DNS resolution problems...

[Robert Dinse <nanook@eskimo.com>]

     DNS, domain name service, is used to convert a hostname into a numerical
IP address, to convert a numerical IP address into a hostname, to find out what
servers handle mail for a particular domain, and many other things. 

     I'm seeing some problems with DNS failing on an initial query and then
after several subsequent attempts succeeding.

     My first suspicion was that something was wrong with a relatively new
name server here, but upon further testing I found every name server I could
try was doing this.

     I tested the various root name servers and I did find two of them to be
not responding.  Unfortunately, commenting them out isn't effective because the
way bind works is that it tries each name server in the list of root servers
until it can connect, and then it downloads a list of root servers from that
root server and discards the list it has.  So commenting out the bad ones
doesn't work because it still picks them up from the root name servers when it
downloads a fresh list. 

     I did ftp a current list from internic.net, and save for some changes to
the comments, it is exactly the same as the list I have.

     I am wondering if there is some generic flakiness released in BIND
version 8.2.3 as that is the version we are running on all machines.  It acts
the same whether it is hosted on a SunOS or Linux box, so it is not an OS
specific bug.

     DNS is really central to the working of just about everything on the
Internet so this is really a major problem and one I feel rather powerless to
resolve at this point.

     I can find nothing on the BIND website relating to this.  If anybody knows
of the cause and/or a fix, the information would be appreciated.