Cache strangeness

[Mark_Andrews at isc.org]

> Hello all,
> 	I'm running 8.3.3-REL on several sun systems.  I have turned off
> nscd on these systems as well.  What I have been seeing is a strange train
> of events repeated a few times.  Another domain will change the
> forward/reverse maps for a host(in this case www.hamiltoncompany.com) and
> their primary reverse dns server will also fail.  The end result in this 
> case was that the forward pointed to 207.13.227.54(new ip) but the reverse 
> failed a lookup.  The problem came when one of their reverse servers came
> back on line, but bind had somehow "cached" the bad answer, and bind
> continued to return no such host errors on lookups.  I restarted bind to
> clear the cache, and the first lookup returned 1 timeout, then found the
> correct info and now works.  Is there some way cached answers are looked up
> differently than uncached ones, with the later trying more servers?
> 
> Please advise.

	Named supports negative caching (caching of NXDOMAIN and
	NODATA responses).  The length of time this will be cached
	for is controlled by the servers for the reverse zone (see
	RFC 2308).  Named itself will put a upper limit on the ttl
	of negative answers it receives (see max-ncache-ttl, default
	3 hours).

	Mark
> 
> 
> --
> -----BEGIN PGP SIGNATURE-----			Robert Weber
> <!-- |**|begin egp html banner/			UnixOps/ITS
> > Thank you for registering There is no catch.	University of Colorado 
>  
> > this e-mail blah blah confidential and yadda yadda legally privileged   
> 
--
Mark Andrews, Internet Software Consortium
1 Seymour St., Dundas Valley, NSW 2117, Australia
PHONE: +61 2 9871 4742                 INTERNET: Mark.Andrews at isc.org