Is data recovery possible? Intel Matrix Raid failure

[mroz]

I've been avoiding sorting out a reliable external back up for the last month, so I guess this was inevitable...

Spec as per 'My System'.

Two 500GB Samsung Spinpoints on a Gigabyte-GA-P35-DS4, in a matrix of half Raid 0, half Raid 1. The system is overclocked, but has been stable for five months.

I've noticed once before it doesn't like booting with an iRiver 8G Clix2 connected to a usb port - the bios will hang at the point it's claiming memory checking. This doesn't happen at any other time. This has only happened on three or four occasions. After it occurs the bios resets certain parameters losing my overclock, as a safety & recovery feature.

Last night this occured so I restored my usual settings & rebooted. XP bluescreened immediately after login; I suspect the Intel Matrix Raid software from what I've found online - eg the thread here. I was foolish enough to reboot once more, only to see the same error.

Next reboot, the machine posts, but can't boot the OS as the drives aren't readable. Checking the bios based Matrix Storage Manager I see the 2nd drive is still correctly reported as Member Disk(0,1), but the first drive now reports as Non-Raid Disk, consequently the Raid 1 is degraded & the Raid 0 failed.

Much important data is on the Raid 0 array as well as the OS install. Yes, I know about back ups. I'm an idiot.

I know if I install XP to a new temporary drive, the OS based Intel Matrix Storage manager should recover the Raid 1 array, but what I'm posting to ask about is does anyone know if I might be able to recover the Raid 0?

I know how Raid works & a physical drive failure would be fatal, but in this case it seems likely all that's been corrupted is the protected sector on the drive with Intel's storage manager uses to record the Raid structure/details. The data itself right down to the NTFS filing system is, I imagine, probably still intact?

Can anyone offer advice?

The best lead so far seems to be from Runtime. Anyone have experience of this especially in relation to Intel Matrix Raid?

I don't have another spare drive around atm large enough to attempt a recovery, so I'm about to order another Spinpoint & an external case, both for any attempted recovery & future back up.

In the meantime, any comments or suggestions?

[Agent]

Has the array degraded (It'll say while posting).
If it as, recreate it with the exact same settings, but don't write anything to the drives.

Install the OS on a separate disk. Download @Active data recovery and let it run. It should get the majority of it back, provided the drives have not corrupted too much

[mroz]

As the Intel Storage Manager now thinks one of the two drives isn't part of the arrays, it reports the Raid 1 as degraded & the Raid 0 as failed.

As for recreating the arrays, I can't do this from the bios storage manager, as it only allows creation/deletion. Before it would let me recreate the 0 or 1 array, I'd have to delete both. Do you think I dare risk this? It won't, unfortunately, let me insert the lost drive back into the two existing arrays.

Are you fairly sure deleting/creating arrays will not touch the data within? It does seem plausible.

Edit: Thanks btw for the suggestion

[Agent]

Oh thats nasty. If both RAID arrays have failed / degraded, its usually by some sort of hardware failure or serious corruption.

Don't delete and recreate them just yet - download what @Active and let it run. It will show you the results / whats recoverable in its current state.

Recreating the array doesn't normally touch the data, but I've only ever done it on one (entire disk) RAID 0 array, so I didn't need to take the RAID 1 section into account. If it altered the areas on the disk it planned to use for the data when making a new array, it could result in data loss. I've not done it though sadly, so can't comment on it directly

[mroz]

Yup, nasty

I would suspect the hardware, but for the fact the symptoms match the OC thread, where in their case a failed OC resulted in a bios reset which subsequently upset the Intel Matrix Storage drivers causing bluescreens then one or more drives mysteriously dropping right out of the array/s; that & that the drive is visible & functioning as a non-Raid unit.

Thanks for the suggestions again. I'm currently reinstalling on an ancient 20GB Maxtor & will then see what I can recover if anything. @Active doesn't know about software Raid, does it? I assume not & at best it might help me recover data from the Raid 1 half. We'll see.

[peterb]

You might wnt to try this

Data Recovery Software - File System Utilities

It isn't one I have tried however, so I can't vouch for it personally.

However before attempting any rescue yourself, I would strongly advise that you buy two more drives and clone each drive from your broken array onto the new drives - then attempt the data recovery from the cloned discs. That way you will still have the originals to work from, and if the first attempt goes horribly wrong, you can re-clone them and start again.

If the data is that important, and you have fairly deep pockets, you might be better off with a data recovery company.

