Determining if 1TB Seagate is really faulty

[RPT]

How serious is 1 UNC error detected by MHDD?
Does the drive has to been returned to manufacturer as failing?

Defect was discovered after SQL had stopped Windows XP workstation whole system, because of this single unreadable sector.

Even if the process stopped wasn't too important, IMO such disk is unreliable and can't be of much value in anyone PC.

Now 2 weeks later the defect has gone, probably listed as bad and hidden by drive logic.
How to decide, if this disk should be returned to manufacturer as faulty???
If once error has already stopped full PC system functioning, isn't that enough to prove, that disk isn't up to the task?

Perhaps there are some tools to obtain more precise info about disk health?

Drive type: ST3100520AS 1TB drive

Thanks!

[Nick_CT]

Have a look at its SMART status, you'll most likely find proof enough there that the HDD should be swapped out.

[RPT]

HDDScan 3.3 smart report is all greens, is there something else to help decision making?

Some years ago HDD manufacturers used to build into drives bad sector remapping ability, I guess it also present in nowadays SATA drives.

When there was UNC, which already has stopped PC normal functioning and UNC is not found anymore, probably it's remapped and drive is just trying to hide it's unreliability. The HDD technology has reached level, when one just can't decide upon pure technical evidence.

[pclab]

Try with MHDD or with the Seagate tool: SeaTools.

[rowan194]

When there was UNC, which already has stopped PC normal functioning and UNC is not found anymore, probably it's remapped and drive is just trying to hide it's unreliability. The HDD technology has reached level, when one just can't decide upon pure technical evidence.

I noticed that one of my Seagates seems to lie about past issues with media. Detailed it here: st3500320as-offline-uncorrectable-600-zero-fill-fixes-t13359.html . You can see from the self-test log that there were at least 5 different UNC LBAs, yet the drive shows 0 uncorrectable/reallocated after a zero fill. It's been sitting unused since then (~18 months ago), I don't trust it any more...

[Vulcan]

I noticed that one of my Seagates seems to lie about past issues with media. Detailed it here: st3500320as-offline-uncorrectable-600-zero-fill-fixes-t13359.html . You can see from the self-test log that there were at least 5 different UNC LBAs, yet the drive shows 0 uncorrectable/reallocated after a zero fill.

I understand that this behaviour is not intuitive, but the likely reason was explained by drc <hat tip> in your linked thread. Put simply: UNC does not necessarily mean that the sector will be reallocated by the drive on the next write. The process used by the drive to decide whether or not to reallocate, is more complex than that.

Whilst I disagree with some of the OP's assertions in this thread, it is certainly possible (as you have also experienced), for there to be evidence that humans can see, showing a drive is not behaving correctly - but for it to be difficult for end-users to get objective evidence from the drive (e.g. SMART failure) to confirm that. (There are some sources of information which are not available to the end-user, but obviously that doesn't help end-users who are trying to make this fault determination themselves. ).

Since the drive manufacturer has little incentive to flag a "slightly flaky but might last a few years" drive as faulty (since a SMART failure == warranty claim), this is one reason why RAID arrays have to try to second-guess the drive behaviour, and potentially decide that a drive is having a problem and take appropriate action, even if there is no specific error reported by the drive itself - although this process is also open to errors and wrong guesses!

To return to your comment: I see no evidence in your thread that your drive is "lying" - it is reporting a non-intuitive result, due to its proprietary treatment of the previously-UNC sectors. You may want to test your Seagate drive using SeaTools, since that utility has previously been shown to correctly report Seagate drives with "odd" behaviour as faulty, even when the drive was unfortunately not yet reporting a SMART failure.

[RPT]

Question is not so straightforward.

Application already has stopped to work because of this disk read failure. MHDD confirmed presence of UNC and I replaced the disk with new one.
Now, several weeks later, prior to shipping faulty drive back to warranty, when running mhdd once again the UNC has gone and SMART doesn't show any evidence of previous failure.

It's obvious that one doesn't dare to use the same disk again in any other PC, because diagnostic and recovering from data loss after first disk error took approximately 2 days.

Now it appears, that there even isn't any proof of past failure and formally there isn't now any reason to return disk under warranty return?

[Vulcan]

@RPT: Are you replying to me? If so, I think you have misunderstood my post above.

I was replying to rowan194 and not to you - that's why I didn't quote any of your text in my reply. I was not trying to tell you anything about your situation. Unfortunately this particular phpBB forum software doesn't make it easy to show exactly who you are replying to...

I hope that helps to explain why you seem to think my previous post did not relate to your situation - that's because I was not discussing your situation! I was just replying to the post immediately above my previous one.

[RPT]

Thanks for input Vulcan.
Sorry, I read your message and thought to elevate my original posting hoping to deserve someone's attention.
But obviously such existence philosophy is out of this forum reach

[rowan194]

I noticed that one of my Seagates seems to lie about past issues with media. Detailed it here: st3500320as-offline-uncorrectable-600-zero-fill-fixes-t13359.html . You can see from the self-test log that there were at least 5 different UNC LBAs, yet the drive shows 0 uncorrectable/reallocated after a zero fill.

