Can a Regular Person Repair a Damaged Hard Drive?

MrSeb writes "There's a lot of FUD when it comes to self-repairing a broken hard drive. Does sticking it in the freezer help? The oven? Hitting it with a hammer? Does replacing the PCB actually work? Can you take the platters out and put them in another drive? And failing all that, if you have to send the dead drive off to a professional data recovery company, how much does it cost — and what's their chance of success, anyway? They're notoriously bad at obfuscating their prices, until you contact them directly. This article tries to answer these questions and strip away the FUD." What has been your experience with trying to fix broken drives?

Comments

One word

[mknewman]

No.

Re:One word

[DarrylM]

One word: Yes.

Longer version: But it may be more difficult to do nowadays; I don't know. About 7 years ago a family member had a computer with a lot of photos that were, sadly, not backed up. The Maxtor drive had suddenly quit. I was able to eBay another drive with the same model number and swap the boards, and voila! We had a working drive with all of the photos (and other data) intact.

Again, I have no idea how easy that would be to do nowadays... It was hard enough to change boards with my clumsy fingers on a 3.5" drive, let alone a mobile drive.

Re:One word

[CastrTroy]

I repaired a drive once by overwriting the entire drive with zeroes, and overwrote the whole thing from /dev/random. After that I repartioned it and it worked fine for another 4-5 years. Before I "fixed" it, it was reporting bad sectors all over the place, and constantly had read and write errors. I salvaged what I could, but wasn't able to recover much. I never really trusted it with important data after that point, but it also never failed me after that. I eventually just stopped using it when I purchased a new hard drive, it realized the old one didn't have enough space to be useful. It was only 12 GB. Most USB sticks are bigger than that these days.

Re:One word

[SuricouRaven]

The old write-over trick. Yes, what you're doing is actually forcing the drive to remap bad sectors. How reliably it works after depends on what caused the bad sectors in the first place.

Re:One word

[t4ng*]

As someone that worked as an engineer in the hard drive manufacturing industry for 15 years I would have to agree, "No."

You might be able to revive a drive if it is a problem with a PCB, but if it is a problem with the disks or heads, forget about it!

Incidentally, a "hard drive crash" used to mean a head touched the disk and physically damaged the head and/or the disk. But for nearly two decades now, heads in hard drives are "contact heads," meaning the smallest part of the gap between the head and the disk is smaller than the mean free path of air molecules. However the heads are "flying" at a fairly high angle of attack, so it is really only the trailing edge of the head that is in contact with the disk at all times. Between that contact head design and auto retracting armatures that pull the heads off the data area of the disks, actual head crashes are extremely rare under normal operating conditions.

Re:One word

[1u3hr]

I repaired a drive once by overwriting the entire drive with zeroes,

TFA is about a physically damaged drive. (Burnt out component on PCB.) The aim is to recover the data from that.

Your method won't work on that kind of failure, and certainly won't recover any data.

Re:One word

[ckedge]

Yes, this frequently works at making failing disks "work again" -- as the manufacturer test sequences and/or simply zeroing the drive gives the drive a chance to find and mark all the bad failed blocks as bad, and the remaining blocks are all the ones that didn't fail and so the disk keeps working for a few more years. I've used this a half dozen times at work.

Of course, this is to make a failing disk "work again", it doesn't help with recovering existing data.

The first thing I'll try with a failing disk is to setup a file by file mirroring program (robocopy is one cli program I use a lot) and set it's "retries" to a moderately high number, like 5 or 10 or 20. Even though you are getting read errors, there are a class of problems where occasionally the read will work, and so each time you try and "rsync" the disk, you get more and more of what's there, till you have a mostly complete copy of the data. This is the same method that some enthusiast utilities use (like grc's disk recovery program, iirc).

I've personally used the freezer trick once. Because of the possibility of condensation, I used the fridge first. I don't recall if I had to use the freezer, but I know I would not have left it in the freezer long (metal transfers heat fast, so it doesn't need to be deep freezed, just a bit colder than the fridge), maybe 5 minutes max, and I recall thinking that I'd end up putting it inside an anti static bag or something with an elastic closing the bag on the cables ... so that the amount of condensation would be limited, either that or run the dehumidifier and/or AC really hard first so that my apt was at low humidity. Definitely would not try it in the middle of a humid summer. Better to wait till winter and turn up your heating system and open your windows so the humidity drops really really low. That's always another option (for those of us that live far enough north), take the system into the chilly cold arid garage so the freezer trick doesn't result in lots of condensation.

Of 5 drives that were failing, 3 I recovered by "retry reads over and over", and 1 I recovered using the freezer trick.

I have one more left that I need to try a "platter swap" with an identical working model number using the "bathroom cleanroom technique". But I'm not looking forward to that, getting the platters out without scratching them on the heads is going to be a massive bitch. I think I'll practice on a few old 9GB drives before I try it with my failed 120GB drive. (I've had it sitting around for forever waiting for me to find the time to do it, I don't actually still use drives that old.)

