MSA1500 SCSI redundant loop question

[Sloan Mitchell]

I have a MSA1500 and can’t find what I am looking for in the documentation. I want to run two loops to my disk shelfs for redundancy. I can’t find anything that specifically tells me that this is supported. Can anyone tell me if they have it setup this way?

Example:
MSA1500
4 SCSI I/O modules
2 SCSI shelf’s

Or
MSA1500
4 SCSI I/O modules
2 SATA shelf’s

[Uwe Zessin]

It is not possible. The parallel SCSI and SATA disk drives have a single interface and the enclosures (MSA20 + MSA30) do not provide a redundant connection either.
And I hightly doubt that the controller firmware could deal with such a situation.

You _might_ have been confused by an MSA30 DB. That is a parallel SCSI drive drive enclosure with TWO connectors. Its SCSI module puts the first seven disk drive bays on one connector and the remaining seven bays on the second one (it is called 'split-bus').

The MSA20 has a single embedded SmartArray controller which is used to convert from 12 SATA links to one parallel SCSI interface. Again, there is no redundancy.

[Sloan Mitchell]

That doesnâ t make sense. If you have an active/passive setup on your msa1500 and redundant i/o modules in your shelf (SCSI shelf), then you should be able to have one SCSI connection on the right side (one I/O path) and one on the left side (different I/O path) splitting the whole shelf. If you have 2 connections on 1 side, yes you will split the disks to 2 separate controllers.

http://h18000.www1.hp.com/products/quickspecs/11742_div/11742_div.html

Fault Tolerance The dual bus I/O module, as well as redundant power supplies and fans, allows the MSA30 family to provide high levels of system availability.

[Uwe Zessin]

Mitchell,

each controller has four (4) parallel SCSI busses and both controllers are connected within the controller shelf to all 4 back-end SCSI busses which are then fed through the outside by 4 I/O modules.

So both controllers have concurrent access to all disk drives.


I thought you were asking for a redundant connection to the disk drives. Again, that is not possible.

[Andy Ivey]

The controller's state (active/active vs. active/passive) only deals with fibre connectivity to the SAN. If you want that kind of redundancy from the shelves (MSA20/30) you need to virtualize your storage at the server level. You can do this by mirroring the data between two shelves in whatever creative way you wish. I recommend HP's SVR (http://h18006.www1.hp.com/products/sanworks/vr/index.html?jumpid=reg_R1002_USEN). Regardless of the method, the MSA cannot provide that level of safety alone.

BTW- even if you use snapshots or some other server-based safety measure, losing a shelf on the MSA will probably take the entire SAN offline anyway. We have two MSA1500cs with 4 MSA30 each. A faulty SCSI I/O module on a single shelf killed one MSA and locked up ever server that could see it. This took our entire SAN offline until the MSA connected to the shelf was rebooted.

[Uwe Zessin]

Andy,
no offense, but have you really mirrored with VR between two storage arrays, turned off one array and still had access to the data?

[Andy Ivey]

no offense taken
I do not like SVR and only recommend it because it is the only HP software I know of that provides some form of storage virtualization for the MSA. All SVR is going to do is create some snapshots that should lead to faster recovery of a failed array. There is not any software suite in the world that will stop the MSA from throwing a fit if it loses a shelf.
I am guilty of this myself, but no good can come from trying to get EVA/XP functionality out of an MSA. Thatâ s exactly what I tried to do with SVR and the end result was disappointing to say the least.