?)".  If I check 'ps', it looks like named IS running, and the
control channel becomes available again in a few seconds.  If I check the
logs, there are no errors.  I just see the server stop/start.

Sorry for the long message, but it really is fairly puzzling.  I checked
the usual (google and bind-users archives), but found no answers so far.
I'm wondering if I may have some file descriptor issue limit issue?

Thanks in advance,
Erik