Re:One word

[Briareos]

Betteridge's Law of Headlines applies again... :)

Re:One word

[t4ng*]

I can see how that might work if it were a pretty old hard drive. Since hard drives have magnetic recording media, you can't just write raw data straight to disk. For example, if you truly wrote all zeros or all ones (the recording bias all in one direction or another) there would be no way to figure out if it were all zeros or all ones or how many bits of zeros or ones that you had recorded. So all data is encoded before writing it to disk to ensure that there is always an alternating magnetic field on the disk. A zero might be expanded to 1001, or something like, that before it is written to disk. Different encoding techniques have certain known data pattern weaknesses, data patterns that when encoded will produce a more difficult to read signal on the disk than other patterns. These bad data patterns are used to test drive designs. Additionally, each data track on a disk is sandwiched between two servo tracks. These help keep the head centered on the data track no matter where it is without having to worry about drive calibration. And finally, drives include a lot of spare sectors that the drive electronics are supposed to automatically swap out, without the OS knowing about it, when bad sectors are detected.

So, it is possible that you had a drive that after a lot of writes and rewrites was having some signal-to-noise ratio problems detecting data written on the drive. Your rewrite operations may have normalized the media on the disks just enough to get a little more life out of it. But what was actually being written to the disk wasn't all zeros. If that were possible, you would really make the drive unusable!

Re:One word

[blackicye]

One word: Yes.

Longer version: But it may be more difficult to do nowadays; I don't know. About 7 years ago a family member had a computer with a lot of photos that were, sadly, not backed up. The Maxtor drive had suddenly quit. I was able to eBay another drive with the same model number and swap the boards, and voila! We had a working drive with all of the photos (and other data) intact.

Again, I have no idea how easy that would be to do nowadays... It was hard enough to change boards with my clumsy fingers on a 3.5" drive, let alone a mobile drive.

This will not work with many newer drives, especially WD Caviar Black and Blue. There is a firmware chip on the PCB that also needs to be transplanted, and this is tricky even with a Surface mount electronics soldering station, the type that uses channeled hot air.

Re:One word

[drinkypoo]

I've personally used the freezer trick once. Because of the possibility of condensation, I used the fridge first.

There is little possibility of harmful condensation if use a limited version of the freezer trick, and simply suck as much air out of the bag as possible (use a vacuum sealer if available) and then let the disk return to room temperature before opening the bag. You don't NEED to run the disk while you thermally cycle it if the problem is stiction. And there is also always the option of using one of the many waterproof enclosures available on the market, and simply slipping some dessicant packets in there before you start your quest. You could also just put it in a tupperware and seal the hole the cable passes through with silicone, and put the dessicant in there with it.

Re:One word

[blackicye]

No.

Any home user who has a Scanning Electron Microscope and the appropriate algorithms in their basement can recover data from almost 90% of mechanical failures.

Re:One word

[Voyager529]

Incidentally, a "hard drive crash" used to mean a head touched the disk and physically damaged the head and/or the disk.

...and today, "my hard drive crashed" can mean basically anything. I can't count how many times I've heard someone say "my hard drive crashed" only to have their response to my question of "how do you know" sound something like, "well, when I go to 'my documents, Word doesn't load" or some other similar error where the hard disk is clearly not to blame. I tried educating people on the actual meaning of the term, but it seemed a losing battle, so now "my hard drive crashed" generally translates to "something is broken", unless it's a fellow computer tech who I'm certain knows what an actual hard disk crash looks like.

Re:One word

[Anonymous Coward]

About two years ago I had a 1TB Hitachi drive go bad on me, and a few minutes with a multimeter told me the board was the culprit. The drive was still under warranty, but they would have replaced it with a new drive instead of trying to repair the old one and 1TB is a lot of data to lose. Fortunately I found the same model on eBay with a bad head for a few dollars.

Simply swapping the boards didn't work for me, so I hunted down the location of the NVRAM and swapped those with a rework station I've owned for a while. After that it worked just fine, and I'm still using it today as a backup for the new drive I received from Hitachi after RMA'ing what was left of the other.

In the end I saved my data and doubled my storage space for less than $25.

Re:One word

[Sancho]

Real data recovery places are too expensive for the average person to use. I think there's a middle-ground between your two extremes.

Re:One word

[danceswithtrees]

You mean like "my program has a bug?" You probably don't mean there is moth interfering with a relay closing.
Certain phrases come to encompass more than their original meaning.

Re:One word

[Mathieu Lu]

As the GP, I once had a burnt component on a PCB of a hard disk. I changed it using the *exact* model of the disk, and it worked. The disk was 100$ for a 500GB off eBay, which is a bit expensive, but afterwards I had a brand new disk to keep (I put the PCB back on the new disk once I finished retrieving the data). The seller on eBay provided the complete serial number in order to make it easy to find the correct replacement disk.

