[Sussex] ext3 data loss

[Steven Dobson]

Brougham

On Tue, 2006-03-28 at 21:40 +0100, brougham Baker wrote:
> The dog pulled out the power to my file server- old bit of junk but it does
> well enough for me. It's a debian box and has been reliable until this
> event.
> Uname -a says
> Linux wesley 2.4.20-bf2.4 #1 Wed Dec 25 13:17:08 UTC 2002 i686 unknown

Bad luck.

> When it came back there were some errors on /dev/hda1 which were fixed and
> then i rebooted again as /dev/hdc1 and /dev/hdd1 didn't mount.
> 
> This time /dev/hdc1 mounted properly but /dev/hdd1 didn't and hasn't ever
> since. I booted the box with a knoppix cd and tried to mount the disk read
> only- it said there were no valid partitions on hdd. I tried took an image
> of the drive to another disk to keep that safe and then tried a couple of
> read only tools on the desk (i forget which, it was a few months ago), the
> drive has sat next to the monitor since then.
> 
> I then booted a Ultimate Boot CD for Windows CD (based on bart but with
> loads of other tools) and ran one of the SMART read only tools- this passed
> saying no errors, i then ran the IBM fitness tool from the Ultimate Boot CD
> (dos) including a full read of the whole disk- this also passed. I can
> presume from this that the disk is physically fit and that only the
> partition info is screwed.
> 
> The drives are mounted like this-
> /dev/hda1       /               ext2    errors=remount-ro       0       1
> /dev/hda2       none            swap    sw                      0       0
> <snip>
> /dev/hdc1       /data   ext3    defaults                        0       2
> /dev/hdd1       /data/video     ext3    defaults                0       2
> 
> I got a list of the superblocks and tried them all- all of which said they
> were invalid. I then booted back into knoppix and looked at the sectors on
> the disk- my files are there.
> 
> I then tried Stellar Phoenix for Linux and that found several partitions but
> none where the correct ones- some started a little far in and some where
> wildly off. Then there was another tool that ran under windows but was
> designed for ext2/3 partitions but that kept crashing and never finished the
> scan- which took over 20 hours.
> 
> All my movies/series are on there (no pr0n)- is there anything else i can
> do?

I sounds to me like it maybe the partition table that got corrupted.
Has you tried the command like tool fdisk(8) or cfisk(8) to see if they
can read the partition table?  Can the kernel see the partitions?  That
would be another indication that the partition table is bad.  

First take a look at dmesg(8) to see if the kernel is seeing any errors
with the disk?  ($ dmesg | grep hdd).  On my system when I look at it I
see three bits of information that are relavent to this e-mail thread:
 hda: MAXTOR 6L040J2, ATA DISK drive           <-- Kernel see drive
 hda: hda1 hda2 < hda5 hda6 hda7 hda8 hda9 >   <-- Kernel see partitions

     And worrying the following errors

 hda: dma_intr: status=0x51 { DriveReady SeekComplete Error }
 hda: dma_intr: error=0x40 { UncorrectableError }, LBAsect=39618818,
                                   sector=39618817
 end_request: I/O error, dev hda, sector 39618817

Looks like I'm going to be buying a new disk this weekend.  Well this
system needed a SATA driver.

Steve
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 191 bytes
Desc: This is a digitally signed message part
Url : http://mailman.lug.org.uk/pipermail/sussex/attachments/20060328/fbf97d05/attachment.pgp