Is data recovery possible? Intel Matrix Raid failure

[mroz]

My two drives are currently operating fine under AHCI, so I suspect my problems were down to buggy Intel Matrix drivers.

Since I disconnected my eSata drive (rather thasn just turning it off), the odd IMSM pop ups about a new device have ceased, so I guess they were down to it partially identifying the powered down drive (another Samsung in an IcyBox enclosure), though I don't even know if that's electrically possible. Certainly another unexpected oddity which doesn't instill confidence in driver robustness.

I won't personally be using Intel Matrix Raid again, ever, unless it's in a hardware set up which has been explicitly stabilty tested wrt a particular driver version & I can't see myself ever having the time & inclination to do this, as these problems took months to become apparent.

[mroz]

Thought I'd follow up on this, a year down the line.

Since giving up with Intel's Matrix Raid I've had no other serious disk related issues with this machine which was in quite heavy use until recently; indeed it's still up 24/7 pretty much.

There have been a few spontaneous reboots which from system logs look to be related to audio drivers. My best guess is that Matrix Raid simply isn't robust enough to cope with such dramatic shutdowns.

[peterb]

Thank you for the update - it really confirms my own thoughts that software (mobo) RAID systems are more selling gimmicks than serious solutions. If you must (or need) to use REAID, either use software RAID built into the OS (especially if it is Linux) or use a dedicated controller card - AND (whatever the solution) back up the data regularly - which you should do anyway. Really, RAID is only necessary (IMHO) where high resiliance is needed and where long periods of down time are unacceptable

[mroz]

Agreed. By an odd coincidence I'm once more considering RAID, this time for NAS - a custom box & FreeNAS looks promising

[steve_cronin]

Simple answer - yes it is
I have a pair of disks configured with one RAID 0 Volume for the OS and a RAID 1 mirrored volume for my data (having moved the default location of the Vista users directory).

I got a spare disk (must be the same size or bigger than the one its replaces) and unplugged one of the existing RAID disks and rebooted. This was to "break" it just so I could test the recovery. Just as expected I got RAID 0 failed and RAID 1 degraded.

Plugged in the spare disk to a spare SATA port and rebooted until the RAID configuration BIOS (ctrl-I on my motherboard). The RAID bios still reported RAID 0 failed RAID 1 degraded. So far so good. The BIOS also gives you the option to select the spare disk to rebuild the mirror to, which I did.

Now it reports "rebuilding" of the mirror will continue in the OS, so the next step is to install and OS image backup (I use Acronis) to the RAID 0 volume, but first you need to delete the existing RAID 0 and recreate it.

The BIOS tells you that deleting a RAID volume will delete all data on the disk - no it doesn't! The RAID 1 volume remains ready for recovery. It only deletes data on the RAID 0 volume (which is already broke anyway because I unplugged the disk).

So I proceeded and deleted the RAID 0 volume and recreated it. In my case it automatically selected the original first disk that the RAID 0 was on and spare space on the new disk that my RAID 1 is going to rebuild to.

Next step is to recover the Acronis image (or your preferred backup solution - I have also successfully recovered my system from Windows Vista Ultimate backup images) to the newly created RAID 0 partition to restore the OS and reboot.

Hey presto, as soon as my machine boots, it is exactly as it was before the RAID broke (or in this case my simulated breakage) and the Intel Matrix software reports that the mirror is rebuilding.

In the meantime, I have a fully functional recovered PC with all my data intact that I can use straight away whilst the mirror rebuilds the redundancy.

I should point out that before I tested this, I had full image backups of my OS and my data on the mirror. I don't believe you should ever rely on one recovery method, and although I don't agree with mobo RAID is a gimmick

Thank you for the update - it really confirms my own thoughts that software (mobo) RAID systems are more selling gimmicks than serious solutions. If you must (or need) to use REAID, either use software RAID built into the OS (especially if it is Linux) or use a dedicated controller card - AND (whatever the solution) back up the data regularly - which you should do anyway. Really, RAID is only necessary (IMHO) where high resiliance is needed and where long periods of down time are unacceptable

, I do believe you should also have an offline backup as well as peterb points out (I use external disk drives).