There are also companies in Hong Kong that specialize in selling replacement PCBs. It's much cheaper, but bigger delays.

Re:One word

[thegarbz]

Condensation won't kill electronics. The water from condensation is pure and free of contaminants. The electronics will quite comfortably run with beads of water on them.

This is a common trick in electrical equipment fault finding. Any fault that is caused by thermal expansion or contraction can be located on a giant PCB by using a localised blast from a freeze spray. The end result is everywhere the freeze spray has been used frosts over then gets wet. I've yet to kill electronics using this method and have been through about 10 cans in my life.

I used the freezer trick to recover a HDD which died of neglect (unused drive for about a year), I assumed the motors just suffered stiction. Out of the freezer and straight into service. Worked beautifully even with condensation running down the side of the PCB. After about 5 minutes it all dried out.

Re:One word

[hairyfeet]

I had to do that trick once as well and I'm glad to say it worked. Man you want to talk about a tense as hell experience, there I was with a gal just crying her eyes out because that drive was the ONLY source of a couple of hundred pics and little video and audio clips of her dead mother.

But I did just as you described, vacu-sealed the drive and let it come back to room temp before opening the bag and it worked just long enough to get that data off. Luckily that one close call was enough for her so I was able to set her up a backup routine using a combination of DVDs plus Cloud plus USB HDDs that are rotated out weekly between her home and that of her sister. that way everything she has is backed up thrice on 3 different mediums at 3 different locations just to be safe.

Please folks, i don't care if its flash sticks, get your friends and family to back up the important stuff, okay? Make digital copies of all pictures and documents and store them in multiple locations. I had a customer come home from a 3 day business trip to a smoldering pile where his house once was so I know how important they can be. Thankfully he had listened to me and all it took was retrieving his backup drive from his brother's house and he had all his records and pictures back in a few hours but he could have lost everything, all the photos of his family, all his childhood videos, the whole thing just gone.

So backup backup backup folks, and make sure everyone you know does the same. Hell you can get a 32Gb flash stick for less than $20 off of Newegg and that will hold most people's precious stuff right there and fit into your pocket or a safety deposit box. Better to have it and not need it than need it and not have it, as I've seen time and time again here at the shop.

Re:One word

[Mathieu Lu]

Here are two PCB sellers in HK who ship overseas:
http://www.onepcbsolution.com/
http://www.hkmingdi.com/enindex.asp

I found them off the forum of this site:
http://www.deadharddrive.com/

I wrote a short post about it in French, you can probably run it in google-translate. It took me a bit of time to figure out the PCB number on my Seagate drive, which is on the PCB, but on the side facing the disk, so I had to unscrew it to obtain it. (both HK sites were helpful and responded to my e-mails in good English).
http://www.bidon.ca/fr/random/2011-04-12-disque-dur-ressuscite

Re:One word

[Solandri]

There is a firmware chip on the PCB that also needs to be transplanted, and this is tricky even with a Surface mount electronics soldering station

Just to be clear, the HDD manufacturers didn't stick it on there just to make it harder for users to fix their own hard drives. It contains mappings of bad sectors which the drive swapped out with reserve sectors through the normal course of operation. Older drives had smaller capacity so proportionately fewer bad sectors and could get away with just mapping them out (reducing capacity) or storing the mapping on the drive itself. The high capacity of modern drives makes it a virtual certainty that it's going to develop multiple bad sectors through its usage life. So you need a more systematic and reliable method of dealing with them. The norm is to set aside some reserve space, and when the drive detects a sector going bad, map it out and replace it with a sector in the reserve space, and note the new mapping in the nonvolatile memory of the firmware chip.

Re:One word

[ArsenneLupin]

For fuck's sake, am I the only one who knows how hard drives work?

Actually, it looks more like you are the only one who doesn't know how hard disks work.

Even if a drive were stuffed full of air (pretty sure they're a vacuum),

Actually, the drive does contain air. The HDD's spindle system relies on air pressure inside the disk enclosure to support the heads at their proper flying height while the disk rotates.

it's not high humidity air

Actually, drives usually have tiny "breather" holes to allow air through for pressure equalization if ambient air pressure changes. These also let through along any humidity that is in the air. The only thing that they are designed to hold back is tiny dust particles, which might otherwise cause a head crash. The environment within a hard drive is merely dust free but not a vacuum.

Re:One word

[Cramer]

This is what we used to call a "low level format". It's next to impossible to do outside the factory these days. For one, because the firmware is stored on the platter(s) -- flash is expensive, so they save $0.03 by not putting any on the board. (the board has the tiniest ROM possible to hold a boot loader.) The FORMAT UNIT command just zero's the drive; it does not actually reformat the surface, which rewrites tracking information, parameter block(s), and padding not normally user accessable. ('tho there are vendor specific commands to get at that. similar to a floppy disk "full track read".)

