3 TB Hard Drive Problems

[Newtun]

From Ars:

Apple announces recall program for 3TB hard drives in 2012 and 2013 iMacs


Quoting from a comment there:

Rosyna (Ars Tribunus Militum)

retrospooty wrote:
Its odd about 3tb HDD's. They are failing in much higher #'s than 1,2 did before it and 4tb did after it. This is all major makers. Seagate, WD and the (formerly known as) IBM Deskstars as well.
​

I'm guessing this is due to the Great Thailand Flood of 2011. It resulted in a massive hard drive shortage. To make up for the shortage, HDD manufacturers (especially Seagate) appear to have neglected proper quality control and quality assurance, resulting in a rash of failures of HDDs made in 2012.

Backblaze has an awesome writeup about it. Over 90% of their 3TB HDDs from Seagate have failed since 2012.

 

[LunarMist]

It's an odd number of TB.

 

[Tannin]

Odd-size drives have always been a bit questionable. Samsung's 750GB drives were nothing like as good as their same-technology 500GB and 1TB drives, for example. I could think of several other cases if I put my mind to it.

What happens, in my view, is that with (say) a 2-head drive, you choose a platter with two good surfaces. With a 4-head drive, you choose two platters with good surfaces on both sides. So (if you are getting 500MB a platter for the sake of example) your 500Gb and 1TB drives are generally good. But what parts do you select for your 750GB drive? Naturally, you don't pick two perfect platters - you could use them to make a 1TB drive which sells for more. So you pick one good one plus one with flaws. The obvious flaws are only on one side of the platter, so there you have your 3-head 750GB drive. Trouble is, the failure rate of the good side of the flawed platter is higher than the failure rate of either side of the perfect one. Result: your 750GB drive has a higher failure rate. The exact same logic applies to 3TB drives (and various other odd sizes over the years).

A more sophisticated (and thse days probably more common) version of the same thing happens where the mid-size drive still uses (say) 4 heads like the full-capacity drive but each surface is recorded at a lower density for the reduced capacity model. Naturally, you select the best media for the 1TB unit (sticking with my 500GB/platter example), and the not-so-good media you use up in the 750GB unit. Same result.

Call this uninformed speculation if you wish, but if you do that you need to provide an alternative explanation for the increased failure rate of countless different odd-capacity drives from many different manufacturers over many, many years, which is not in doubt.

 

[mubs]

My 320 GB (odd-size?) Samsung drives have worked flawlessly. But I hear you.

 

[Tea]

http://redhill.net.au/d/43.php

 

[Buck]

Tony is correct when it comes to hard disk drives that have a higher prime capacity then what you have in hand. When Seagate released the 3TB drive related to the news article, 3TB was the prime capacity. They were manufactured with the prime format and all heads operating. A 3TB drive today would be a product with an alternate format and/or head(s) de-popped (as Tony described).

It should be noted that HDD manufacturers will produce two prime versions of a product. One version is manufactured with the maximum amount of platters (e.g. 5 for 3.5" drives) and another is manufactured with only 1 platter. From the outside they look the same, but you can definitely feel the difference in weight. This is done because lower capacity products still sell well (especially with OEMs) and a 1 platter drive costs less to manufacture. So, a 6TB and a 1TB could both be prime drives. The 1TB is just manufactured with one platter.

 

[Buck]

I am wondering why Backblaze would use desktop drives in Storage Pods. They should be using Datacenter drives. They will always see a higher failure rate when using a product in an environment for which it was not designed. Granted, a 90% failure rate is related to quality issues. But taking a drive designed to live for 3 years with less then 8 hours of use per day, and only moderate use at that, and then sticking it into storage pods is stupid. We've come across Backblaze having data traffic balance issues in their storage pods too - where a high amount of read/write cycles are disproportionately sent to only some drives, thus resulting a higher then normal failure rate.

 

[Mercutio]

Backblaze had some blog posts indicating that just after the Asian floods, it was seeking Cost/Capacity above all else, to the point of sending employees to purchase and de-shell external drives from discount retailers like Walmart and Costco. It wasn't buying enterprise drives because postdeluvian (ooh what a fun word!) pricing was out of line with any previously acceptable standard. I remember the $99 3TB Hitachis hitting $400 a few weeks after the flood, so it was entirely rational to buy Target's $125 3TB external drives instead.

Seagate 3TB drives are definitely less reliable than any other drive I am willing to install and tolerate, but my personal experience with them is still very different from what Backblaze is reporting. I've lost about 10% of my Barracuda 3TB drives per year since I started using them, where I'd call losing 5% of all drives "normal." I've also not lost one to a head crash; every failure I've had has been recoverable.

... And the last three WD Black drives I've wound up owning have all died in under a year, though the 3/4TB Reds at least don't seem to be a storage holocaust.

 

[Buck]

... And the last three WD Black drives I've wound up owning have all died in under a year, though the 3/4TB Reds at least don't seem to be a storage holocaust.

WD Black drives are a waste of time. I am sure you have the good sense to never own one again. Although I use WD drives quite often, that is just one model line that I find to be lame. A SSD + some 5400 rpm drive for storage is much, much better. The WD Red drives (6TB) have been good for me too. I do not have much experience with the Red Pro series.