Re: Quorum partition - does it have to be raw?

Figured so. Thanks for the response.

I recall, a long while ago (6.2 version days) that there was some 
chatter to have an alternative to the raw partitions for cluster node 
state sharing for Piranha. Something along the lines of how IBM's HACMP 
(for AIX) accomplishes it - along the heartbeat networks (serial, 
network and/or across SSA loops) via cluster daemons. Each node has a 
local memory store of the state of the cluster, resource groups and 
nodes and all shares with all. It was unofficial chatter, but the 
discussion did come up. Do you know if there is any current discussion 
for that?

Thanks again for your speedy response.

Tim Burke wrote:

> Philip Greer wrote:
>
>> My initial thoughts is 'yep'. You MUST have a raw partition (10MB
>> primary, 10MB shadow) for the quorum partition for clustering.
>>
> Correct.  It must be raw.  NFS doesn't cut it.  Even when enabling 
> synchronous writes, it doesn't provide the required semantics.



-- 
---------------------------------------------------------------------
PG..                                         [email protected]
Law of probable dispersal: Whatever it is that hits the fan will not
be evenly distributed.