[I've had a few Maxtor drives (bulk) that weren't programmed. The boot rom has just enough brain to show up on the bus. Maxtor will not give you the necessary tool to fix this -- i.e. there is no firmware upgrade tool available outside the factory. Other manufacturers are not as lame.]

It Depends

[Anonymous Coward]

Sometimes just the controller portion fails. If you remove it and replace it with a working one from a identical drive you're back in business. Only tool needed is a torx driver I believe.

Re:It Depends

[OAB_X]

The sysadmin where I work once bought two Seagate drives, they had sequential serial numbers. They failed within 5 days of each other (killing the RAID-1 because the array had not finished rebuilding).

Freezer "fix"

[Georules]

Don't ask me how, but I had a failing drive that couldn't even manage to be on for 30 seconds before being unreadable. Since I was curious, as a control, I first let the drive sit at room temperature for an hour. Afterwards, again, only 30 seconds of read time. I then put it in the freezer for an hour, and was able to read for 10 minutes, just enough time for the data I needed. I have no idea what actually happened, and am still skeptical to attribute the success to the freezer, but I did get what I wanted.

Re:Freezer "fix"

[toygeek]

Something on the PCB was cracked. The freezing caused everything to pull back together, and heat separated it. So, bringing it to a lower temperature kept it together longer. Simple enough. Is it a repair? No. Its a workaround. A temporary one.

PCB swap is cheap, quick, and often works

[thatseattleguy]

In the very limited (3) cases that I've had to try and revive a client's dead desktop drive, replacing the PCB board from an identical model - usually purchased cheaply, used or new, online - has always worked.

The other advantage of this approach is that if the first drive becomes revivable, even a time, you now have a second same-capacity drive to transfer the data to (using intermediate storage media if in fact it was the PCB that was the problem and you can only get one drive working at a time).

If it doesn't work, you're no worse off and still have a replacement drive to load data from your (hopefully recent) backups.

AU $2600 to repair...

[HeadlessNotAHorseman]

I broke an external USB hard disk once (it tipped over while running). It cost me AUD $2600 to get it repaired. They got most of the data off; some was corrupted but fortunately nothing important. I take more regular backups now!

my experience

[Anonymous Coward]

Had a disk at work with our sourcesafe database on break. Due to responsibility falling between chairs, there was no backup at all. Sent it to one rescue firm, came back without successful restore, sent it to another one, got more than 99% back, lost nothing important, cost somewhere in the low 4 figures.

With private disks where data rescue is out of the question, I've had good experiences with freezing and in other cases replacing the circuit board. If doing it yourself, always mount RO and have somewhere with enough with enough space to make first a "cp" of selected really important stuff, a recursive "cp" of everything, and last a "dd" or "rescue_dd" of the whole disk. I've had better luck copying files from within a read-only mounted filesystem at first, you are fighting the clock after all.

Re:For the 57th time on Slashdot

[txoof]

I've tried the freezer trick to help what sounded like an ailing bearing , but with limited success. No amount of freezing seemed to help. To make things worse, when I took the drive out of the freezer, moisture started condensing immediately on the cold PCB. I tried to place it on a sponge to help sop up the water, but I can't imagine this helped the drive at all.

I have some friends that swear by this, but I am extremely doubtful especially because of the condensation problem. I feel like this is an a apocryphal bit of "knowledge" that has been passed down from a time when drivers were larger, slower and had less precise bearings. I can imagine that on a big old drive freezing the drive *may* have helped. But then again, perhaps it's something like throwing a pinch of spilled salt over your shoulder or touching wood--something your grandma told you to do, but doesn't actually do anything.

It's obvious

[Anonymous Coward]

Just 3D print whatever new part you need. A new read/write head? Just pop some plastic in the 3D printer and print one out. Then head over to the clean room and the tool box and jigs and use your dexterity and skill to change the head. Bad IC somewhere? 3D print out a new chip. Yes, 3D printing is the future!!

Re:It's obvious

[Georules]

Broken 3d printer? Print a new one!

One trick not listed

[Anonymous Coward]

I took my hard drive to the Geek Squad and they wanted $500 to send my hard drive away to get the data.

I yelled at them and I told them that was robbery. Asked for the manager. But, when I was leaving one of the Geeks told me a secret.
He said just go home and drill a hole in the hard drive and then set it on top of your new hard drive with the hole facing down. All the data will just pour out to your new drive.

It didn't work for me, but maybe I didn't do it right?

Re:One trick not listed

[SuricouRaven]

Probably used the old 32mm hole. Newer computers are built using 64mm bit technology, you need to use a 64-bit to compensate. It's a very delicate operation though, as 64mm is a very big hole and you need to be careful to keep it perfectly round, else the data will be unevenly distributed.

Re:One trick not listed

[Waffle Iron]

Did you remember that you also need to drill a hole on the top of the new drive?

Otherwise, the data will just spill out onto your desk.

