DNS resolver problems in old version of Unix

Sheldon Glickler sheldong5178 at my-deja.com
Tue Dec 28 18:23:04 UTC 1999


This has been just posted independently in comp.unix.admin
==========================================================

I tried searching Deja News but was unable to find help there.  So here
goes, maybe someone can help me.

We are running an old version of Unix.  In is Unixware 1.1.2.  There is
no support out there, and we will be changing soon.  Right now,
however, we are stuck with it.

On our two servers we are running named and have two independent
ethernets going into each server.  Thus we have four possible targets
for our resolver.

Basically there are two problems.
(1) I know of no way to change the initial timeout period for the
resolver.  Currently it is 6 seconds and I want to reduce that to one
second.  How can I do that?
(2) The books, and the man pages, say that when multiple nameservers
are listed in resolv.conf that the processing is at the first timeout
setting for the first nameserver, then the second nameserver, etc.  It
then wraps and tries the first nameserver with double the timeout, then
then second, etc.  However, the observed performance (I killed the
first nameserver named and did a "nslookup -d machine_name" showed that
the order is the first nameserver, then at double the timeout, then
double that, then double that and then the second nameserver, etc.  How
can I fix this to the proper behavior?

All our processes in our application have timeouts of about 15 seconds
or so.  Thus, the second nameserver is never looked at if the first is
down or hung as those processes, themselves, timeout.

Can anyone give some advice (other than get a new OS)?  In addition to
posting here, an email reply would also be greatly appreciated at
sheldonlg@(my service provider is listed below).

Shelly



--
Please reply to sheldon + lg at earthlink.net.


Sent via Deja.com http://www.deja.com/
Before you buy.



More information about the bind-users mailing list