SCSI disks access mutual exclusion

[Luigina D'Amico]

Hi All,

I have a Cluster HP composed by three nodes( R servers) and by SCSI chain of shared data disks (16).
When I configured the first R node I select 3 disks as boot disk and data disks: they are mounted on the node and after that they are in charge of this node.
When I have to select the disks for an other node (the second one) I can see all the disks, also the disks already selected for the previous node, and if I select one of these all the data referring to the previous node are lost. It is possible to have more informations about the status of the disks already in charge for an other node?

Any help is GREATLY APPRECIATED.
Thanks very much,

Luigina D'Amico

[Duncan Edmonstone]

And lo you find the joy of managing clustered systems! The fact is clusters might allow an application to be more available, but the upshot of all that complexity is that they're much easier to break.

Seriously though, good practice dictates that local disk is not attached to shared channels - so all disk down the shared channels can be seen by both sides. If you can't do this, the best you can hope for is some kind of warning to other sysadmins in /etc/motd - or if your feeling brave you could have a startup script which rmsf'd the disk you don't want to be able to see (I've a feeling this wouldn't stop SAM showing it as available though...)

If you had buckets of money, and were using fibre channel/SAN disk arrays from HP?EMC/Hitachi then you can use fabric zoning and LUN masking technology to solve these kinds of issues.

HTH

Duncan