It depends

[woboyle]

Some HD problems (stuck platters so it doesn't spin up) are user-fixable. Most are not. There is a syndrome called "sticktion" where the read/write heads settle on the platters when shut off (most modern drives will elevate the heads when shut off, but some, including many older drives, do not). Because the platters and heads are so flat, they mechanically weld themselves together over time. To fix this (a technique I have used often in the past), you need to remove the drive, and then snap rotate it on the plane of the platters, so that the momentum of the platters trying to counter rotate against the impetus of the rotational momentum you are applying to the drive will break the "weld" loose. If you then quickly re-install the drive and turn it on, it will most likely spin up and continue to operate without problems. Other failure modes include head "crashes", spindle bearing failures, drive motor failures, controller circuitry problems (bad electronic components), and mechanical breakage of connectors, solder joints, etc. These typically are not user repairable.

won't work

[Anonymous Coward]

Maybe the PCB swap used to work, it almost certainly won't work anymore. When a HDD powers on, it needs to load some parameters for the servo system (i.e. positioning the arm) and other tuned parameters for the controller to read back off the disk. These parameters are probably stored in flash memory on the PCB and the parameters will vary from disk to disk. So, parameters for drive A will not work to spin up drive B because of small variances in their manufacturing even if they're made on the same day in the same plant on the same line by the same underpaid employee

You can't swap disks because even if you get a tiny fingerprint on the disk, it's the size of Mt. Everest compared to the distance between the read head and the media. You'll be putting your own home-grown media defects all over it. Forget about getting your files back.

Aside from common firmware related problems (search for "reparing 7200.11" in google for an example), you're not going to have much luck.

The only other thing I've seen work: a guy took his neighbors HDD (which was not responding in Windows) and had to use an oscilloscope to realize the read waveform from the read head was a low amplitude. He built a small in-line amplifier which brought the amplitude back up to spec so the data could be read off. I was impressed.

Source: I have work experience on manufacturing processes for HDDs.

Yes, maybe.

[Fished]

My first job in "the industry" was in a PC repair shop in 1991. Back in those days, we had a huge crop of bad Seagate 40MB (yes, that's "mega" children) hard drives. The usual problem was that the spindle had frozen up, and if we took the circuit board off and gently tapped the spindle, you could often (about 75% of the time) get the drive to start spinning again long enough to get your data off.

Hard drives have gotten a lot more reliable and a lot smaller since then. I don't know whether this would be a wise thing to do with a modern hard drive.

A Better Word

[epp_b]

Backup

Re:A Better Word

[hazem]

Backup seems to be one of those things people don't learn until they've learned it the hard way.

After my unrecoverable hard-drive crash, I set up a NAS with RAID that my computer backs up to every night (incrementals daily and full every 7 days). I then backup that NAS every couple months to an external USB drive that I store in another building.

The key there, though, is it go and try to do a restore from your backup every now and then to make sure you can. There are few things worse than feeling secure in your backup process only to find it was not running correctly all along - or that you can't actually restore all your data from the process after all.

Platters no way

[ArchieBunker]

You will have great difficulty taking the platters out. The read heads have to be removed without physically coming into contact with the platters. You'll need specialized fixtures and tooling to even begin. If the data is that important then send it to a professional.

Been there, done that, time wasted..

[brokenin2]

We fixed a drive by trading the pcb with another *IDENTICAL* drive (same rev of board etc)..

The funny part was that when we went to recover the files they desperately needed back from that drive, all we found were shortcuts to a network drive, where the files had been safe and sound the entire time.. The user just had no idea that they hadn't lost their files..

moving serial memory chips

[petermgreen]

Sometimes it's possible to ressurect a dead drive by swapping the controller board with another from the same model (or a very close model from the same range). Unfortunately with modern drives there is often information stored on the controller board which is needed for the drive to start. This information seems to be stored on a serial memory chip (usually an 8-pin device in a SOIC or similar package) on the controller board.

What i've found you can do is remove the serial memory chip from the dead controller board and solder it to the donor controller board. Provided you have a hot air rework station it's pretty easy to remove and re-fit the serial memory chips. So-far i've tried this twice and it's worked both times, YMMV of course.

Re:For the 57th time on Slashdot

[petermgreen]

If you do that you want to put a dessicant in the bag with the drive. Otherwise you are just sealing the humid air in.

Re:It depends - Sticktion Y2K Repair

[InitZero]

In preparation for Y2K, we had to turn off our text archive server (at a newspaper) for the first time in, literally, years. The machine itself has been in production for six years, the last two or so of which without a reboot.

It was an IBM AIX machine with an array of 4.5GB SCSI drives. After sitting with its power off for a couple hours, we turned it back on and Nothing Happened. No drives were spinning. Crap.

Called IBM tech support. Got the run-around. Finally got to a guy who said something along the lines of "you're going to think this is crazy but do what I say in this order" followed by...

