[52] in Hesiod
Re:
daemon@ATHENA.MIT.EDU (Theodore Ts'o)
Wed Jul 10 21:22:39 1991
Date: Wed, 10 Jul 91 21:11:32 -0400
From: tytso@ATHENA.MIT.EDU (Theodore Ts'o)
To: swartp@prl.philips.nl
Cc: hesiod@ATHENA.MIT.EDU
In-Reply-To: Paul Swart's message of Fri, 5 Jul 91 10:09:14 NED,
Reply-To: tytso@ATHENA.MIT.EDU
Date: Fri, 5 Jul 91 10:09:14 NED
From: swartp@prl.philips.nl (Paul Swart)
Problems with a secondary BIND/Hesiod server.
We try to start up BIND/Hesiod in a configuration with small
databases, a database corresponds with a zone, and a
primary and several secondary servers. At this moment we
have a local working 'named' server (version 4.8), but
the problem we are facing at the moment is a failing refresh
of the databases of the secondary server. The connection with
the primary is established correctly, but then the 'SOA' and
'AXFR' queries fail both, and the result is a failed zone transfer.
The problem is that BIND is not really set up to do zone transfers for
anything other than class IN; what was happening was that zone-transfer
program tried to query for the SOA record using class IN, and since the
SOA record was for class HS, it failed.
BIND also has a bug that once it starts to do a zone transfer, it
transfers everything underneath the requested DNS zone, no matter what
class it is. It will also stop its recursive descent of the namespace
when it sees a NS record of any class. This last problem causes two
problems. First of all, too much data is transferred in a zone
transfer, since both class HS and class IN data will be transferred.
Secondly, if the DNS zone boundaries are different for class HS and
class IN, some information may not be transferred since it will stop
when sees a class IN NS record.
Patches to BIND 4.8.3 to fix both of these bugs can be found in
~ftp/tytso/hesiod-xfer-changes and ~ftp/tytso/more-hesiod-xfer-changes
on NET-DIST.MIT.EDU. As always, these patches are distributed with an
AS-IS basis, with no guarantees of success; as far as I can tell they
should work, though.
- Ted
