ST3500320AS offline uncorrectable 600+, zero fill "fixes" it

[rowan194]

Hi all, I found this forum last night and have enjoyed browsing through it. Some of the recovery stuff goes way over my head but I'm still interested in understanding the technology and inner workings of hard drives, since I maintain around 25 of them.

My question today is about strange behaviour from a Seagate ST3500320AS (500GB) drive that is about 14 months old. It dropped out of a RAID1 mirror, a self test failed, and the current pending sector and offline uncorrectable SMART values were non zero. I performed a zero fill to try to evoke any further errors but all this seems to have done is DECREMENTED the bad sector related values back to 0 (and it also has NOT incremented the reallocated sector count - that is still 0!). If it wasn't for spin retry count being >0 (and two UNCR events in the SMART error log) you would never know there was a problem with this drive.

So I put it back into the mirror to see what would happen; a few seconds into rebuilding it started clicking again and eventually dropped out with the bad sector SMART values going >0 again. One more zero fill, perfect, back into the mirror, click click click. It's reporting 600+ bad sectors but a zero fill consistently "fixes" the problem. At this point I left it for a few days.

It gets stranger... now that I've removed the drive and put it in another machine it hasn't skipped a beat. An MHDD scan looks reasonable, with only 45 sectors <150ms.

I'm confused why this drive is not incrementing the SMART reallocated sector count, if it's remapping bad sectors then surely this should be >0 ?

Here's some relevant bits from smartctl -a:

Code:

  5 Reallocated_Sector_Ct   0x0033   100   100   036    Pre-fail  Always       -       0
10 Spin_Retry_Count        0x0013   100   099   097    Pre-fail  Always       -       948
197 Current_Pending_Sector  0x0012   100   100   000    Old_age   Always       -       0
198 Offline_Uncorrectable   0x0010   100   100   000    Old_age   Offline      -       0

  40 51 00 2d eb 10 00  Error: UNC at LBA = 0x0010eb2d = 1108781
  40 51 00 d8 ca 08 00  Error: UNC at LBA = 0x0008cad8 = 576216

SMART Self-test log structure revision number 1
Num  Test_Description    Status                  Remaining  LifeTime(hours)  LBA_of_first_error
# 1  Short offline       Completed without error       00%     10340         -
# 2  Short offline       Completed: read failure       90%     10302         597072
# 3  Short offline       Completed: read failure       90%     10294         22095
# 4  Short offline       Completed without error       00%     10289         -
# 5  Extended offline    Completed: read failure       90%     10264         3910350
# 6  Short offline       Completed without error       00%     10264         -
# 7  Short offline       Completed without error       00%     10249         -
# 8  Short offline       Completed: read failure       90%     10247         211917
# 9  Short offline       Completed without error       00%     10218         -
#10  Extended offline    Completed: read failure       90%     10216         256509
#11  Short offline       Completed: read failure       90%     10216         256509
#12  Extended offline    Completed without error       00%      8969         -


Thanks in advance for any comments...

[rowan194]

PS: I've had bad luck with drives so every server I look after has a minimum of RAID1 protection, and I do regular backups, so I'm not going to argue with you guys about how I can recover data myself for less than $50.

[drc]

Could be soft bads that are fixed by rewriting them.

[rowan194]

Could be soft bads that are fixed by rewriting them.

Yep, I understand that much, but shouldn't that increment the reallocated sector count? This is the behaviour I see with other drives.

[rowan194]

After re-reading your reply I guess you meant that it's a marginal error, and rewriting the same sector (rather than remapping) ends up restoring it? Hmmm.

[drc]

Right. Of course if it keeps happening over and over again then I would say the drive is having problems.

[rowan194]

I'm going to use a simple program that I wrote which reads and writes random lengths of data on random portions of a disk, I don't really trust this drive any more and without it [currently] playing up I can't RMA it, so I may as well give it a bit of a workout and see what happens...

[drc]

If you really want to kill it just use spinrite or something...

[rowan194]

If you really want to kill it just use spinrite or something...

Yeah, I actually did a search on this forum for spinrite a day ago, it's entertaining to see some of the replies.

I used it probably 15 years ago on an old MFM drive. Even back then terms used in the readme such as "scrubbing" sounded a little wacky and far fetched. I presume with today's proprietary electronics presenting a digital only interface low level access isn't anywhere near possible these days.

I once wrote a program to "refresh" floppies on an old (circa 1982) microcomputer. If it came across a bad sector it would repeatedly try to read it until valid data was returned, then it wrote out the sector again. It usually worked. Sounds familiar?

[BlackST]

About floppies : how can you write a sector if the beginning mark or gap or preamble or guard band are messed up ? Maybe reformat entire track then rewrite the sectors... but what if the problem is physical i.e. magnetic layer dropout or media damage ?
On modern drives it is practically impossible to do such actions, except ask the drive to read or read without CRC control a block, then rewrite it.
You have some sort of control on reallocation and other mechanism(s) if you have enough info about firmware.... but forget direct access to internal DSP/decoding/encoding unless you work for the manufacturer or you have done extensive research on firmware and hardware structure of THAT drive/family - and, final note, it is totally useless except for self-educational purpose....

[rowan194]

About floppies : how can you write a sector if the beginning mark or gap or preamble or guard band are messed up ? Maybe reformat entire track then rewrite the sectors... but what if the problem is physical i.e. magnetic layer dropout or media damage ?

This was just a simple program to try to recover a floppy that had degraded through normal use, formatted in one machine, half the sectors written in another machine, then some of those updated and more written out in another... like I said, it usually worked, I was a 15 year old kid, lighten up a bit.

On modern drives it is practically impossible to do such actions, except ask the drive to read or read without CRC control a block, then rewrite it.
You have some sort of control on reallocation and other mechanism(s) if you have enough info about firmware.... but forget direct access to internal DSP/decoding/encoding unless you work for the manufacturer or you have done extensive research on firmware and hardware structure of THAT drive/family - and, final note, it is totally useless except for self-educational purpose....

That's why Spinrite's claims seem so far fetched. I saw a screenshot on grc.com which showed some sort of head amplitude graph... through an IDE/SATA interface??