MSA1000 configuration question.

[Doug_81]

A couple of us are having a "discussion" about what can and cannot be done with an MSA1000 controller. Perhaps someone can help resolve the question.

We have recently installed new system which utilizes an MSA1000 Storageworks SAN storage system.

Two VMS (non-clustered) nodes are connected to the MSA1000 via an MSA-8 port switch.

The question is can a 3-node cluster be added to this configuration and share the disk storage system? The end result would be 1-3 node cluster and 2 other non-clustered nodes connected to the MSA1000 via the MSA-8 port switch.

One person is saying that this would violate VMS cluster security and another person thinks that it would not be a problem because the MSA1000 is a network device.

This isn't just idle conversation. We are putting a plan together to upgrade a cluster and are thinking that if this configuration would be "allowed", then we wouldn't need to purchase another disk controller. These systems share the same computer room but are running very different applications.

Actually, the clustered nodes run one application and the two stnd-alone nodes run one other application.

Any words of wisdom?

[Uwe Zessin · Accepted solution]

Yes, that setup is possible and even supported. When talking about AlphaServer support you can have up to a single cluster on an MSA1000 and share it with other, non-clustered nodes of the same operating system.

If it is set up properly it will not 'violate VMS cluster security' and it has nothing to do with the MSA1000 being a Storage Area 'Network' device.

The key to this is a feature called Selective Storage Presentation (SSP) which is used to control which server has access to which logical disk(s).

As you already have two non-clustered servers connected to the MSA1000, it is possible that SSP is already in place so that one server does not have access to the other server's disks. If not, it really should be done quick, because it is quite easy to mount or even initialize a wrong disk.

[Jan van den Ende]

And one more remark (hopefully this is totally superfluous, but I just want to make it explicit)

Of course the cluster and the non-clustered systems must NOT try to access each others disks! (other than via the node/cluster that owns such disk, eg, DECnet, or NSF)

Proost.

Have one on me.

Jan

[Ian Miller.]

It sounds like itsd just like you can do on a HSZ70 connected to multiple alphaservers, where it can be configured so that each disk can only been seen by one of the alphaservers.

[Doug_81]

Thank you for your prompt and informative replies. Very helpful.

[Doug_81]

Thank you - very helpful, even if it does mean that I loose the bet ;-)