I understand that this behaviour is not intuitive, but the likely reason was explained by drc <hat tip> in your linked thread. Put simply: UNC does not necessarily mean that the sector will be reallocated by the drive on the next write. The process used by the drive to decide whether or not to reallocate, is more complex than that.

Understood. I should probably mention that the uncorrectable errors started appearing during a RAID rebuild (ie, all writes), but then a subsequent zero fill passed without a hitch. Put it back into the RAID, more errors. Zero fill, she's fine. Both the self-test log and the relevant SMART variables (600+ uncorrectable/current pending, after zero fill 0) showed there were bad sectors, so it was the drive itself doing that. Definitely something funny going on...

I do have a Seatools CD but the SATA DVD-ROM I have doesn't want to work with my test rig. Running an MHDD scan instead.

[Vulcan]

@RPT: I don't understand the meaning of your comment about "philosophy", nor why you seem to criticise this forum. You can try posting your questions to the Seagate forum, if you are unhappy with this one...

I chose not to reply to your earlier posts in this thread, because there were too many questions - some (many?) of which were a matter of opinion, and therefore had no conclusion. Therefore I could give lots of info, and still not satisfy you - so I decided not to try. IMHO I have offered enough help and spent lots of time, with your previous threads.

As pclab suggested to you, running SeaTools would be a good idea IMHO - because as I explained to rowan194, that utility has sometimes been seen to correctly recognise a faulty Seagate disk, even when the SMART data does not give the same evidence. In any case, I wish you good luck.

===

@rowan194: In this specific case, running MHDD is not an alternative to running SeaTools. The evidence from other users here, is that SeaTools can (correctly) recognise a faulty disk (sometimes), even when other test software showed no obvious indications. I know that fzabkar hangs out in the Seagate forums, so he might have seen more examples of that, over there.

Modern disk drives typically have proprietary internal error logs, which are much, much more detailed and are not shown in the SMART output. Therefore one hypothesis for how SeaTools testing is different than normal SMART checking, could be that SeaTools has a method of getting at least some proprietary information from Seagate disks, in order to make a better determination of disk drive health - but since it's proprietary software, we don't know exactly how it works.

You mentioned RAID "rebuild" and then said that this was "all writes". A rebuild would not be all writes to all disks. If you are doing a rebuild of an existing, live, RAID volume by inserting a replacement disk, then there will be reads of the existing disks in the same volume. Did you mean that there were "all writes" to this disk because this was the new disk during the rebuild of an existing RAID volume? Even in that case, I know of at least one RAID controller family where there would be some reads, during the rebuild.

It would be interesting to see the SMART error log, not the self-test log (specifically what commands returned the UNCs) before you did the zero-fill. Did you capture the SMART error log (which is part of the "smartctl -a" output) at that point?

As I mentioned earlier, the lack of reallocations after UNC, means that the disk tested the UNC LBAs when it did the next write to those LBAs, and found those physical media locations to be reliable - and hence did not reallocate them. This suggests that something happened during the writing to those LBAs to make them unreadable, but a later write is successfully readable.

Unfortunately, as you said in your previous thread, unless you could provoke this reproducibly by testing, or make the drive much more "clearly" faulty than it currently is, I don't see how you can prove that the drive is faulty to the manufacturer (bearing in mind that they have their own idea of what "faulty" means). That's why it would be great for you, if SeaTools does find a reason to report the drive as faulty - but Seagate won't care about any MHDD output (other than a SMART failure), which is why I said it's not a substitute for you, in this case.

I have one other suggestion, based on a fault which I saw some years ago. In your previous thread, you said:

now that I've removed the drive and put it in another machine it hasn't skipped a beat

I wonder if yours was a power-related issue, in the original machine? I have seen power fluctuations cause problems when writing, thereby making seemingly random sectors unreadable, later rewrites were OK, and a replacement drive was not affected (it turns out, in that case, to have been a problem with the power-connector, but obviously that's not the only power-related possibility).

Hope that helps.

[fzabkar]

Now it appears, that there even isn't any proof of past failure and formally there isn't now any reason to return disk under warranty return?

If you can get the drive to fail any of the SeaTools tests, then SeaTools will generate a Test Code that is uniqe to your drive. This error code is derived from your drive's serial number, and the number of the failing test.

Once you have this test code, your drive will automatically qualify for an RMA, either now or at some future time.

