Trying to change scsi ID on controllers, but ISL doesn't see the controllers

[Matt Hearn]

I'm setting up a serviceguard cluster with 2 rp5470s. Each has two internal 36GB drives (root disks), a DVD-ROM, and an external tape drive running off of a PCI controller.
I also have two A7381AZ disk arrays with two 36GB drives each; each array is attached to both servers.

I was attempting to determine the SCSI IDs on the primary machine so that I could then go into the secondary and change them, eliminating contention from having two controllers using the same SCSI ID on the same chain.

Still with me? Good.

When I go into the ISL Service menu and type "scsi," the box only appears to recognize three paths:
0/0/1/0
0/0/1/1
0/0/2/0

(All three have an initiator ID of 7, which is normal.)

1/1 is one of the root disks; 2/0 is the other. I assume 1/0 is the hardware path of the cpu, memory, etc.
Where are my 0/12/0, 0/8/0, and 0/9/0 paths to which my tape drive and disk arrays are attached? For that matter, my DVD-ROM is on 0/0/2/1, why doesn't that show up?

If I attempt "scsi init 0/12/0" to see the initiator ID, I get:

ERROR: Path entered does not point at a valid SCSI address.

The back of each disk array has a 3 way switch, labelled "A B C." I have no idea what this does, and unfortunately all the hardware came from HP pre-racked, and I can't find any manuals for the disk arrays in the pile of boxes they left sitting on my floor.

I can't boot the systems up and attempt an ioscan until I make sure I'm not gonna have a SCSI conflict from the other server. Could I boot them up one at a time without that conflict occurring? Is there a command I can run from the service menu that does an ISL/GSP level hardware scan of some kind that would make the system aware of all these extra stuff attached to it?

Thanks!!!

[Steven E. Protter]

No real solution but some help gathering info:

ioscan can be run while the system is running. You can look for conflicts.

mstm or xstm(X windows) will let you poke around and make sure all the expected hardware is where its supposed to be.

Its a start.

SEP

[Helen French]

I think it's a SCSI ID conflict between devices or system. Start the first system and see if you can list all devices with 'scsi' and 'search' commands. Set the SCSI IDs and make sure that it's not conflicting. Now, shut it down and start the other one. Set the IDs there so that it won't conflict with the other.

[Helen French]

To add, Check the SCSI termination too. Make sure that the SCSI bus is terminated at the end and not anywhere else.

[Matt Hearn]

>To add, Check the SCSI termination too.
>Make sure that the SCSI bus is terminated
>at the end and not anywhere else.

In my case, it doesn't terminate, although I ran the cables myself and I make no pretense of experience or intelligence.

I have it set up so scsi goes from system one, into the disk array, back out, and then into the controller in the same slot on system 2. So it doesn't terminate anywhere. Did I set this up wrong? I can't figure out any other way I could possibly have done it, since each disk array has exactly 2 scsi connections on it.

[Matt Hearn]

Latest update:
I'm a moron, and didn't notice the little buttons on the back marked "power." :sigh: So now at least the disks are on.
However, this doesn't do much more good. When the OS comes all the way up, the disks show up in the hardware scan, which is good, but unless I can resolve the fact that the initiators on both boxes are at scsi ID "7," I won't be able to bring up both boxes at the same time!

Despite the disks appearing in the ioscan, I still can't see them from the ISL service menu. Maybe I need the firmware upgraded?

[Helen French]

(Ohhh! I 've been trying to reply and at last I got it working!)

Again, I am not very familiar with this disk array, but from the model number given I think you are using a DS2100 (or DS2110?) disk system. From what I read before, this disk system is meant for a single host connection and the end of the SCSI bus should be terminated unless it's a daisy chain. You can have up to 3 DS21XX in a single bus and the bus needs to be terminated at the end.

Having said, your configuration might work (I haven't tried it) with 2 systems if you set all SCSI IDs without conflicts. But to make sure that your controllers and disks are working, terminate SCSI (without connecting to 2 systems) and see if it works.

This manual would be helpful for you:
http://www.hp.com/products1/evolution/e3000/download/52430406a.pdf

[Joshua Scott]

The A B C switch on the back is for setting the SCSI ID's of the disks in the DS2110 array. I'm attaching a txt file that outlines the IDs.

[Helen French]

To add, the "A B C" switch at the rear side of your disk system is used for Enclosure Addrressing and SCSI IDs. The SCSI address for disks are fixed by it's slot position. So, when you change the switch to A, B and C, the SCSI ID of each slot also changes accordingly. This will be most useful when you have 2 or more disk systems connected to the same SCSI bus. Read the above manual for more information on how the SCSI ID is set.

[RolandH · Accepted solution]

Try a

scsi init 0/12/0 6

this sets the id of the scsi adress in one system to 6. On the other system leave it as it is. Now you have two differant scsi id´s for you controllers. But be aware if you do taht you lost one scsi id and you can´t use the disc slot, which is normaly set with this id.

HTH
Ro