On Sep 30, 2008, at 1:13 PM, Martin, Terry R. (CMS/CTR) (CTR) wrote:
As a matter of fact that is one of the commonalties of these hosts.
There is Oracle Clustering going on in these hosts. Can you explain
more about this?
Well, I don't know exactly how Oracle does it.
But I do know, for instance, that for Veritas clustering, you need to
define a set of shared disks on a controller that implements SCSI
reserve-release, and that Veritas uses that to see who's actually
active in the cluster. They recommend that you use the tiniest LUNs
you can, because they basically just write a few bytes to the start of
the physical device and those bytes differ depending on who wrote it.
So I wouldn't be surprised if the following were true: Oracle does
something similar (and certainly you CAN do something like reserve/
release with channel-attached DASD), but owing to miscommunication,
you're actually putting filesystems and data on the disks you were
supposed to be using just for cluster arbitration management. Then,
every so often, one of the cluster members stomps on part of your
disk, and everything falls apart.
Mind you, this is total guesswork. It might be something else entirely.
Adam