I built a RAID5 array with 4 2Tb devices a few month ago.
- I used LVM2 to build partitions.
- I also used GRUB2.
- Finally I installed Debian.
For an unknown reason the RAID failed. The BIOS "see" only two devices.
I am aware that my RAID is lost.
For the two failed devices smart tool return nothing.

1) with the reduced RAID array:

  • is it possible to retrieve logical volumes and backup them ?
  • is it possible to separate them, I mean to have two devices and not one RAID of two devices, and retrieve data ?

2) When I try to boot with the two good devices, grub rescue takes the control.
The ls command gives:

(hd0) (hd0,msdos1) (hd1) (hd1,msdos1) (md/0)

ls .../boot gives unknown filesystem
ls (md/0) gives error disk 'md/0' not found.

3) I boot an old device and install on it the HDDsupertool application. This application detects the failed disk by the command: hddsupertool -d

ata1 - cff0 cf7e cfa0 1 wdc2003FYYS-02w0b1

But this application is complicate and I don't know if I could retrieve data.
Any ideas I could retrieve data or repair RAID ...


  • The problem is, for a four-drive RAID five configuration, the data are striped across three disks, so you're going to be missing a third of everything.
    – DopeGhoti
    Feb 28 '16 at 17:54
  • I ask myself if really both disks are ko. the probability is low. that's why I search also a solution in this direction ...
    – Bertaud
    Feb 28 '16 at 18:07
  • It sounds like a problem with your BIOS or motherboard rather than a hardware failure of the drives themselves. If hddsupertool can see the 'failed' drive, then so can the linux kernel (try booting with a rescue CD or USB stick). If the kernel can see the drive, then the RAID array can be re-assembled and the data retrieved - (you should immediately copy the data to somewhere else!). I recommend installing all of the drives on another motherboard - or simply replacing your motherboard.
    – cas
    Feb 28 '16 at 23:37
  • I tried to boot an old hdd by moving it in each slot and the boot was always correct. hddsupertool sees it as an IDE disk, uses the direct io mode. I would see rather a problem with the s.m.a.r.t data (hddsupertool does not see them).
    – Bertaud
    Feb 28 '16 at 23:47
  • does blkid or lsblk show the drives? if yes, then the kernel sees them and the RAID array can be re-assembled (and the data copied!) as i mentioned earlier.
    – cas
    Feb 29 '16 at 10:32

If you have a RAID5 device with 2 failed disks, you will only be able to recover bits and pieces of the device. In this case, ⅔ of the blocks would be recoverable. If there is some really important piece of data that you hope to recover and that's small enough to fit in a single RAID stripe, then with a fairly large amount of effort you can see if you can scan for it block-by-block (no guarantees), but that's about it. You will have little luck recovering any kind of intelligible filesystem. Sorry.

  • bad news. scan with hexdump ?
    – Bertaud
    Feb 28 '16 at 18:04
  • @Bertaud something like that :-(
    – Celada
    Feb 28 '16 at 18:06
  • Someone on internet used hexdump /dev/sda1 -C | grep 'LABELONE' : is it possible to find the beginning of an LV ?
    – Bertaud
    Feb 28 '16 at 18:10
  • Well, you'll have about a ⅔ chance of finding anything you look for, so yes, maybe. But what will you do once you find the beginning of one of the LVs? With ⅓ of its contents outright missing it will be too far gone to fsck or mount.
    – Celada
    Feb 28 '16 at 18:13

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.