The procedure I followed is not the same as the example that the Intel Matrix documentation gives for when you have the OS on RAID 0 and a RAID 1 on the same pair of disks. In fact their procedure sort of negates the benefit the RAID 1 gives in the first place (unless you have it on a pair of disks on its own).

So there you have it - a small RAID 0 partition for the OS giving you speed and a RAID 1 mirror - both volumes on the same pair of disks. If it fails the only bit you need to restore is the OS from an OS image backup and the RAID 1 takes care of itself when you follow the procedure above.

[mroz]

Simple answer - yes it is
[snip]
The BIOS tells you that deleting a RAID volume will delete all data on the disk - no it doesn't! The RAID 1 volume remains ready for recovery. It only deletes data on the RAID 0 volume (which is already broke anyway because I unplugged the disk).
[snip]

Nice. Any chance you could post the version numbers for the Intel Matrix software & the related hardware involved? The motherboard manufacturer's name & the mobo name/version itself should be enough, if you don't know the hardware details.

Thanks for posting - good to see this thread still contains imho useful info about software raid & IM in particular

[Edit] I've just reread this thread; ah, the memories ; & realized that while your post contains very useful observations & a procedure to recover your similar system where a two disk arrangement is split to give RAID 0 & RAID 1 volumes, it doesn't actually answer my original question! You describe recovering the RAID 1 volume, recreating the RAID 0 (something we didn't know was possible, as we feared it would destroy the RAID 1 volume in the process) & then restoring the RAID 0 volume from a backup.

I was actually asking for more than this. I wanted to restore the RAID 1 & also the RAID 0 without recourse to any backups (as I stupidly didn't have any). Ordinarily this isn't possible due to data loss, however in this case the hardware was fine as was the filing system data on the volumes. It was only the RAID 0 metadata that had been corrupted. In theory I think this ought to be possible, & swiftly at that without needing the significant data movement of restores or un-backing up, but in practice I couldn't find the necessary tools. I did however find a way to recover the data from the RAID volumes, including from the RAID 0, proving it was only the metadata that had been corrupted (that process is described earlier in this thread).

For anyone else facing this problem & having no back ups, a combination of our two approaches is probably best, as your method will restore the RAID 1 volume faster & with much less hastle

[Aside] How does one thank a poster? Despite my post count I've not been here in a while & currently can't see a suitable link anywhere, let alone remember where it used to be .
[Edit2] Ok, I now see the thanks button on many posts - I just don't know why it doesn't appear in this thread on anyone's posts - why can't I thank peterb or steve_cronin?

[steve_cronin]

System specs (was available on my post profile)

Motherboard: Abit IP35 Pro
Intel Rapid Storage Technology 9.6.0.1014

My plan of using RAID 0 for OS and RAID 1 for data is that not having a backup for the RAID 0 didn't matter as you could just rebuild the OS onto this, and meanwhile your data partition is safe. However, I do have an image backup (of the RAID 0) which is updated every few months or so just to avoid the pain of a complete OS install. Not a lot of help to you and your question though I realize. I don't know if you were trying to recover the RAID 0 assuming the data on it was not corrupt, but in my experience of RAID 0 failures, it was caused by a physical disk fault so was completely unrecoverable (without a backup).
As for whether RAID 0 actually gives a performance boost for the OS I have no idea, but it does save some valuable disk space by not mirroring something which can be recovered by a re-install of the OS from an image really quickly (rather than having data and OS in the same partition which can take ages to recover).

[peterb]

<---snip


[Edit2] Ok, I now see the thanks button on many posts - I just don't know why it doesn't appear in this thread on anyone's posts - why can't I thank peterb or steve_cronin?

There is a time limit and the post you wanted to thank was outside that time limit You cannot thank old posts.

[mroz]

@Steve: Interesting, cheers.

@PeterB Thanks

Merry Xmas to all