Increasing rate of hard drive failures

[mrcaffeinex]

I'm not trying to start a fight over what brand is better or which company outlasts the other, but I feel compelled to express my concerns over the quality of manufacturing in newer hard drives and I am interested in seeing if anyone else out there has had similar experiences.

I do a lot of repair work on desktop and notebook PCs. Over the past few months I have seen literally dozens of Western Digital and Seagate hard drives fail on customers. They exhibit high numbers of reallocated sectors, spin-up retry counts and raw read errors when viewing their SMART attributes. I have shipped three Western Digital drives out for warranty replacement so far this week alone and worked on a few more that were no longer covered under warranty.

Did the components change that much moving from the 200GB to 320GB 2.5" and 3.5" models that 500GB+ models are just displaying technological "growing pains" or do I just happen to be located in a Twilight Zone of bad luck? I still use several old Maxtor 200GB IDE drives and a Maxtor 250GB SATA drive that have been chugging along in high-use machines for years without exhibiting any of the aforementioned issues, although I have lost a few 500GB Western Digitals in the same time frame.

I have a basic understanding of statistics, so I realize that this is a small sample size. It does bother me though to see literally dozens of newer drives failing; in many cases less than a year from the date of purchase. Hopefully the warranty replacements will fare better...

-MrCaffeineX

 

[RebateMonger]

Apparently, "spin-up retry counts" are meaningless in SMART on Seagate drives.

Disk makers in general are having a hard time keeping uncorrectable error rates down with the latest high-density platters. Some disk makers seem to have fewer problems than others, but it's a concern all around. That's one of the reasons for moving to the new 4K sector disks which, theoretically, have much better error correction capability.

 

[FishAk]

A 1Tb drive is 5 times bigger than a 200Gb drive, giving it five times the likelihood of running into bad sectors as well. This is one of the main reason RAID 5 is not reliable on larger arrays.

 

[Rezident]

Every external drive I've bought in the last three years has failed (WD and Seagate), only one out of five internals has died over the last five years. I just find external HDDs to be rubbish tbh.

 

[pitz]

Haven't seen a 3.5" hard drive failure for quite a while now. From my vantage point, 3.5" 7200rpm drives are more reliable than ever. I have quite a number of 200gb Seagate's from 5-6 years ago that are pushing 50-60k hours on them, without issues (or anything even remotely suspicious in the SMART stats). And plenty of 320gb and 500gb Seagates that have 25k-30k hours on them now.

 

[mrblotto]

At my work, getting lots of failing SeaGate and Hitachi laptop drives (on Lenovo lappys) - mostly 160G 7200rpm. I've been seeing about 1 a day on average..........keeps me busy trying to retrieve data from them, especially since a lot have PGP installed...........sigh

 

[pjkenned]

From the unscientific NewEgg user ratings, I compiled 1TB v. 1.5TB v. 2TB results a little bit ago. 1TB's were higher rated than 2TB's by a significant margin... in an unscientific sample.

 

[Binky]

Haven't seen a 3.5" hard drive failure for quite a while now. From my vantage point, 3.5" 7200rpm drives are more reliable than ever. I have quite a number of 200gb Seagate's from 5-6 years ago that are pushing 50-60k hours on them, without issues (or anything even remotely suspicious in the SMART stats). And plenty of 320gb and 500gb Seagates that have 25k-30k hours on them now.

Your list of reliable hard drives (older and smaller and slower) is what he's comparing these newer drives to...

Saying that drives are "more reliable than ever" when you're talking about drives that are 4+ years old doesn't really work.

 

[Emulex]

i put my freeagent 2TB on a boob-matt (silicon gel). i was pissed at something and smacked the table and the freeagent clicked and went offline and then had a bad sector. it dawned on me how sensitive drives are to vibrations/noise/etc.

i get a ton of freeagent units back. bad power supply or using a real (short,more insulated) usb cable corrects 95% of the issue.

 

[jjmIII]

At my work, getting lots of failing SeaGate and...

I'm trying to coin Seabait....please work with me :awe:.

 

[jjmIII]

From the unscientific NewEgg user ratings, I compiled 1TB v. 1.5TB v. 2TB results a little bit ago. 1TB's were higher rated than 2TB's by a significant margin... in an unscientific sample.

cool :thumbsup:

 

[HendrixFan]

I've found most hard drive problems have to do with lack of adequate cooling. I personally experienced this back in 2003 or so when I was running my first RAID array. Since then my anecdotal evidence has been the same, anytime someone is having problems with a drive it has been fixed by improving the cooling.

The lower operating temperatures of the WD green drives are why I now have three of them.