* turn machine off
* remove drives
* turn the machine on
* bang the drives on their edge a few times on the floor - don't go crazy but harder than you think is a good idea
* spin the drives flat on the ground as though they were tops
* immediately, put the drives in the enclosure
* reboot the machine but do not power it off

Damn if the guy wasn't right.

His guess was that the drives had been powered for eight or so years and the lubricant had either broken down or the heads were simply stuck to the platters. The thumping dislodged the heads and the spin gave the grease a fighting chance. {shrug}

In any case, we dared not turn it off for another year and a half until at such time it was replaced. We thought about buying replacement drives but IBM wanted something along the lines of $600 for a 4.5GB drive. Even on eBay, they were three times what we felt was reasonable.

Cheers,
Matt

If you're crazy enough to try the freezer trick...

[IonOtter]

At least do it the right way.

1. Get ahold of long cables that can reach *outside* the freezer to your machine.

2. Wrap the drive *before* you put it in the freezer. Heat a towel in the oven to make sure it's dry, then wrap the drive in the towel. Now stick it in a plastic baggie, along with some silica gel packs to suck up more moisture. Try to close the mouth of the baggie around the cables as much as possible. Use duct tape if necessary.

3. Put it in the freezer, route the cables through the door seal, and make damn sure the door is shut tight as possible. Seal it with more duct tape if you have to. Let it sit in the freezer for at least 6 hours to get really cold.

4. Make all your preparations before plugging in the drive. Situate your primary machine right next to the freezer, make sure you're ready to go. If you can somehow manage it, and you know what you're doing, boot into an old copy of DOS, or a command-line interface of your preferred *nix distro. Don't waste time loading Windows if you can help it.

5. Turn off your machine, plug the drive in, then reboot.

6. Move *fast*. Start copying the drive contents over to the backup drive as fast as you can. If you can do it via a script or batch file, then even better. Speed is of the essence. In fact, if you know the locations of the files you need, as well as their general file names, then creating a batch file BEFORE starting would be your best option. Just tell it to copy everything in C:\MyLifesWork\coldfusion*.*

7. MOST IMPORTANT STEP!!! If this does not work, and you can't pull anything off the drive, then don't panic just yet. Turn off your machine, unplug the drive, then unplug the freezer.

Do NOT open the freezer until it has reached ambient temperature, which will take at least 24 hours or more.

This will prevent the drive from getting roached from the condensation, and make it more feasible for a drive recovery company to save your data.

Freezer trick...

[bayankaran]

Few years back a 20GB drive I salvaged from an old dead Thinkpad stopped working. No whirring sound, nothing...the green light on the USB enclosure stayed on.

There was no important data, but I thought "this is the chance to learn how to salvage a hard drive".

I did the freezer option. I had already used the freezer to kill ants in sugar and bugs in rice. Froze the drive overnight, took it out and immediately connected and waited for whirring sound. No sound. The drive is dead.

Gave the drive couple of almighty whacks. Still no sound. No life at all.

I threw it in the dust bin.

The next day I tried to connect a camera. The SD card on the camera failed to be identified on Windoze and Linux.

I tried another USB cable. And the camera connected fine.

It took me a few seconds to remember the old hard drive. Took it out of the trash, wiped it clean and connected.

The drive works perfectly fine even today. But it still got the smell of decomposed tea leaves.

Correct

[Sycraft-fu]

It really depends on what has failed and how. I've repaired a number of drives at work well enough to get data off with just basic software tools, like using Knoppix to force mount problematic partitions and so on. The drive may be failing but not completely so a software-only solution can do the trick.

Also I've had Spinrite work. It has about a 40% success rate but on drives that nothing else could read, I've had it make them readable again. In one case I ran Spinrite (it takes many hours, put a fan on the disk), copied the data to a new disk with Ghost, did a chkdsk, did a repair install of Windows and the system functioned flawlessly, no data or app loss. Of course the other 60% of the time it destroys the disk beyond any repair so it is a "Use only as a last resort and only if the data isn't important enough to pay for professional recovery," tool.

Replacing controllers can work if the controller is what has failed. Needs to be the precise controller so one from a like disk but different size won't work and occasionally even the firmware version can matter.

However if the problem is with the heads themselves or the platters then no, you can't do shit. You need a clean room to open the drive up without destroying it, and then of course you need something to put the platter in to for reading them.

So you can try to self repair a drive. As I said using recovery software (Knoppix with force mount is a great thing to try first) is a good first step, so long as the BIOS can see the drive. May be that you can just copy the data and call it good. However there are also plenty of situations where you can't repair it so don't count on it working. If the data is really important, send it to a pro.