[Agent]

Yup, nasty

I would suspect the hardware, but for the fact the symptoms match the OC thread, where in their case a failed OC resulted in a bios reset which subsequently upset the Intel Matrix Storage drivers causing bluescreens then one or more drives mysteriously dropping right out of the array/s; that & that the drive is visible & functioning as a non-Raid unit.

Thanks for the suggestions again. I'm currently reinstalling on an ancient 20GB Maxtor & will then see what I can recover if anything. @Active doesn't know about software Raid, does it? I assume not & at best it might help me recover data from the Raid 1 half. We'll see.

Yup, it does software RAID
Its really good software

[mroz]

You might wnt to try this

Data Recovery Software - File System Utilities

It isn't one I have tried however, so I can't vouch for it personally.

Thanks - I thought it looked promising, which is why I linked to it in my first post

However before attempting any rescue yourself, I would strongly advise that you buy two more drives and clone each drive from your broken array onto the new drives - then attempt the data recovery from the cloned discs. That way you will still have the originals to work from, and if the first attempt goes horribly wrong, you can re-clone them and start again.

Thanks, that is a good idea. Unfortunately I'm pushed to pick up the single drive I've just ordered right now & balancing value of the data against the time the machine would need to be out of use until I could find the cash, I can't take that route on this occassion. No doubt there will be another though & I'll do this for precious data.

If the data is that important, and you have fairly deep pockets, you might be better off with a data recovery company.

Pockets /very/ shallow. The data has little monetary worth; it's more a matter of personal docs/photos & to a lesser extent the time it will take to rebuild from scratch.

Atm I have XP on an old drive running in the machine. I'm currently backing up the Raid 1 array across my lan. Once that's done, around 11pm, I'll look into possible solutions for data recovery from the Raid 0.

[mroz]

Yup, it does software RAID
Its really good software

Cool, that gives me two candidates to consider later. Thanks to all for the quick advice. Appreciated.

[peterb]

Soz - missed the significance of the link! Hope it goes OK - let us know!

[mroz]

Currently I'm running Active@ Partition Recovery on what it sees as the RAID 0, however while I get the impression it would do its job on a lost partition within a working RAID set up, I don't think it can possibly help in this case, as it must be relying on the Intel drivers to tell it what's on the array, but those drivers have lost half the array. So far it's 23% of the way through & has found 0 partitions & 0 files.

When it's finished I'll reboot & disable RAID so that the drives can be accessed separately, then I'll give Raid Reconstructor a shot. It has to be told which drives to consider & will attempt to infer drive order & blocksize (if necessary), before building a disk image from the raw data. I'll not be able to do the latter until my new drive from Scan turns up, hopefully on Wednesday.

I'm hoping that option will work & either I can ghost that back to a new RAID 0, or at worst reinstall to a new RAID 0 & manually restore critical files.

If that all fails, I'll use the Storage Manager in the bios to delete/create the arrays & hope this doesn't corrupt the data further, then try running Active@ Partition Recovery on that.

At that point I start crying, give up & make do with a clean install, followed by regular back ups & self flagellation.

Edit: When I run Raid Reconstructor it would obviously help if I don't need to let it guess the parameters. I assume Block size is just strip size; 128kB in this case. How about the data start sector? For hardware RAID it's usually 0. For Intel's Matrix RAID, is it 1? This is essentially a software RAID, yes? So it must store the RAID parameters on the first n sectors of each disk.

[mroz]

Update, in the end I terminated Active@ at 50% as it had found nothing & I wanted to start Raid Reconstructor before going to sleep.

I let it make whatever guesses it wanted & it came up with the expected parameters rapidly & with high confidence.

I saved the output as a mysterious virtual image file, only to discover this is, reasonably, a tiny xml file describing the raid construction.

As an aside, Raid Reconstructor only supports, aiui, raid 0 with 2 drives (also raid 5 upto 11 drives). If you use more, they suggest you create a profile of your array, which Raid Reconstructor will build for you, that you send to them & in return they supply the manually constructed details - this costs $299 payable only if they can provide the information.

