1

After experiencing slow copying from an 2.5" USB hard drive, I copied off all data, except for a few files which were unreadable, and then deleted and recreate the partition, and quick formatted the drive under Windows XP without error. However, the SMART status on the drive indicates a "read error". Both the short and long SMART tests abort after a few percent and report a "read error". I am now doing a full format on the drive. After a long (about half way through the format), the drive light stopped flashing, but the format did not abort. I canceled it and successfully completed a quick format again.

I would like to understand why the drive can be formatted successfully if it appears that the drive is about to fail or is unusable? If the SMART tests abort with an error, how can the drive still be formatted? This answer suggests that SMART is not so smart; does it work the other way around though -- a drive which SMART thinks is bad is actually not bad?

I presume I should use this drive as a door-stop or paper-weight? It's a few years old, but has not had very much use. It has been carefully looked after and is supplied in it's own "shock resistant" enclosure.

Edit: smartctl -H under Debian sid shows "PASSED".

Edit: badblocks (default read-only mode) has run for ~75 minutes and is 45% complete. It has found 4 read-errors.

Edit: I stopped badblocks and noted the numbers of the four bad blocks. I then ran it again with -n just before the bad blocks and let it run over them. I repeated this with the destructive write (just the first pattern) and let it run over the area again. I then tested in read-only mode and the blocks were not detected as bad. I returned to palimpsest and refreshed the SMART data, which now reported the drive as healthy. I repeated the short self-test in palimpsest which I had done before, and now it did not abort, but completed and reported that the disk is healthy. Possibly writing to those bad blocks forced the drive to find the problem and swap the bad blocks out.

2 Answers 2

4

"The purpose of S.M.A.R.T. is to warn a user of impending drive failure while there is still time to take action, such as copying the data to a replacement device." - Wikipedia

S.M.A.R.T errors are predictive and not always an actual sign that the drive has failed. Quick formats do not read/write to every potentially bad block on the drive, hence why it might have been successful in your case.

If you're getting any kind of errors, I would definitely think about replacing the drive.

1
  • 2
    @SabreWolfy Consider a paper book, written in pencil. A quick format merely erases the title and the index. The book is now 'empty' and later writes to it are actually erase a page contents, write new data, adjust index. A Full format would erase all the pages and draw new horizontal lines to help you write. That full format will see all pages and will error if one is torn. A quick format never touches most of the books and will pass, even if part of the drive/book is dead/broken.
    – Hennes
    Feb 10, 2013 at 15:12
0

Run a SMART selftest:

smartctl -t long /dev/sdX

It will take some time (several hours for big hard disks), after that you can see the result with

smartctl -a /dev/sdX

in the SMART Self-test log block. This selftest feature has been a 100% reliable so far - if it says "Completed without error", the disk is OK, if it reports any errors, throw it away.

3
  • As indicated in my question: 'Both the short and long SMART tests abort after a few percent and report a "read error".'
    – SabreWolfy
    Sep 16, 2013 at 12:46
  • @SabreWolfy - A read error would indicate the drive should replace. The fact you also have bad blocks indicate the disk surface has serious problems.
    – Ramhound
    Sep 16, 2013 at 12:50
  • 1
    @SabreWolfy: Sorry, must've overlooked that when I was answering. The "if it reports any errors, throw it away" of my answer still stands, though :) Sep 17, 2013 at 17:55

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .