BLU's server crash and data recovery

[Bill Bogstad]

On Tue, Sep 9, 2008 at 11:05 AM, Jerry Feldman <gaf-mNDKBlG2WHs at public.gmane.org> wrote:
> On September, 1 we experienced what appeared to be a disk failure.
> We have a Dell Poweredge 1750 with hardware RAID and 3 Ultra SCSI 320GB
> drives that JABR set up as Raid 5. I also have in my possession 2 Dell
> Poweredge 1750s, and I can confirm on these systems that the drives are
> problematical. Currently the 1750 is reporting 2 failed units and 1 good
> unit. I also think that oe of the drives may have been unused. I also
> confirm that all 3 drives successfully spin up in both of the Dells and i an
> Intel whitebox system.
>
> One thing I am going to try to do is to dd each of the drives as soon as I
> can acquire some space (possibly on a USB SATA). When my home 40GB drive
> died while installing a new system at home, I did some research on data
> recovery companies and the cost usually runs at or above $1000 per drive.
> This would place the BLU in debt :-)
>
> I'm going to try to see if I can do a 'dd if=/dev/sdx of=/dev/NULL' to see
> if the drives are readable. When I placed those drives into the Intel
> system, when Centos started up, the initial vgscan reported some I/O errors
> on one of the 3 drives, but on subsequent boots no errors were reported.

Look up ddrescue.  It's a variant of dd which tries hard to skip over
bad blocks in order to recover every possible block of data it can.  I
believe you can run it multiple times and it will keep track of which
blocks it has already successfully recovered and will retry failures.
If your problem is in any way intermittent, this might let you recover
everything...

Good Luck,
Bill Bogstad