Why do most hard drives fail?

[Nocturnal]

Is it luck? Is it heat? Is it a combination of both? Is it due to power surges? What normally causes a hard drive to fail?

 

[Sunner]

Depends on what you define as normally.
Heat can kill a harddrive, but I'm not sure that counts as normal.
Dropping it in the floor is a good way to kill it, but that's not really normal operating procedure either.

Generally, if you take care of them(no excessive heat, don't drop them, etc), the most common fault in my experience is the motor or heads going bad.

 

[MrDudeMan]

it is either physical or logical. if the head crashes into the platter or the electronics wear down, you have suffered an electromechanical loss. this is usually because of wear and tear or machining faults. remember the parts in there are super high precision, so it is possible to have a defective drive straight out of the box because of the difficulties in making the drives.

also, the surface of the hard drive deteriorates over time and all drives are susceptible to this. i work in an office with extremely high precision parts (we can move 400-500 pounds anywhere from 10 nanometers to 1 millimeter at a time) and i can tell you with full confidence that getting it right everytime is impossible.

 

[Jiggz]

Heat and most of the time it is "heat" again!

 

[fire400]

http://en.wikipedia.org/wiki/Failure_rate
-rate of failure

http://www.hddtemp.com/
-drive temp status

<a target=_blank class=ftalternatingbarlinklarge href="https://www.regnow.com/softsell/nph-softsell.cgi?item=3605-3&vreferrer=homepage">https://www.regnow.com/softsell/nph-......cgi?item=3605-3&vreferrer=homepage</a>
-claims to help predict HD failures

 

[programmer]

All drives have a "MTBF" rating -- Mean Time Between Failure. All hard drives will fail....eventually. Everything does.

In my experience, of the dozen or more hard drives I've owned over 20 or so years I've only had two truly fail before they were so "small" I couldn't really use them anymore (remember 10MB hard drives? 500MB? etc.). Both failures were physical components -- one drive was a real crash where you could hear the drive head scraping the disc platter. I believe the other was the arm that moves the drive head. I couldn't hear it seeking across the disk anymore, though the platters were spinning.

Its not at all a usual thing, unless like the others have mentioned, you drop the drive or soak it in high heat all the time.

 

[CalvinHobbes]

I think heat does it most often. I've seen some devices with very poor cooling that had HD dying left and right.

I wish more PC cases made HD cooling a priority. Lots of them have fans in front but too often they only cool some of the HD slots in the case. It's fine for one or two drives but if you have 6 HD's in a PC, they all need cooling.

 

[Eddieo]

Moving parts. Anything that moves will break more than non moving parts. Including on cars, printers, etc...

 

[deadseasquirrel]

I just finally lost my 27gb IBM Deskstar from '99. I couldn't believe it lasted as long as it did with pretty much everyday use over the last 7 years.

 

[Sunner]

Originally posted by: programmer
All drives have a "MTBF" rating -- Mean Time Between Failure. All hard drives will fail....eventually. Everything does.

In my experience, of the dozen or more hard drives I've owned over 20 or so years I've only had two truly fail before they were so "small" I couldn't really use them anymore (remember 10MB hard drives? 500MB? etc.). Both failures were physical components -- one drive was a real crash where you could hear the drive head scraping the disc platter. I believe the other was the arm that moves the drive head. I couldn't hear it seeking across the disk anymore, though the platters were spinning.

Its not at all a usual thing, unless like the others have mentioned, you drop the drive or soak it in high heat all the time.

I have a 20 MB external SCSI drive left somewhere...
Still using my trusty Quantum Fireball 2.5 GB'ers though, my firewall only needs about 130 MB of space, plus another 50 or so for logs

 

[Thor86]

1) Heat. Mechanical/electronic devices can only tolerate a certain level before failure.
2) Shock.
3) Electrical surges/brown outs.
4) Constant un-clean shutdowns/power outages.
5) Wear/Tear.

 

[drum]

Originally posted by: Eddieo
Moving parts. Anything that moves will break more than non moving parts. Including on cars, printers, etc...

exactly. its just a matter of time

 

[homercles337]

Originally posted by: Sunner
Depends on what you define as normally.
Heat can kill a harddrive, but I'm not sure that counts as normal.
Dropping it in the floor is a good way to kill it, but that's not really normal operating procedure either.

Generally, if you take care of them(no excessive heat, don't drop them, etc), the most common fault in my experience is the motor or heads going bad.

Actually, i just had a failure at home and was talking to the guy in charge of the 10th largest cluster/super computer in the world (MIT) and he said that he was currently working on getting > 50 drives replaced. This room is chilled to like 50 degrees, and every machine (blade farm) is cooled. Seeing it gave me goose bumps, and not because of the temp--the sheer computing power is just WOW!

 

[corkyg]

And - actually MOST hard drives do not fail! Maybe the question should be, "what causes most harddrive failures?"

I got my first HDD in 1985 - a "hard card" - 10 MB. Within a year,I had my first Seagate ST-251 - about 20 MB. They were famous for "stiction" failures. But - that never happened to me. Since then I have had probably three dozen hard drives, and out of that pile, only one failure!

All of the things in the above notes can and do contribute to HDD failure. But again, most hard drives do not fail.

 

[Tostada]

I think all you guys claiming heat is a primary killer of hard drives are wrong. There are tons of people keeping tons of hard drives in hot cases with totally inadequate cooling, and you just don't hear about it killing hard drives. The only heat-related drive failure I can think of is all those old stories of 1st-gen Cheetah hotswap cages bursting into flames from overheating ... and I'm going to go out on a limb and say those incidents have been exaggerated.

For 5 years I had three Quantum 10K drives and two IBM 10K drives all stacked up in a case with no venitilation across the drives and shut in a desk drawer. You could hear the drives constantly doing their little ticking thermal recalibration thing. None of them ever died or got any bad sectors. I've also had Maxtor and Hitachi drives die in a month when they were running as a single drive in a well-cooled system. Obviously Quantum/Maxtor and IBM/Hitachi had higher quality 10K SCSI drives than the more recent IDE drives products I'm talking about dying on me, but I've still never seen a drive die which I would even begin to suspect had anything to do with heat, and I've seen a lot of drives die.

I've had multiple drives from every brand die on me except Fujitsu and Samsung, but that's just because I haven't gotten many of them.

I think the most common ways for a drive to die are:

1) the drive gets knocked around and causes a head crash, either killing it right then or giving it tons of bad sectors
2) the click of death, where the head keeps seeking and banging against the stop. I've heard people say this is usually because of the sensor dying which tells the drive when the head has reached the first track and lets it calibrate itself, but I haven't really seen any fully credible explanation.
3) the reliability of the drive starts to degrade, bad sectors start slowly popping up, sometimes files can't be read, sometimes Windows gives you the "delayed write failed," etc, until the drive totally dies or you just stop using it because it's unreliable. This could be because of manufacturing defects, poor tolerances, defective materials, excessive heat, etc. Most of the time this can be considered dying of natural causes.
4) ESD or possibly heat or even just age damages the electonics. Usually the drive spins down and never spins up again. Sometimes the drive just gets totally flaky. I've seen the BIOS report a scrambled ID string instead of the actual name during POST half the time the machine boots up when the drive is close to dying.