 

[Modelworks]

I haven't had a hd failure in about 5 years. One big problem with drives currently is people moving them around when they are running. They see the portable label on external drives and think that means they can pick it up, toss it on a shelf or slide it around the desk while it is running.

The other thing is people moving their pc to get to cables, or something they drop behind it. That would be fine if the pc is off but doing it while the pc is running is a bad idea.

Same goes for laptops, they were not intended to toss around like a book while running.

 

[corkyg]

Echo Modelworks! I also have a suspicion that ever increasing larger capacities also have an effect, especially when folks seem to fill them up beyond a sensible 75 percent.

 

[Voo]

especially when folks seem to fill them up beyond a sensible 75 percent.

Huh? What should the amount of data on the drive have to do with the failure rate? Yeah the chance for a bit error is bigger, but for the HDD itself it's completely unimportant if you fill it up with 100% or 0% of useable data..

 

[corkyg]

HDDs need some freeboard in which to manuever things. Once you start filling beyond 3/4, they start to lose efficiency. You have many dynamic programs that need room to move things around. When they are constrained, it shows. The drive has to run longer to do simple things.

 

[pjkenned]

HDDs need some freeboard in which to manuever things. Once you start filling beyond 3/4, they start to lose efficiency. You have many dynamic programs that need room to move things around. When they are constrained, it shows. The drive has to run longer to do simple things.

Spindle disk = file storage, SSD = program/ local cache. No reason to be doing anything other than big sequential reads/ writes on a spindle disk at this point.

 

[Gooberlx2]

Seems like a good argument for error correcting filesystems like zfs and btrfs. These ever higher densities and disk sizes are only going to have increasing numbers of issues going forward, at least until disk technology (is SSD more reliable?) can catch up to the sizes. As far as data corruption goes, it's definitely something that could/should be watched for at the filesystem level.

Of course, a filesystem won't do anything about a dying drive, but hopefully it can help to maintain data integrity until it can be transferred elsewhere.

That reminds me that I need to transfer my data volume to an ext4 partition, in order for a more seamless upgrade to btrfs whenever it's ready for a production environment. But I gotta upgrade to SLES11 SP1 before I can do that..ugh.

 

[DrMrLordX]

Let's hope ZFS doesn't die a premature death due to patent trolling:

http://www.sun.com/lawsuit/zfs/

Otherwise, while I have had no problems with any IDE or SATA harddrive I've ever owned (7200.7, 7200.10, WD Black), I have and will continue to steer clear of any harddrive of 1.5TB in size or larger.

 

[Cerb]

Seems like a good argument for error correcting filesystems like zfs and btrfs. These ever higher densities and disk sizes are only going to have increasing numbers of issues going forward, at least until disk technology (is SSD more reliable?) can catch up to the sizes. As far as data corruption goes, it's definitely something that could/should be watched for at the filesystem level.

Don't you mean error checking? I was under the impression that no data recovery mechanisms were in either FS, for single-drive use. I would gladly sacrifice a few % of drive space, myself, for some parity data, to protect against data loss from growing bad sectors. One sector can go, and whatever that sector was used for just became void.

 

[RebateMonger]

Microsoft claims to have incorporated disk error correction into its in-development "Windows Home Server V2" that's currently out in Beta form. I haven't heard of any conflict with NetApp's patents.

 

[Voo]

HDDs need some freeboard in which to manuever things. Once you start filling beyond 3/4, they start to lose efficiency. You have many dynamic programs that need room to move things around. When they are constrained, it shows. The drive has to run longer to do simple things.

If at all, some programs need space to move stuff around (obviously), but that has nothing to do with the HDD itself. The only reason HDDs have spare space is for error detection and to reallocate corrupt sectors, but contrary to SSDs you don't have any problem to write a single sector of data, so how should that affect performance? Sure you get less performance on the inner tracks, but you've got that performance hit independend on how full the drive is, only where you're reading.

Any links for that claim?

 

[corkyg]

No links. Just pragmatic experience - optimizing a nearly full drive takes a very long time because the program has little free space in which to operate. The result is constant fragmentation. SSDs are exempt from this.

 

[Gooberlx2]

Don't you mean error checking? I was under the impression that no data recovery mechanisms were in either FS, for single-drive use. I would gladly sacrifice a few % of drive space, myself, for some parity data, to protect against data loss from growing bad sectors. One sector can go, and whatever that sector was used for just became void.

You're right. Corruption detection, and correction when there's redundancy in the pool.

 

[Voo]

No links. Just pragmatic experience - optimizing a nearly full drive takes a very long time because the program has little free space in which to operate. The result is constant fragmentation.

Yeah sure you get more fragmentation on full drives and obviously it takes longer to defragment them, but I just don't see how that has anything to do with reliability.