RMA new Seagate drive for 187 SMART error?

[SimonAG]

I have completed 1 cycle of preclear with pre read and post read on 4 Seagate 6 TB NAS drives. 1 has come up with 1 187 SMART error. I have looked in to this and it seems there is a higher chance of failure of the disk if this SMART value is greater than 0. If I do warranty the disk they will send me a factory disk and not necessarily a new one. Would you RMA a brand new disk for a reported uncorrect 187 error? 

[John_M]

If it's "brand new" as you say, I wouldn't RMA it to the manufacturer. I'd return it to the vendor for a brand new replacement.

[SimonAG]

I didn't think they would accept that, I will try send it back to the seller and request a new one. 

[John_M]

26 minutes ago, Simontv said:

I didn't think they would accept that

 

They might not where you live. Where I live we have the Sale of Goods Act, which says that legally the contract of sale is between the vendor and the purchaser and that the goods must be fit for purpose. You said that the item was brand new and faulty and asked what people would do. So I told you what I would do in the same circumstances.

[pwm]

An important thing here is that one relocated sector doesn't really mean the drive is faulty.

 

Just about all drives have faulty sectors - you can't manufacture perfect disk surfaces. This was originally handled when the end users formatted their new drive and all older file systems have support for mapping sectors as broken in the file system layer.

 

But then the manufacturers moved the bad block mapping into the drive itself, making it possible to access the drive as if it has zero faults.

 

The initial factory test is intended to find all bad sectors and remap them. And the SMART data will not inform you about sectors relocated already at the factory - just as the SMART data will not report about the running time or number of read/written sectors at the factory.

[John_M]

5 minutes ago, pwm said:

An important thing here is that one relocated sector doesn't really mean the drive is faulty.

 

A good point. I didn't even consider that. I just answered the question "would you RMA a brand new drive?"

[SimonAG]

54 minutes ago, pwm said:

An important thing here is that one relocated sector doesn't really mean the drive is faulty.

 

Just about all drives have faulty sectors - you can't manufacture perfect disk surfaces. This was originally handled when the end users formatted their new drive and all older file systems have support for mapping sectors as broken in the file system layer.

 

But then the manufacturers moved the bad block mapping into the drive itself, making it possible to access the drive as if it has zero faults.

 

The initial factory test is intended to find all bad sectors and remap them. And the SMART data will not inform you about sectors relocated already at the factory - just as the SMART data will not report about the running time or number of read/written sectors at the factory.

 

This is sort of what i was reading in some other articles, which is why I created this thread, as I was unsure. I have asked the seller if they will send a replacement, I find that highly unlikely. I could ignore the error as some have suggested in other threads, I would have to acknowledge the smart alert every boot it seems. 

 

If they won't accept the RMA then i will probably hold on to it and RMA it before the warranty is up. 

[pwm]

29 minutes ago, Simontv said:

I could ignore the error as some have suggested in other threads, I would have to acknowledge the smart alert every boot it seems.

 

No, I don't think so. unRAID has logic to "watermark" SMART counters and just report if any counter has increased.

 

If the counter stay at one, then this is a perfectly fine drive. It's only if the counter keeps ticking that you know you have a bad drive.

[Frank1940]

Most vendors (either manufacturer or retailer) will give the customer that benefit of the doubt as long as the customer is not habitual  in making claims.  I have never really explained what the problem is on any HD that I have returned.  And, in each case, they have always shipped the replacement drive the same day as the RMA'ed one hit their Dock.  So they are not testing them at that point before shipping you the replacement.  I am certain that the manufacturers expect that a certain percentage of the HD that they get back don't have a problem and they accept that as simply purchasing customer good will.  (Remember that removing a HD from a system is not a trivial task for most people...)   

 

EDIT:  you might want to read this article about attribute 187:

 

    https://www.backblaze.com/blog/hard-drive-smart-stats/

 

[pwm]

5 minutes ago, Frank1940 said:

I am certain that the manufacturers expect that a certain percentage of the HD that they get back don't have a problem and they accept that as simply purchasing customer good will.

 

And that's the reason why replacement drives can be quite old and well used - the HDD manufacturers can't afford to throw all the returned drives so they give them a checkup and then sends them out again.

 

But the costs to handle RMA needs to be added to the sales price of new drives.

[Frank1940]

2 minutes ago, pwm said:

 

And that's the reason why replacement drives can be quite old and well used - the HDD manufacturers can't afford to throw all the returned drives so they give them a checkup and then sends them out again.

 

But the costs to handle RMA needs to be added to the sales price of new drives.

 

You do take that chance when you RMA a drive under warranty.  Especially, when the drive is been in production for a year or more.  It is also true for those drives with 5 year warranties which are manufactured discontinued after two years.   But I bet that they have test programs and methods that can test and condition drives in ways that we can't do.  After all, we very seldom ever get a drive with any defects even when we test them for period of 50 to 100 hours when we get them.  They 'know' how to find problem areas on a disk and 'hide' them so that we never see them.  I have no idea how long this process takes them but I would bet is it substantial.

 

Remember, we unRAID folks (and a lot of other server folks) need disks with zero errors on them.  Most folks would never realize for three years (minimum) if the entire inner 10% of their new drive was completely unreadable! 

[pwm]

Just now, Frank1940 said:

But I bet that they have test programs and methods that can test and condition drives in ways that we can't do.

 

I wouldn't buy into that bet. You don't take a returned drive and bring into the clean room and open it. The final testing will be performed with the drive itself as the test equipment. The only addition is that you can set the drive in test mode - sometimes by a strap field on the outside and sometimes by use of test points on the PCB. And the factory test can measure how much the drive vibrates and possibly the signal strength/integrity of the heads. But the actual surface testing is basically the same as you and me can do.

 

3 minutes ago, Frank1940 said:

Remember, we unRAID folks (and a lot of other server folks) need disks with zero errors on them.

 

Disk health isn't a digital state. So there isn't a 1=perfect and 0=broken. The manufacturers don't produce disks with zero errors on them, even if a small trickle might be produced that doesn't have any mapped sectors when leaving the factory. A drive that ends up with one remapped sector during the initial burn-in might be in way better state than 10 other drives that has zero remapped sectors after the burn-in. And it may have lots fewer remapped sectors - remember that you can't see how many sectors the factory test did remap.

 

The reason you see figures like 1 in 10^14 for nonrecoverable bit errors is because the function of the disks are very much related to statistics. There is a probability that the factory test will detect all bad sectors - but that probability isn't 100%. There is a probability that the data you write will land perfectly on the surface - but that probability isn't 100%. There is a probability that vibrations during the write will affect the write. There is a probability that noise on the supply voltage will affect the write.

 

All these probabilities are also a reason why you can find videos like this:

 

And the accompanying blog post:

http://dtrace.org/blogs/brendan/2008/12/31/unusual-disk-latency/

 

If you buy a new car and stalls the engine while driving from the dealer, you have to ask yourself - was it caused by a factory defect? Some particle in the fuel? User error? Divine intervention?

[Squid]

25 minutes ago, pwm said:

A drive that ends up with one remapped sector during the initial burn-in might be in way better state than 10 other drives that has zero remapped sectors after the burn-in.

I'm old enough to remember when MFM/RLL drives always had the bad sectors labelled on them, and the ones that didn't have any were always the worst ones to use as you could almost guarantee that additional blocks were get found by chkdsk (and a file corrupted in the process), but the ones that had a couple of sectors already labelled by the manufacturer tended to stay stable.