Dying Hard Drive

drasnordrasnor Starship OperatorHawthorne, CA Icrontian
edited October 2009 in Hardware
A coworker just handed me his laptop's dying hard drive with several years worth of not-backed-up photos on it. The drive spins up fine and mounts in my USB enclosure but file operations fail with "Delayed Write Fail" shortly after I try to do anything. On Linux, I can't get a drive image with dd because the drive won't let me pull more than a few gigs before it won't transfer any more. On Windows, I can't seem to get anything to copy before the Delayed Write Fail. The fun part is that I've got write caching disabled on the drive in my Windows Device Manager so delayed writes shouldn't be happening. I've managed to take ownership of most of the photos I intend to save before the delayed write fail prevents me from getting more but I keep thinking there's got to be a better way to do this, especially since Explorer hangs before I can copy any photos off of it.

What I really want to do is get a raw drive image and work on the filesystem on a good disk. Any ideas?

-drasnor :fold:

Comments

  • kryystkryyst Ontario, Canada
    edited September 2009
    I've had luck before using Norton Ghost to recover drives suffering from that exact problem. Also throw a bag of frozen peas on top of it. Basically just ghost it to a new drive, or partition and then recover from that.
  • edited September 2009
    Since it starts reading first then fails, I don't think the problem is bad magnetic media. It seems more like a controller board failure (if not r/w head or electric motor). If you can find another disk of the same brand/model, you can try swapping the controller boards. We had a WD disk saved by this method. I am not sure if you can swap the motors but it might also be possible.

    PS: Just in case, make sure you have connected the auxiliary power of external USB enclosure. Especially with the old/failing disks, 500mA USB power may not be sufficient.
  • drasnordrasnor Starship Operator Hawthorne, CA Icrontian
    edited September 2009
    It's an ancient Toshiba drive. I'm having some luck accessing the drive with the deprecated Linux raw I/O driver and dd but I've been getting I/O errors at the 3.1-3.2GB mark consistently. I'm going to try ddrescue later today since it's designed to deal with that sort of thing. The trouble is now that I can only get data off the drive at 45 kB/s so it's going to take 10 days to get a drive image.

    -drasnor :fold:
  • edited September 2009
    If the failure is consistently at the same point, then it might be a bad magnetic media too. Have you checked what it is reading? I would check it before waiting for 10 days, it might be reading garbage.
  • drasnordrasnor Starship Operator Hawthorne, CA Icrontian
    edited September 2009
    mirage wrote:
    If the failure is consistently at the same point, then it might be a bad magnetic media too. Have you checked what it is reading? I would check it before waiting for 10 days, it might be reading garbage.
    I have no idea. Short of firing up the image in a text editor and looking for strings of ASCII-encoded text I wouldn't know what I'd be looking for. The drive image I'm taking includes the MBR so it's not mountable as-is. I'd have to write the image to a drive and then try to mount it and I'm not sure that 8% - 10% of a NTFS filesystem will have enough integrity to mount. Since I can mount the drive under Windows for short amounts of time I think I'm probably not getting garbage.

    -drasnor :fold:
  • drasnordrasnor Starship Operator Hawthorne, CA Icrontian
    edited September 2009
    Success! dd_rescue was able to work around the bad blocks and get an image. The image didn't have enough filesystem integrity for chkdsk to work but did have enough to where I could get the files he wanted off. The disk turned out to have 1 MB worth of bad blocks in three locations in addition to bad integrity from being turned off during write operations a bazillion times. We got lucky.

    -drasnor :fold:
  • edited September 2009
    Congrats! Glad that it did not take 10 days :)
  • drasnordrasnor Starship Operator Hawthorne, CA Icrontian
    edited September 2009
    mirage wrote:
    Congrats! Glad that it did not take 10 days :)
    Yeah, dd_rescue and dd both use the direct I/O ioctl by default which can access block devices faster than with the raw I/O driver. My original concern was that some problem with DMA and read/write caching was screwing me up but dd failing on both raw and direct I/O at the same location seemed to suggest media damage. Dd_rescue with direct I/O got the job done in about 16 hours for 37 GB.

    -drasnor :fold:
  • edited September 2009
    Yes, failure at the same location is usually the sign of bad sectors. Your friend is lucky that his/her files are not destroyed by those sectors and there was someone who could use Linux (or was it BSD?).
  • drasnordrasnor Starship Operator Hawthorne, CA Icrontian
    edited September 2009
    The software I used is available on any POSIX OS but I used Linux because I have it and am familiar with it.

    -drasnor :fold:
  • ardichokeardichoke Icrontian
    edited October 2009
    I'd just like to chip in here with I <3 dd_rescue, I saved quite a bit of data with it back when i was a student system admin in college. Great software.
Sign In or Register to comment.