I asked about simplification of maintenance of disk shares. I wanted to
mantain host names only in the DNS and no longer in /etc/hosts.
> ...and then Lucio Chiappetti said...
> %
> % - Thomas Anders, Joel Lee, David Thorburn-Gundlach and
> % Gustavo Chaves all indicate that if one uses unqualified names
> % in the share command the query is resolved via NIS, if one uses
> % fully qualified names it is resolved via DNS
>
On Thu, 26 Mar 1998, David Thorburn-Gundlach wrote:
> You seem to have misunderstood me...
> I never, ever said this. What I said was that unqualified names on
> your share line will not be matched by the FQDN returned from dns.
I apologize with all the persons quoted (who replied again to me) for the
misunderstanding. It is actually the other way round :
- one has to use in dfstab either the short form or the FQDN form if
one is using respectively NIS or DNS
- that's because "share" (or actually "mountd" later) will do a reverse
lookup using the appropriate resolver.
> share really does *not* behave differently from other commands as far
> as I have been able to see
this is questionable if one does not appreciate in detail what is going
on (the reverse lookup)
> (what I'd recommend, too, BTW), then when you try "ping atlas", you
> will get "atlas.ifctr.mi.cnr.it alive" (or whatever the FQDN is).
Yes that's what happens (so ping can "expand" and append the domain)
I finally managed to use the following approach :
- all hosts defined in the DNS SOA file
- all hosts having nsswitch.conf = files dns (or equivalent, I have some
Alphas which *always* had that contrary to the old SunOS machines)
- no machines in /etc/hosts but the local one and NIS master
- in the NIS distributed netgroup file I replaced all entries of the form
> kalypso (kalypso,,.ifctr.mi.cnr.it)
with
kalypso (kalypso.ifctr.mi.cnr.it,,.ifctr.mi.cnr.it)
apparently, although the documentation is not clear on this, one CAN
put the FQDN in the first field, while the additional (NIS) domain in
the third field is used for further checks.
We already had such entries in order to define an overall institute
netgroup containing all machines (for hosts.equiv), so is no additional
effort.
- in the share command we used and continue to use strings of the form
rw=machine1:machine2:machine3
but now such "machine*" are the netgroup names (each one containing a
single machine). These now resolve to the FQDN.
Probably our previous problems were due to the fact they resolved to
the simple name.
BUT we thought (and the documentation is TOTALLY SILENT in this respect)
that to use a netgroup we should use rw=@groupname
(as in all other files using netgroups)
while experimentally it is not so.
Thanks to everybody
----------------------------------------------------------------------------
Lucio Chiappetti - IFCTR/CNR - via Bassini 15 - I-20133 Milano (Italy)
----------------------------------------------------------------------------
Fuscim donca de Miragn E tornem a sta scio' in Bregn
Che i fachign e i cortesagn Magl' insema no stagn begn
Drizza la', compa' Tapogn (Rabisch, II 41, 96-99)
----------------------------------------------------------------------------
For more info : http://www.ifctr.mi.cnr.it/~lucio/personal.html
----------------------------------------------------------------------------
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:12:35 CDT