BIND 8.2.7 master ixfr to 9.2.2 slave

mayer at gis.net mayer at gis.net
Thu Apr 28 21:55:35 UTC 2005


----- Original Message Follows -----
> I have a BIND 8.2.7 master supplied by a third party (MetaInfo).  It
> is feeding zones to a 9.2.2 slave.  Occasionally the 9.2.2 system will
> complain about missing NS records for a zone and I have to delete the
> zone file from the slave and restart the nameserver.  An AXFR from the
> 9.2.2 system shows a zone with an SOA and one or two other records,
> when it should contain a hundred or so records.  After deleting the
> zone file and restarting 9.2.2, the AXFR shows the correct
> information.
> 
> I have a packet trace where the 9.2.2 system performs an IXFR query. 
> The answer section from the 8.2.7 system only contains two identical
> SOA records, with serial numbers one more than the query contained. 
> Immediately after the response the 9.2.2 system complains about the
> zone missing NS records and I have to restart the nameserver.
> 
> It is probably an error that the 8.2.7 system sends two identical SOA
> records, but the 9.2.2 system shouldn't truncate the zone.  I can't
> upgrade the 8.2.7 system because it's supplied by MetaInfo.  I can
> upgrade the 9.2.2 systems.  Was this bug fixed in 9.2.5?  I found this
> in the CHANGES file:
> 
> 1722.   [bug]           Don't commit the journal on malformed ixfr
> streams.
>                         [RT #12519]
> 
> Has anyone seen anything similar?
> 
> Mike Mitchell
> Mike.Mitchell at sas.com
> 

Don't use IXFR on BIND 8. It never quite worked right and it got
rewritten
3 times. It works correctly in BIND 9.

Danny



More information about the bind-users mailing list