Tests

SMART self-tests (“short” & “long”) results.

Perceived results' randomness made me create a script to automatically reiterate the read-verify cycle.

Its results prove that a issue exists: SMART self-tests fails (at a seemingly random sector), but reading data from said sectors succeeds.

Writing zeros on the entire disk (i.e. dd if=/dev/zero of=/dev/sdg bs=64M oflag=direct) succeeds too, but then filling the drive with real data made disk restart throwing SMART (197/Current_Pending_Sector) errors (after a few hours?).

These steps were repeated several times for a week with same results.

Warranty

Obviously, the disk was scrubbed and uniquely marked before being sent to warranty.

On Dec. 28 2012 disk was sent to vendor, including a note describing aforementioned issue, tests & behavior.

It was returned to me on Jan 10 2013 with the resolution “test ok”.

They did not run any write test! All signatures were still on disk.

Seemingly, the only SMART test done by service techs was the conveyance test.

As they left no (other) trace in SMART logs, I guess they decided disk's functionality relying on Data Lifeguard Diagnostic for Windows's extended (read) test results. (Wrong assumption, they didn't run this test either, as proven later).

Though this software offers a write (zeros) test! Even if it wasn't really useful in this case, using it would have at least proven their commitment in diagnosing my hard drive.

Pitifully, this software fails to read SMART attributes properly:

Can it be (re)used?

I'll assume vendor's techs have smarter methods to determine disks state, even without “touching” it (alas, “writing”).

Yet before reinstating the disk I want to confirm that the diagnostic tool provided by disk's manufacturer (“WinDlg_124”) can't see any issues:

smart attr. info (as of WD);
initial smart status;
disk properties;
quick test;
“Extended Test” crashed along with WinDlg after around 4 hours; I decided to not try running it again;
write zeros;
write zeros results;
quick test.

WinDlg says „everything”'s ok.

(Trying to) reuse

Disk was reinstalled, formatted and filled:

Jan 11 04:07: powered up
Jan 11 04:10: run a series of short SMART self-tests and various raw read/writes. (No errors occurred, so I run a conveyance test too, as “log mark”)
Jan 11 06:02: formatted, mounted and started data fill up;
Jan 11 17:55: disk full
Jan 11 18:14: clients got access to disk (i.e. back in production – reads 99%, seldom writes)
Jan 12 10:07: SMART errors restarted throwing
Jan 12 12:20: started data salvage (copying files to other disks)
Jan 13 01:50: data salvage completed

Test again

Jan 13 03:06: run a series of short SMART self-tests. No error occurred.
Jan 13 04:03: running a long SMART self-test;
Jan 13 14:00: finished long SMART self-test: failed.
Jan 13 17:00: running extended read test in WinDlg;
Jan 14 06:00: test hasn't finished yet, but already generated many SMART errors
Jan 14 07:00: test finished and WinDlg acknowledged the errors, even if it still shows wrong SMART info.

Ok, given the latest test results obtained using disk manufacturer's endorsed diagnostic software, I stop here.

On Monday, January 14, at 8:00, PC Garage was informed about this analysis.

Conclusions?

Unfortunately PC Garage proves once again the “competence” of Romanian IT&C retailers. :(

Here is WD's answer (ref. #011313-6484706):

As our diagnostic software shows, the drive has too many bad sectors, which means, that it can't be repaired and needs to be replaced.
If the vendor refuses to do it, you can have it replaced by us. To do so, please follow these instructions: http://websupport.wdc.com/rdsfdc.asp?linktype=rmacreate&portaltype=wd&custtype=end&fs=&ss=&lang=en

I'm still waiting for PC Garage's position.