ProLiant Servers (ML,DL,SL)
cancel
Showing results for 
Search instead for 
Did you mean: 

RAID on 6i - replace failed disk, different SCSI ID?

Jonathon Austin
Occasional Contributor

RAID on 6i - replace failed disk, different SCSI ID?

Hi there,

We have a DL380 G4 with the following:

SA 6i - 2 x 72GB drives in RAID mirror. One logical volume which is the system (C:) drive.

SA 642 - 4 x 300GB drives in RAID 5 with one logical volume which is the D:.

One of the two disks in the volume built on the 6i has failed. HP has attended and replaced the failed disk, but it is appearing as a different SCSI ID to the failed disk.

See the screenshot from ACU - the original two disks were Port 2, SCSI ID 0 and 3. The replacement disk which has gone in the exact same drive bay is appearing as Port 2, SCSI ID 1.

I've upgraded to the latest 6i firmware (2.68) which makes mention of an issue around SCSI ID assigning and dual port disks, however this has not changed the situation at all.

Any suggestions on where to go next? Ideally, I'd like to remove the second failed disk from the array all together which may trigger it to bring the new unassigned disk into the array and start rebuilding the array.

Rebuilding is a big job in terms of man hours and is really my last resort - we have a case open with HP but the server is in Hong Kong (I'm in Australia) and the communication is difficult.

Attached is some screenshots and a dump from ADU.

Any help is greatly appreciated!
2 REPLIES
Brian_Murdoch
Honored Contributor

Re: RAID on 6i - replace failed disk, different SCSI ID?

Hi Jonathon,

The fact that your original mirrored pair were at scsi id 0 and 3 means it was bad to start with since you have your DL380 G4 drive backplane in duplex mode. This normally provides scsi id 0 and 1 for a mirrored pair on one controller and id's 2,3,4,5 on another controller. If you had 0 and 3 originally then the drive in bay 1 (which should be SCSI ID1) had a problem with addressing.

The newly inserted drive is seen as SCSI ID1, which is correct and means the backplane addressing is ok but this gives you a big problem since the good drive (ID0) is looking for a partner at SCSI ID 3 now to rebuild.

A couple of suggestions.....

If you have another DL380 G4 which is in simplex mode (all 6 drive bays on one controller) you can take the good drive and the new one and place them in bays 0 and 3 and just let it rebuild (Don't boot the O/S just). Once rebuilt you should be able to put them back into the original system in bays 0 and 1. It should be OK once the RIS tables are built on the new disk since drive roaming (moving it to another bay) is supported if it has been rebuilt ok.

Alternatively change the config of the system involved to simplex and try the above with the 642 cable and its 4 disks removed. Just have the 6i seeing all 6 bays.

You may also be able to remove the new drive from its casing (4 screws at the side) and put a jumper on the address jumper pack to add 2 to its assigned backplane id. Depending on the manufacturer the jumpers can be marked 1248 (binary powers of 2) or 0123. When inserted into bay 1 the drive will be assigned scsi id 1 but become 3 due to the additional jumper.

Unfortunately you need to get the original good drive at SCSI ID 0 to be able to see another drive at SCSI ID 3 to fix it.

I'm sure others may think of an easier solution and I appreciate you will need help at the Hong Kong end so good luck.

Regards,

Brian

gregersenj
HPE Pro

Re: RAID on 6i - replace failed disk, different SCSI ID?

OOOhhh You're in a bad situation.
Your sever is at risk.

Picking up on murdoc's input. Here's another possibilty.

This is entirely at your own risk. I have newer tried doing this myself, and you need a good disaster recovery plan.

There is a way to split the mirror:
http://h20000.www2.hp.com/bc/docs/support/SupportManual/c00378986/c00378986.pdf?jumpid=reg_R1002_USEN
when it has been split into 2 seperat array's, you can try to delete the array with the incorrect ID
then recreate the mirror using the new drive.

Or you could simply assign the disk ID 1, as hot spare to the array, then at least you would have a running mirror.

This is risky, so if you try, and fail don't blame me.

p.s. latest fw is 2.80(B)
But if problem is initially caursed by the failed disk, then a fw upgrade won't fix it.