Manually-Confirgured Software RAID Under NT?

Mandoric asks: "I recently had a RAID-0 array die due to the controller refusing to recognize it as anything but a pair of single drives. Are there any software RAID programs (preferably under NT, as it's an NTFS partition I'd be restoring) that allow manual entry of a preexisting array's stripe size, rather than forcing all stripe data to be rewritten before reading from an array? Or ways to modify the volume data of an existing program to do this?" Are there decent RAID tools, for either NT or for Unix, that will allow one to recover from errors such as this?

Comments

Recovering broken RAID

[sofar]

Well, it's probably not the easiest approach but one that is definately looking into IMHO:

- pop the disks into a linux or un*x system on a free controller port

- pop in an extra empty disk to salvage the data to. preferable larger than both disk together ;-)

- try to figure out the block size of the stripes (usually a multiple of 512 bytes)

- write a simple script that reads block n from broken RAID disk 1, and writes it to block 1 in the empty disk, then block 1 from broken RAID disk 2, to block 2 in the empty disk (ask your local unix guru)

You can fiddle with mount, df and cat /dev/hd?? > hexdump to see how data is arranged on the disks. Hardware RAID does usually not put special data blocks on the disks that make it unaccessible for these tools.

As an example, I run a software RAID and unmounted it cleanly, then remounted only one of both disks at the same point (so without the mirroring). Worked like a charm, also a nice way to continue working when one disk completely fails or you want to make copies fast ;-)

good luck

What hardware?

[bedessen]

You didn't mention anything about your hardware, but I had the exact same thing happen to me with the Highpoint 370 controller on a KT7-RAID motherboard. I rebooted one day to find that the HPT BIOS would say only "broken stripe." I figured my NTFS partition was toast. But there is a utility, raidrb, which I think is written by either Via or Highpoint. Anyway you just boot from a floppy and run it and it "reassures" the controller that the stripe is really ok.

http://download.viahardware.com/faq/kg7kr7/downloa ds/utils/raidrb.zip

There is some discussion of other methods to try if this one does not work in the Raid section on Paul's KT7 page. (Google search for "kt7 faq" for the url.)

Raid 0?

[EvilOpie]

I am curious as to what the job of the server that the raid card was in was. After all, if he truly meant Raid 0, then that is striping without fault tolerance. If you had one drive fail, then the whole raid array would fail and he'd be in the same situation. Now there is an advantage in speed when it comes to drive access, but still, a single disk failure would take out the whole array. That is not unlike what happened here in that the raid array failed when the controller bombed.

That being the case, he should have backups somewhere to restore the data from. I wouldn't trust anything mission critical to a raid 0 array. Besides, when the controller went down, who's to say that it didn't basically shred the data on the disk apart? If the data is being spread across 2 disks, and now the computer sees them as two separate disks instead of one larger disk, then the data on the two disks is probably toast to begin with. (with the exception of going to those companies that do sector by sector data recovery from a hard drive for ungodly sums of money)

I guess the only thing I could suggest is to redo the server with Raid 1 minimum, and hope that your backups are good. There's no telling what that raid card did to your system when it failed.

Veritas can do this

[duffbeer703]

Veritas Volume Manager for NT/2000 can do this.

http://www.veritas.com/products/category/Product De tail.jhtml?productId=volumemanagerwin

Only one problem. It is very expensive.

Here is an article

[Aknaton]

on Tom's Hardware entitled: "RAID without additional hardware: Do It Yourself With Windows 2000"

http://www4.tomshardware.com/howto/01q3/010906/i nd ex.html