About SeaTools Test Codes:
http://www.seagate.com/ww/v/index.jsp?l ... 48090aRCRD

My understanding is that the second last digit identifies the failing test, as follows:

xxxxxx4x - SMART Fail
xxxxxx5x - Long Test, long LBA test
xxxxxx6x - Short Test, short LBA test
xxxxxxAx - Self-Service SeaTools Test Code
xxxxxxCx - Short Generic
xxxxxxDx - Long Generic
xxxxxxEx - Short DST
xxxxxxFx - Long DST

[rowan194]

Thanks for the detailed explanation, Vulcan. In my case doing the test would be for curiosity's sake only, a 500GB drive isn't worth it - for the cost RMA'ing the drive to get back a dodgy refurbished unit, I'd rather buy a brand new WD or Hitachi.

This drive was the last time I purchased from Seagate. I'd made that policy decision well before this drive started playing up, but it seems appropriate (in an ironic sort of way) that my experience with this final drive wasn't really any different to many of the other Seagates I've had.

My worst Seagate experience (but not the only bad one) : I purchased 4 identical drives, and over time several failed (including their replacements). The original 4 were replaced under warranty with 2 new drives, and 6 recos. There were 7 failures in all over around 2.5 years. At that point I gave up and changed my RAID array (only had one at the time) to WD.

[BlackST]

Just my 50 cents , always the same coin:

1) 90 out 100 failures on DISK subsystem are due to SYSTEM flaws especially at power.
2) On these, one recurrent is the use of cheap, mainstream disks on systems 24/7 on. Why not use enterprise class drives or move to SCSI if availability / reliability is a must ?!?
3) WDs are even more unstable than Seagates on many circumstances.
4) SMART is not reliable for failure prediction anymore (too many variables at play).
5) Time ago there was shortage (here) for seagate brand new drives while the distributors had a lot of FACTORY RECERTIFIED drives. I installed a lot of them, in NON critical systems, and they are running flawlessly (on normal/moderate use) since end of 2006. I constantly monitor these systems and the drives are just aged, no problems. These systems have redundant power supply and are under UPS.Many are 24/7 on ,the vast majority have 2 to 4 power cycles every day. Either I am lucky, or there's something weird elsewhere.
6) Back to point 3) : remove the cause and the effect will be gone - but first you have to find it. A post-mortem diagnose could be useful.
7) Last: the bigger the drive, the more the short term failure probability.

I could be more technical but I am not allowed to...

[Vulcan]

@rowan194: Thanks for the update.

In my case doing the test would be for curiosity's sake only, a 500GB drive isn't worth it - for the cost RMA'ing the drive to get back a dodgy refurbished unit, I'd rather buy a brand new WD or Hitachi.

Understood - and, of course, if you don't find a test which reports a fault, that doesn't mean that the fault has been solved. I was just suggesting ways in which you could make more progress, if you wanted to do that. Your comment about the suspect drive seeming to have no further problems, when installed in another system, is very interesting IMHO. Good luck!

[RPT]

I have run Seatools, Smart is "Passed" and I have run several Basic tests -> long generic to my Drive type: ST3100520AS 1TB . All passed.

There is no sign of past headache this drive has caused to me.

When there was a failure to database access because of UNC, there was also error in application log, describing that data not accessible ...
I dismounted HDD switched to other PC and run HDD to get UNC confirmation. The UNC error was there without any hurry to leave.
I replaced failured disk 2 days later with WD drive and when copying from failed Seagate to WD, the error was still there, exactly the same file not able to copy.

What else proof one needs, what to do next with this disk?

A month later just for fun prior to returning to RMA I run a test once again and UNC had left the building.

From my point of view there is no choice - the whole disk must go.

[BlackST]

It's a completely pointless discussion. Drives DO self heal the grown defects when and if possible. Live with it or use different drives for storage as these cheap drives are NOT intended NOR designed for enterprise use.
Better to have an array of smaller, enterprise class 250-300 GB maximum GB drives, possibly SCSI and in some redundant / fault tolerant configuration (if a monster capacity is needed) than have such timebombs if reliability is a problem.
Of course it's gonna cost you much more for same capacity than a single, 50$ drive.

[RPT]

It's a completely pointless discussion. Drives DO self heal the grown defects when and if possible. Live with it or use different drives for storage as these cheap drives are NOT intended NOR designed for enterprise use.
Better to have an array of smaller, enterprise class 250-300 GB maximum GB drives, possibly SCSI and in some redundant / fault tolerant configuration (if a monster capacity is needed) than have such timebombs if reliability is a problem.
Of course it's gonna cost you much more for same capacity than a single, 50$ drive.

YES, I understand this but most of people do not understand.

The expected resolution of this thread was something like that ...
that nowadays cheap hard disk technology only pretends to save our data and so damned well, that most of time we even don't understand how close is total data loss.

Therefore I do backups and promise to keep the secret.

[fzabkar]

2) On these, one recurrent is the use of cheap, mainstream disks on systems 24/7 on. Why not use enterprise class drives ...

What is the difference between "mainstream" and enterprise class drives? For example, how do 7200.11 and ES.2 drives differ, apart from firmware?

[BlackST]

There is , and the world is not limited to Seagate.