topic Re: Sharing Surestore 12H disks between two nodes. in Disk Enclosures

Sharing Surestore 12H disks between two nodes.

Kim H. England — Thu, 02 May 2002 14:36:13 GMT

Has anyone been able to connect two nodes (K580s) to a single 12H disk arry, and be able to share the disks betreen the nodes? If so, how?

Re: Sharing Surestore 12H disks between two nodes.

harry d brown jr — Thu, 02 May 2002 14:41:57 GMT

Kim,

using the search feature and this string, I found the following link:

+12H +shared +nodes +disks

http://forums.itrc.hp.com/cm/QuestionAnswer/1,,0x63741cc6003bd6118fff0090279cd0f9,00.html

and this search string produced a lot of hits:

+12H +cluster

live free or die
harry

Re: Sharing Surestore 12H disks between two nodes.

A. Clay Stephenson — Thu, 02 May 2002 17:34:15 GMT

This is actually a very standard configuration especially in MC/Service Guard configurations.

This is essentially what you do:
1) I'll assume that 12H Controller X is set to SCSI ID 0 and that Controller Y is set to 1.
2) The FWD controllers in HostA are set to SCSI ID 7.
3) The TWD controllers in HostB are set to
SCSI ID 6.

HostA c1 (7) -------- 12H X Controller (0) ------ HostB c1(6).

HostA c2 (7) -------- 12H Y Controller (1) ----
-- HostB c2(6).

The LUN's can now be accessed by both HostA and HostB and both can have primary and alternate paths to each LUN. Unless you are using raw/io, each LUN should only be accessed by one host at any one time.

It is perfectly legal to define LUN's and allow some to be accessed/mounted on HostA while others are accessed/mounted on HostB.

Re: Sharing Surestore 12H disks between two nodes.

Derek Engelhaupt — Wed, 15 May 2002 18:25:59 GMT

Just to let you know, sharing disks without ServiceGaurd is an unsupport configuration because it can cause data corruption. Data corruption can be caused by the user inadvertantly writing data to volumes not owned by that box. ServiceGaurd locks the volumes so that there is only one owner of the volume and therefore prevents this from occuring. EMS will also spit out errors due to the fact it can't talk to the other controller since it is physically connected to a different machine. There is no physical internal connection between the two controllers on an AutoRaid.