non HA disk system configured in MC/SG?

[Thomas Schler_1]

We are running MC/SG on HP-UX 10.20 using two cluster nodes (K460). An array of HA disks is attached to the cluster nodes using F/W Diff. SCSI interfaces.

Now, we want to attach a new array of four LVD disks (HP disk system 2100) to both nodes. At the moment the ds2100 is attached only to one node using an adaptive SCSI cable, LVD -- single ended. This works.

Would it be possible to use a second adaptive SCSI cable, LVD -- single ended, to attach the ds2100 to the second cluster node?

Of course, the ds2100 is not a HA disk array (no mirrored disks, no redundant SCSI interfaces, no redundant power supplies), but in this case we do not need the HA feature. We just want to have access to the ds2100 from both servers,
even in the case that one server is down (therefore, NFS is not an option).

Our HP representative says, it may be possible, but we have to test it.

Do you have experience in such a configuration? What happens if one server is down? Is the ds2100 correctly SCSI-terminated, can it still be used by the other server without problems?

[A. Clay Stephenson · Accepted solution]

Yes, you can use SE/SCSI in MC/SG but you have to observe the standard stuff: SCSI controller ID's set to 7 on host A and 6 on host B; termination only on the controllers. The main problem with doing this in SE/SCSI or LVD SCSI is cable length. You very rapidly hit the cable length maximums.

If it were me, I would use HVD SCSI with 2 HVD/LVD bus converters on the bus.

e.g.

Host A (HVD SCSI Controller SCSI ID 7) -----
Bus Converter --- LVD SCSI -- Array --- LVD SCSI --- Bus Converter --- Host B (HVD SCSI Controller SCSI ID 6).

This overcomes your length problem and gives you higher throughput than the SE-SCSI controller for the K-boxes. A source of HVD/LVD SCSI Bus converters is Blackbox Corp.

http://catalog.blackbox.com/BlackBox/Templates/blackbox/mainscreen.asp

Food for thought, Clay

[Insu Kim]

>Do you have experience in such a >configuration? What happens if one server is >down? Is the ds2100 correctly SCSI->terminated, can it still be used by the other >server without problems?

In the DS2100 configuration guide, it's refered to supported configuration.
That is, if host was to bacome non-functional, the remaining host would still be able to access the data in the DS2100 with no problem.
In this configuration,
Keep in mind that SCSI address range must be selected so as to not conflict with the host bus adaptor addresses so you have to set the SCSI address switch on the enclousure rear panel.
Maximum bus length is 12M for LVD, 6M for single-ended.
HBA (SCSI ID:7) -> DS2100 with SCSI address switch set <- HBA (SCSI ID:6)
Termination is required at the end of the bus because DS2100 does not have built-in termination circuitry but in the configuration where DS2100 is connected to the dual servers, it's not required because it's done by the HBA (Host Bus Adaptor) in each host.

In some platforms like L,N-Class, you should go to BCH to see if HBAs are properly terminated.
For older models such as T,K,H,I,D-class, termination packs are seated on the HBA so you have to check that out visually.

Hope this helps,