[svlug] Files That Will Not Delete

[Michael Eager]

On 11/02/14 17:18, Scott DuBois wrote:
> On 11/02/2014 05:10 PM, Michael Eager wrote:
>>
>> When it completes, run "smartctl -a /dev/sdd".
>>
>
> Ok, now since badblocks gave me _zero_ errors, what could we
> _theoretically_ expect from running this test? Since this is my first
> time running these tests, I do not know what I'm looking for or should
> expect as a "baseline".

You'll see a list of the internal counters in the drive.  These
counters list the current value, the worst value, and the threshold
to be considered an indicator of a potential failure.  You should
see just dashes under "When Failed", otherwise there might be a problem
with the drive.

There should be a listing of errors logged.  It isn't unusual to
see that errors have been corrected.  Lots of corrected soft errors,
or an increase in soft errors, can indicate that a drive is aging.

There will be a listing of self-test results saying that whether
any errors were detected.

Smartctl is a pretty conservative test of drive health.  A "PASSED"
result does not mean that the drive has no bad blocks or is in otherwise
perfect condition.  But a disk which shows failures in a test or is showing
age indicators should be replaced before a potential failure becomes an
actual failure.  (This can happen much sooner than expected, speaking
from experience.)

> I'm also a little concerned since I was expecting badblocks to have a
> runtime of between 3-4 days and it completed in just over 6 hours. Since
> I posted my entered string, I'm confident I typed the operation to be
> executed correctly. The file it produced had a null value associated
> with it.

Time estimates depend on many factors.  YMMV.

If badblocks did not encounter any bad blocks there are several explanations:
  1) There are no bad blocks
  2) You ran it in read-only mode and that didn't detect any bad blocks.
     (Write faults are not detected.)
  3) The disk failure mode is data dependent (not very likely).
  4) The disk hardware successfully corrected any soft failures.
  5) Bad blocks on the disk have been reassigned to good blocks.
  6) Some other reason.

There are a number of reasons why you might have gotten a I/O read
failure.  The most likely is a disk problem.  Next is connector
problems.  Then there are disk controller problems, flaky wires,
random Alpha particles, and many more things that could happen.
If after testing the disk you can't find any errors, use the disk.
Run smartctl every so often to check disk health.  Run the smartd
service to get messages if your disk starts acting flaky.

There are other disk test programs.  Hitachi makes a good one and
Seagate has one for their drives.  Both are available for download.
At some point you have to decide that you've tested the disk
enough to have confidence it isn't going to fail, or you decide
to put it in the recycle bin.

-- 
Michael Eager	 eager at eagercon.com
1960 Park Blvd., Palo Alto, CA 94306  650-325-8077