Conference turris::digital_unix

9021.0. "BIND and ASE service IP address + BIND and alias" by EVTAI1::9dhcp21.evt.DEC.COM::DORNANO () Tue Mar 04 1997 06:08

Hello

I'd like to come back on a well known subject : IP alias on an
interface of a BIND server.
I already read the whole conference about this subject, but I need
some clarification.

Should one put in named.boot the alias name and address and the
reverse stuff (as seen in note 5887.1) or restarting named after
putting the alias on the interface is enough (as stated in 5887.3) ?

I guess the first method is the good one, isn't it ? In this case,
should the patch for named be useless ?

Now, with the second method (assuming there is no declaration of the alias
in named.boot), how is named aware of the alias ?
By reading a structure in the kernel ?

All these questions, because a customer of mine is using ASE 1.4 with
2 UNIX V4.0B machines, and is having NFS as a service
(thus with an IP address for this service).

When machine A is down, the NFS service points to machine B.

In /etc/resolv.conf of the clients, there is first the IP adress of
machine A, then the IP adress of another BIND server, and that leads
to long delays when A is down before reaching the other BIND server.

So customer tried to declare B as a BIND server as well, and put the
IP address of the NFS service (sort of alias I would say) in the
resolv.conf of the clients. 
But this configuration doesn't work, it times out.

nslookup says "Can't find server name for address.."

My thinking is that it looks like the alias bug, but it is not.
I think the customer should declare the IP address of the ASE service
in named.boot.

What do you think of that ?
Thank you also for clarifications about the begining of my note !

Pascal d'ORNANO