SUMMARY: dns failover problem

Many thanks for your prompt and useful replies

Dan Astoorian, John Riddoch and Brett Lymn all provided the correct
information, and that is:

nslookup is flawed. it is a troubleshooting tool and doesn't use
the same resolver mechanism that is used when a normal
dns lookup is being performed. It uses only one server at a time-
the first one listed in resolv.conf.

as I was testing a failover configuration for our dns, it became
apparent that I should use something that mimics the dns lookup
process more closely.

"getent hosts some.box.name" was the appropriate command
and did produce the desired results. when named was running
on box 1.2.3.4 and getent hosts www.some.name was entered,
the appropriate ip address was returned. I killed the named pid
on 1.2.3.4 and getent hosts www.some.other.name came back
with the proper ip address.

Pat Novak


Original email:

>I have 2 dns's, joined by a hub, set up as master and slave.  both have
each
>other
>entered as default routers, and both have each other listed in the
>resolv.conf file as
> second nameserver entry, and themselves as the first.  The slave finds and
>updates
>from the master as expected.

>the box at ip 1.2.3.4 has resolv.conf entries of:

>domain some.domain.com
>nameserver 1.2.3.4
>nameserver 1.2.3.5

>and  box at 1.2.3.5 has:
>domain some.domain.com
>nameserver 1.2.3.5
>nameserver 1.2.3.4

>if,  on dns A,  i say "nslookup   some.box.name"      it resolves.
>if,  on dns A,  i say "nslookup   some.box.name   dns.B.ip"  it resolves.
>and vice versa

>however, if i test a failover on box A by killing named-pid, then do:
>"nslookup some.box.name"  i get no response and truss shows the
>process sleeping. i get the same result going the other way.
>has anyone else seen this,  or have i missed something obvious.


>tia
>Pat Novak
_______________________________________________
sunmanagers mailing list
sunmanagers@sunmanagers.org
http://www.sunmanagers.org/mailman/listinfo/sunmanagers