Why Might 2 out of 3 Bind9.1.3 Systems Fail a Lookup?

[Martin G. McCormick]

	A customer of ours correctly reported that he couldn't
resolve cybershelter.org while on our network.  This happened
yesterday October 26.  I looked it up in whois and the data were
as they should be according to the customer who was familiar with
the domain.

	A third system on the same network, also running
bind9.1.3 never failed to resolve the domain.  While all 3
systems have somewhat different named.conf files, all are
supposed to resolve foreign domains normally with recursion on.
They all resolve thousands of domains daily with no trouble and
we get occasional peaks of 200,000 successful hits per hour.
Serious problems are usually noticed in minutes if anything goes
wrong.

	I even killed named on the one system that could resolve
it so as to restart a new cache.  It immediately began to
properly resolve cybershelter.org as soon as I restarted named.

	Today which is October 27, all three systems are
resolving cybershelter.org.

	What is going on?  I hate finger-pointing, but I didn't
change anything and it started working all by itself.

	The only difference I can think of is that the one system
that did resolve this domain gets frequent rndc reload commands
throughout the day while the other 2 don't get this treatment
very often and simply receive updates from the third system.
Would that normally have any effect on outside domain resolution?
All three systems use the same named.ca file.

Martin McCormick WB5AGZ  Stillwater, OK 
OSU Center for Computing and Information Services Network Operations Group