However if you know the structure yourself it would be easy to author the file by hand. Here's my two drive raid 0 virtual image file (vim):

<?xml version="1.0" encoding="windows-1252"?>
<virtualimage>
<info>
<!--NOTE: The INFO tag is not necessary for the VIM to function. It is only a comment.-->
<creator>Raid Reconstructor, Version 3.32</creator>
<creationdate>2008-02-04T01:23:39.187Z</creationdate>
<!--
RAID-0, 2 drives
Drive 1: drive 129, X, Geo (60801*255*63=976771055), LBA 976771055, 466 GB
Drive 2: drive 130, X, Geo (60801*255*63=976773168), LBA 976773168, 466 GB
Block size: 256
Start sector: 0
Rotation: Forward
-->
</info>
<loop>
<drive>HD129:</drive>
<drive>HD130:</drive>
<cycle blocksize="256">
<stripe>1,2</stripe>
</cycle>
</loop>
</virtualimage>

Noting the info tag is optional, the only important but is:

<virtualimage>
<loop>
<drive>HD129:</drive>
<drive>HD130:</drive>
<cycle blocksize="256">
<stripe>1,2</stripe>
</cycle>
</loop>
</virtualimage>

I'm not familiar with the source of the drive numbering scheme, but the next piece of their software which you need will show the mapping between these numbers & the physical drives, so the info is readily available.

The next step is to attempt imaging or file recovery from this data, depending on how corrupt it might be. I used RunTime's GetDataBack for NTFS.

Initially this scans for available drives. I was somewhat concerned as it twice told me drive A wasn't ready - which isn't too surprising as this box doesn't have a floppy - however repeatedly skipping the error was all that was needed to proceed. It is able to open the .vim & lists it with all the other drives from which recovery can be attempted.

Now comes the time consuming step. Scanning the data. On this system it took 4-5hrs. The result was a list of possible ntfs file systems, along with a recommendation as to which are legitimate & possibly recoverable. It found all of mine.

From here one chooses a specific file system & after a short delay to scan the mft & indexes, a file browser is presented.

One is also able to save the scan results, so that if you exit the application before completing all restoration, you'll subsequently be able to resume it without repeating the entire scan.

The browser lets one view some simple file types directly, or to launch any file in an external application, to spot check integrity before starting a mass copy. Any group of files or directories can be exported with the standard save dialogue. Unfortunately I found it wouldn't let me export anything across my network, complaining that the destination directory couldn't be created, regardless of whether it already existed or not. Local extraction was fine.

However as I don't have enough storage space to permit recovery atm, I wasn't able to do anymore. I'll have to re run the program on Wednesday after setting up my eSata drive.

Fingers crossed I will get much, if not all, of the data back - it is looking promising. The Runtime software seems to be good at its job. Even though the gui is a little rough this can be forgiven. Of course as my problem was more than likely a corruption of the raid structural info only, all the remaining data is probably there to be found, so I'm not able to comment on how well this would work with badly corrupted data. Nevertheless, I do get the impression under the bonnet this software is sound.

Edit: I'll post back when I have more results.

Query: In the case of the OS partitiion, if I merely copy /all/ files to a new partition, then create a new RAID, what's the best way to transfer the data back into the new initial primary partition within that array, so that (if the data's intact) it will be bootable & won't upset XP? Is it enough to set it active? How about the MBR? Can I patch this up from my XP installattion disk, either via the recovery console or a repair install of XP?

[peterb]

Thank you very much for posting that - its good to get a hands on review! I have mentioned this software to a few people on the forumns, when they have had problems, but only on the basis of the web site - nice to know it works in practice!

[mroz]

I know what it's like when you're on the other end trying to find a solution to an immediate problem & the value such info can have, so it's good to help add to the searchable knowledge base when the chance arises.

Hope my drive turns up on Wednesday; couriers are often hit & miss around here.

Can anyone comment on my query re restoring C: from the recovered files, making it bootable etc?

[peterb]

Can anyone comment on my query re restoring C: from the recovered files, making it bootable etc?

Probably not - reliably! You are probably better off rebuilding the system from scratch, and then importing your data.

Without knowing your particular set up, my thoughts would be...

"Do I really need a RAID 0 set up?" - as you have found out, there are issues if one drive fails, and unless you are running specific types of applications (large databases) the performance gain (which only applies to read operations - write operation are the same or slightly slower) really isn't worth it.

You might also consider your partition strategy - having your data on a separate partition (wich you may want to have as RAID 1) makes backing up a lot easier, and if you do have to reload the OS, the risk to the data is minimised.

[mroz]

Thanks for the feedack.

I move quite a few large files about - mostly video. Also do a fair bit of image processing.

I was using RAID 1 for some data. My main mistake was that I left lots of stuff in scratch space that was only ever meant to be there a few days; that & no backups of anything . Instead a few days turned into a few months. Also for some reason that escapes me now, I put My Docs into the 0 RAID.

When I rebuild I'll shrink the 0 RAID in favour of 1, so that it's only used for scratch & the OS.

Scan have my drive delivery down for tomorrow, but it seems City Link can't read; the drive is out for delivery now & there'll be no one here in a few minutes for the rest of the day.