[Info-vax] Idiots Guide To Clustering

[Jan-Erik Söderholm]

Michael Moroney wrote:
> H Vlems <hvlems at freenet.de> writes:
> 
>> On 17 mrt, 17:36, moro... at world.std.spaamtrap.com (Michael Moroney)
>> wrote:
>>> H Vlems <hvl... at freenet.de> writes:
>>>> Basically what happens is that a few parameters in sys
>>>> $system:modparams.dat are set correctly.
>>>> ALLOCLASS : each cluster member in a VAX NI cluster must have a unique
>>>> and non-zero value
>>> ALLOCLASS does not need to be nonzero.  A zero ALLOCLASS will change
>>> how a member's disks are named in the cluster.  They will be node$ddnu
>>> rather than $m$ddnu.  It was the norm during the time when local area
>>> vaxclusters with many nonvoting satellites to have the satellites use
>>> alloclass 0.
>>>
>>> You want ALLOCLASS nonzero for any node serving shared storage, but I
>>> don't remember if that's even necessary with port allocation classes
>>> for shared SCSI.
> 
>> Correct, but one might be tempted to set up a shadowset across the LAN.
> 
> Yes, if the node is going to serve a disk to a shadowset, that disk must
> have a nonzero alloclass. name$ddcu disks (zero alloclass) cannot be
> shadowset members.  A node with a zero alloclass can still mount shadowsets.

If I'm not wrong, you can't run HBVS with alloclass = 0 even on a
single (non-clustred) box...

Jan-Erik.