My experiences

[drinkypoo]

WAY BACK WHEN I took the cover off a 40MB RLL disk, pushed on the spindle by hand (it had so much stiction it could not be repaired by any other means) to free it, put the cover back on, and it worked. Note, 40 megabytes. I didn't even try to improvise a clean room. My cleaning procedure was to blow on the top platter gently before I closed the drive. The drive spun up and I was able to recover 100% of the data from it, and it was nearly full. Before this happened this disk actually burned a power-carrying trace off the board and I replaced it with a wire jumper. Then later that wire got so hot that the solder melted and it fell off, and I put it back on and used it some more. You guessed it, Seizegate.

MANY TIMES I have got a non-spinning drive spinning again by whacking one corner (from the side of the drive) with a screwdriver. The last one I did this with was 80GB or so, but there's no reason why this technique should not be valid today. Connect to power, give it a sharp rap in the appropriate direction, listen to it spin. I started doing this with ST-225s which needed it very often, but I've applied it to many different disks successfully over the years.

My experiences aside, many people have put disks in the freezer or even the oven (not hard to stay below reflow temperature) and got them to free up. If it's a stiction problem it's all about thermal expansion and contraction. If the drive spins but does not work, if you're very lucky you might have a PCB problem, and if you can find a disk of the same model and version then the PCB from the other disk might work on your disk.

Re:I've never met...

[Antarius]

You need to hang around with people who eat more insoluble fibre.

you must be new here

[anarcat]

Regular person? This is slashdot, there are no "regular persons" here.

Removed shorted diode, fixed!

[neurocutie]

My boss handed me the hard drive from his laptop. He said it was totally fried. It had ALL his work on it, no backups.

What had happened was that he had some minor NTFS corruption problem, so he went to our IT dept. Some IT monkey removed the laptop drive and tried to hook it up to a SATA - IDE converter. However he managed to wire up the power backwards. That fried the drive, but actually all it really did was burn/short the power polarity protection diode.

So with magnifying glass and soldering iron, I simply removed the shorted diode, and voila (not wahlah or viola), the drive was working again. I was then able to easily clean up the NTFS problem. Boy was he happy to get all his stuff back.

Re:I've never met...

[erdos-bacon sandwich]

You need to hang around with people who eat more insoluble fibre.

They typically don't have to worry about being backed-up

More detail, please, neighbors!

[Anonymous Coward]

Please put dates on your war-story recovery experiences. If you don't remember, give the drive's size.

Please don't be pointlessly coy about naming the recovery vendor who could or couldn't recover your data, the price they quoted, the price you finally paid them. Don't say nothing. Don't say "a certain major vendor, har har." Just tell us who they are. You don't have long before these sketchy tradies see the article and flood the comments with self-serving FUD.

Please state what was wrong with the drive before you tried to recover it, ex:

  * didn't spin

  * didn't identify

  * latent sector errors (ex. "OS keept crashing")

None of the upvoted comments have these details which is making the article useless for me.

For my own experience,

(a) I had full, easy success swapping the controller of an IBM 9GB SCSI drive, but the old controller had been visibly damaged. SMD gull-wing (?) pins were smashed. This is a rare failure case for me.

(b) I'm often able to recover data from the third category of drive with 'dd if=/dev/olddrive of=/dev/newdrive bs=512 conv=noerror,sync'. (there is also dd_rescue, but I don't use it). This works on Linux or Mac, but on Solaris you must use GNU dd for it to work---the included one is just broken. Obviously you need to boot off a drive other than the one that's failing. The drive must be unmounted when you use dd. It takes several days to read a failing drive this way, and about a quarter-day to copy a fully-working drive this way, so the excessively slow/dumb retry cycles in the firmware and storage stack mean your drive is spinning long enough to get worse, if it's decaying (something to consider for self-service vs. pro recovery).

Re:For the 57th time on Slashdot

[Enigma2175]

The issue is also what the humidity was in the room in which the head-disk module was assembled/sealed.

Except that hard drives are generally not sealed. They have a filtered breathing hole to exchange air with the outside. Otherwise, the casing would balloon when you took your computer on an airplane or when the drive is shipped via air.

personal experiences

[v1]

I've been a repair tech for the last 10 yrs. (and I don't mean I'm a "I built my own PC, I'm a computer god! I fix my friend/family''s computers" I actually know what I'm doing and have electrical engineering experience) I'd estimate I've seen around a thousand bad hard drives in that time. Of those, I'd say 65% would tap repeatedly, 25% had some io errors but were still working, 8% would sound normal but would never post on the bus, and the other 2% were the other weird issues like chirping or no power at all.

The tappers were very rarely recoverable by me. Every now and then I'd see one that if you powered it up dozens of times, you might get lucky and it would post properly and you could get data from it. None of the other common methods were helpful.

Over 90% of the drives with io errors and slow blocks could be recovered from. Most of those simply required a file level copy from bad drive to good. Most would have a handful of unrecoverable files. Depending on what was lost, an OS reinstall was sometimes required on the new drive, but not usually. A small percentage of them would have a large number of errors and require days to recover, or would fail completely during the recovery. A few of them would look promising but then quickly becomes apparent that almost nothing will be recoverable.

