[OpenAFS] DB-Server & virtual ethernet interface

[R.Laatsch]

I checked out running an OpenAFS database server with a virtual ethernet
interface (eth0:afs). Purpose: it should take over the sync site if
that goes down. Result is: failure. Did someone else try that?

Used OpenAFS versions: 1.2.11 + 1.2.13
Setup: 5 servers ; 3 on physical ethernet interface ( $network.14 ,
$network.214 , $network.217 ) and 2 on virtual interface ( $network.8 on
physical $network.35 , $network.55 on physical $network.196 ).
The standard sync site was $network.14. Shutting down this sync site, no
quorum was reached. The servers on $network.8 and $network.55 only voted
for themselves; the 'physical' servers $network.214 + $network.217 both
voted for $network.214 . This looks like only 'physical' interfaces are
used for quorum voting. Might be a bug?
The Netinfo files were set up to point to the virtual interface (lower
IP than its physical IP). Arp caching was not the problem here. The
/usr/afs/etc/CellServDB were in sync on the servers. The
/usr/vice/etc/CellServDB on the afsd's had the corresponding entries.
The afsd's got informed with 'fs newcell ..' giving the 5 servers names.
The server names were ok in DNS. The afsd's partially ran for another cell
(ThisCell) than the server site; but that should not be problematic.
No firewalling was involved. No IP numbers 192.** or 10.** were used, but
official ones 134.**
Comments or help someone?
Mit freundlichem Gruss
Rainer Laatsch
________________________________	______________________
E-mail: Laatsch@rrz.Uni-Koeln.DE	Universitaet zu Koeln
					Reg. Rechenzentrum (ZAIK/RRZK)
Fax   : (0221) 478-5590			Robert-Koch-Str. 10
Tel   : (0221) 478-5582			D-50931 Koeln