bind-8.2.3-T2B after zone transfer error

[Mark.Andrews at nominum.com]

> >> Received no replies to this one.  In retrospect I confounded the issue
> >> slightly by the twaddle at the end relating to "_"s.
> >> 
> >> However we have seen another zone problem with the ac.uk zone today
> >> (server is currently running 8.2.3-T5B) ...
> >> 
> >> May 23 15:24:13 cblslca named[80]: zone: ac.uk/IN: non-glue record below b
> ott
> >> om of zone: ies.lmu.ac.uk/NS
> >> May 23 15:24:13 cblslca named[80]: zone: ac.uk/IN: non-glue record below b
> ott
> >> om of zone: ies.lmu.ac.uk/NS
> >> May 23 15:24:13 cblslca named[80]: zone: ac.uk/IN: non-glue record below b
> ott
> >> om of zone: mis.lmu.ac.uk/NS
> >> May 23 15:24:13 cblslca named[80]: zone: ac.uk/IN: non-glue record below b
> ott
> >> om of zone: mis.lmu.ac.uk/NS
> >> May 23 15:24:14 cblslca named[80]: slave zone "ac.uk" (IN) rejected due to
>  er
> >> rors (serial 2000052305)
> >> 
> >> the previous serial which was ok was 2000052301.
> >
> >	This is the result of a bug in the master
> >
> > 990.   [bug]           we could loose track of a bottom of zone cut if the
> >			write buffer filled up at just the correct moment.
> 
> Oops.. I neglected to update the Subject: the problem spotted yesterday
> was spotted with 8.2.3-T5B which included 999 according to CHANGES.
> 
> >> Obviously the contents of the ac.uk zone file is a matter for the
> >> maintainers of ac.uk but what is more concerning to me is that after
> >> one of these failures to load we immediatly start to see errors for our
> >> own zone lookups (the server is slaving leeds.ac.uk)...
> >> 
> >> May 23 15:24:34 cblslca popper[23286]: warning: can't verify hostname: get
> hos
> >> tby
> >> name(food-pc81.leeds.ac.uk) failed
> >
> >	I really believe that this is totally unrelated to the purging of
> >	non-glue.  The basic code for purging of non-glue uses the same
> >	framework as the general purging of a zone when reload / removing
> >	a zone, i.e. they use the same function to detect when they have
> >	encountered another zone in the database.  The only way this
> >	function fails is if there is an invalid zone with no SOA record.
> 
> Yesterday (and previously when I hit the same problem with T2B - my
> previous message unfortunately pointed the finger towards the "_"s)
> the gethostbyname failures started immediatly after the error message
> and it was necessary to update the named.conf and stop/start named.
> 
> Andrew
> 
> 
	On further reflection I can see how
	gethostbyname(food-pc81.leeds.ac.uk) could have failed.

	If the C / resolver library on the machine in question is
	not RFC1535 aware and the search list includes "ac.uk"
	then the search would have been aborted at
	food-pc81.leeds.ac.uk.ac.uk.

	Mark

--
Mark Andrews, Nominum Inc.
1 Seymour St., Dundas Valley, NSW 2117, Australia
PHONE: +61 2 9871 4742                 INTERNET: Mark.Andrews at nominum.com