Sometimes a drive would stop responding during recovery and require a break. Trips to the freezer helped on about 30% of the drives. Some drives required numerous trips to the freezer, using rsync to resume copying where it left off last time, a process which could take days but could result in a complete recovery. I pondered ways to cool a drive during the recovery such as using a peltier, but never got anything implemented. I also use ddrescue and another custom script I wrote that works in a similar way, doing block-level recovery while splitting problem areas for smaller recovery chunks. That's useful for windows or other foreign OS where you can't do a file copy. (mac shop here)

I've never dried "drop therapy" or "impact maintenance". I'm sure it could help under specific circumstances like a stuck spindle or loose connection but I've never witness it.

I've done a little bit of onboard controller card ("OBCC") swaps for identical drives where the bad one wouldn't power on at all. About 25% success there. For that reason I tend to keep old tapping drives because their cards can work in dead drives. I assume the tapping drives have head failures, which isn't related to the OBCC. I've talked with multiple data recovery places about this process, and to my surprise every single one of them has told me "that won't work". They usually explain the remaps are stored on the OBCC, which makes sense, but isn't a good excuse not to try when the remaps probably don't account for more than one in a hundred thousand blocks. I think they just want me to send the drive to them.

The sled you place the drive into makes a HUGE difference in recovery. Avoid usb. I don't care if you insist on windows, install a firewire card. Almost all USB bridge chips handle misbehaving drives very badly. Only use one of those little external adapters with the build-on 2ft usb cord on it as an absolute last resort. OWC's "mercury elite aluminum" series are the best (reasonably priced) recovery sled I have found, and I have tried many. USB (39MB/sec, not 36, 26, 16, 12, etc), FW400, FW800, AND esata interface. In the past I used a Granite Digital "fireview", those absolutely rocked for drive recovery (LCD panel with diag menu....) but they stopped making them and they were IDE only. Someone needs to make a modern sled like that for sata please.

As for paid recovery, results seem random. Techs tend to have a recovery place they swear BY, and others they swear AT. But my observation is simply that methods vary and different places handle different problems with varying success. I think many techs' impressions are based on their first few experiences - if good they like, if bad they don'

Re:personal experiences

[Anonymous Coward]

We have personally had to send them a drive of ours - clowns at corp accounting didn't back up their data and lost an accounting drive. That was a $10k lesson I do hope they've learned from.

Accountants? Their backups are paper, and the backup method is glaring suspiciously at the computer while they use it.

Re:personal experiences

[v1]

Firewire? Why the heck would you...

Quite a few reasons really.

- I like to get my work done quickly, I'm impatient. MOST internal IDE/sata attachments require a reboot. It's inconvenient to have to reboot one of my service machines every time I attach a drive to it. Which happens many times every day.

- I can replace a laptop drive with a new drive, boot off a USB stick, and plug the failing drive in and run the recovery directly. There have been days when I've been running four recoveries at once. Often more efficient than trying to get several things going on a service machine at once. And with having to reboot it would get that much worse. You can't attach an old hard drive internally to a laptop at the same time as the good drive. (most cases)

- failing drives have a habit of hanging their interface until power cycled or unplugged. Again, not something you want to be doing with your service machine or even target machine. If it hangs, I pull the fw cable which un-hangs my service machine, power cycle the sled, plug back in the cable and resume. Cranky drives can require a dozen sled reboots, or periodic trips to the freezer. Having a recovery hang a service machine while I was trying to use it for something else at the same time causes TWO of my jobs to have to be started over.

- When necessary, I can get more than one recovery going at the same time on a single service machine. Can't reboot a machine for one drive while you're doing another.

- Firewire isn't as fast as many sata and most ide, but a lot of the time I need to be attaching one drive or another over firewire anyway. If I have a laptop booted into target mode and attached to a service machine for recovery, attaching the bad drive internally doesn't save me any time since it will have to go TO the good drive over firewire anyway.

- When I have a lot of things going at once around here, speed becomes less important than the ability to multitask, and firewire offers me many more options and flexibility. Careful consideration of your options, how you do things, and in what order greatly reduces turnover time. This can be the difference between a process taking four hours of time and monopolizing a service machine, vs taking 90 minutes of time and using one sled and a usb stick off in the corner while I do other things. OR 30 minutes of my time focusing on a single job vs shuffling something like that off to a corner to run for a few hours on its own while I get other things done. Most things I do can arguably be done in many different ways, requiring very different amounts of clock time, hands-on time, and various resources. Prioritizing, availability of options, and planning my methods are very important to my job productivity.

Re:Give Spinrite a try

[gweihir]

Pure snake-oil today. It used to have some merit in the MFM and RLL days, but these are long over. The only thing SpinRite can do today is to cause more damage to the drive if it has mechanical problems. If the drive is mechanically fine, repeated read accessed do exactly the same as SpinRite does, because it does not have any other possibility on modern drives.