question about caching of lame servers

[JINMEI Tatuya / 神明達哉]

>>>>> On Thu, 12 Oct 2006 15:46:48 +0200, 
>>>>> Klaus Darilion <klaus.mailinglists at pernau.at> said:

> I tried to test the caching of lame servers in bind9 (9.2.4-1sarge1), 
> and reviewing the source code, I came to the conclusion that this 
> feature is only used as long as there is at least on authoritative name 
> server which is not in the lame-cache. As soon as all authoritative name 
> servers of a certain domain are in the "lame" state, bind "restarts" the 
> lame settings and tries again although all servers are marked as lame.

> Is my conclusion correct?

That's basically correct.

> If yes, is there an option to change this behavior (if all servers are 
> marked as lame bind should not try at all)?

Not with 9.2.4, but BIND 9.4 caches the lame information per
<qname,class,type> and doesn't try any lame server (whether or not
there is a non-lame server):

1880.	[func]		The lame cache is now done on a <qname,qclass,qtype>
			basis as some servers only appear to be lame for
			certain query types.  [RT #14916]

I think this implements what you want to see.

(Note: the behavior in 9.2.4 made sense because there are known
(broken) nameservers that behave as being lame only for certain query
types.  See Section 2.2 of draft-ietf-dnsop-bad-dns-res-06.txt)

> Further, I not only want to cache lame name servers, but also name 
> servers which are down. Is this possible?

Not exactly, but the fact that a server is down is cached as
a penalized RTT, which makes that server less preferred in subsequent
server selection.

					JINMEI, Tatuya
					Communication Platform Lab.
					Corporate R&D Center, Toshiba Corp.
					jinmei at isl.rdc.toshiba.co.jp