HP EOB022 RAID

[barak]

Hi Gurus!

I have a three disk SCSI-controller HP EOB022. One of the disks is out of game and I am in a middle of a battle to find out the RAID settings. Anyone here to help me with this one?

[einstein9]

- Create a Dump of all working drives (if you have the gear)
- Import those Dumps into UFS, and create a screenshot of it and post it here
- If possible Sector 0 (via Winhex) for those as well and post the screenshot as well

good luck

[barak]

Hopefully I was able to follow your instructions. Here we go:

Attachments

[einstein9]

from Winhex image i see the working drives are Disk3,4
Disk4 is 1TB

from UFS, Disk3 is 500GB,

am i right here? if YES >> then how did you DUMP ?

Compaq disk is SCSI 73gb (as in UFS), where is the other 73gb drive? or HDD DUMP?

[barak]

Well I did not have 73GB SCSI drivesat hand so I used other drives as targets while cloning.

[fzabkar]

ISTM that the Hitachi clone is the only drive whose partition information makes any sense. AFAICT, the two clones appear to constitute a RAID 0 pair. The third member of the RAID must be a parity drive. Does this make sense?

Code:

  start     size
----------------
     63    16002    MBR -> 0x3e82 = 16002
   1088   203968    1088 + (203968 / 2) = 103072
103072 16385280    103072 + (16385280 / 2) = 8295712
8295712 16385280    8295712 + (16385280 / 2) = 16488352

The 2nd, 3rd and 4th partitions appear to have consistent boundaries, but the first partition appears to be bogus. Wikipedia suggests that a partition ID of 12h is usually associated with bootable FAT service partitions. Perhaps the user wiped the MBR and partition table, and then created the single 12h service partition. In fact the code in sector 0 is a standard Microsoft MS-DOS MBR.

http://thestarman.pcministry.com/asm/mbr/STDMBR.htm

[einstein9]

Well I did not have 73GB SCSI drivesat hand so I used other drives as targets while cloning.

I mean by "DUMP" : HDD to Image file , use Winhex to do it, (73gb is not a big deal)

the way you did it needs some extra work by creating REGIONs (can be done, but longer path)

@fzabkar i think its R5 not R0 (needs quick winhex look @the end)

[fzabkar]

1088 is divisible by 64 (1088 / 64 = 17). Could the stripe size be 32KB?

[Dmitri]

Here we go:

Look for the partition beyond the 50 GB point on each of the drives.

Should find yourself stuck with this recovery, I'll be happy to look into it remotely.

[Martin]

What is the problem? One is with data in first block, one is with parity... What is in sector 64 in each drive?

[fzabkar]

Assuming that the OP's RAID looks like this ...

Code:

A1 A2 Ap
B1 Bp B3
Cp C2 C3

... then ISTM that there should be an obvious visual difference as one transitions between a data stripe and a parity stripe, in which case a quick scan with a hex editor should identify the parity stripes.

FWIW, the starting sector of each valid (?) partition is divisible by 32 but not by 64.

[Doomer]

the two clones appear to constitute a RAID 0 pair. The third member of the RAID must be a parity drive. Does this make sense?

[data-medics]

That's an HP Smart Array, meaning that you're almost certainly got a parity delay (as Lamer pointed out) as well as virtual disks over the physical RAID.

The reason you're only seeing the 15Gb worth of partitioned space is that you're only seeing the first virtual disk (and one that's pretty small too), but there's likely to be other VDs later on.

[DR-Kiev]

Offset 1088 sectors
Stripe size 16Kb (32 sectors)
Left asynchronous
Delay 16