On Monday, 04/24/2006 at 12:39 EST, Brian Nielsen 
<[EMAIL PROTECTED]> wrote:
 
> There are several more iterations of Stop Lan and Start Lan, but it is
> always on the backup OSA addresses, never on the primary OSA addresses.
> 
> So the question is: why didn't it ever try to restart on the primary OSA
> when the backup OSA received the Stop Lan?  If both OSA cards were the
> same it might not matter that much, but the primary is a Gig-OSA and the
> backup is a Fast-OSA.

There is no permanent association of "primary" and "backup" in the 
VSWITCH.  There is just a list of available OSAs.  The VSWITCH starts with 
the first one and fails over until it finds a working one.  Whichever one 
is active is the "primary".  The others are "backups".  (You can see that 
in QUERY VSWITCH.)

So, I'm not 100% sure I'm reading your post correctly, but I would have 
expected identical behavior vis a vis the decision to fail over, without 
regard to which OSA was active and which were backups.  A StopLAN on the 
currently active OSA should have caused a failover.  If none of the 
backups worked (i.e. all OSAs unplugged at the same time), I would have 
expected CP to come back to the OSA that is currently 'active' and sit and 
wait for StartLAN on the active OSA or one of the backups.  At that point, 
whichever one comes up first would become the active OSA.

Of course, with the abend in the controller, bizzare things were obviously 
going on.  The Support Center definitely needs to look at it.

FWIW, if you want it to go back to the first OSA in the list, set up some 
system automation to do a SET VSWITCH DISCONNECT followed by a SET VSWITCH 
CONNECT when you get an adapter-initiated StartLAN on the desired device 
(as seen on the controller's console).  CP will go back to the beginning 
of the OSA list and start looking for a working device.

Alan Altmark
z/VM Development
IBM Endicott

Reply via email to