DIY HDD platter swap?

[codeboy2k]

(I'm wordy) TLDNR:  I agree with what you're saying and I know I need to keep the limits of my abilities in check.

Hi thanks! yes your comments hit home with me, it certainly is worthwhile data and does not have $0 actual value, but I'm trying to decide if it has $1600 value ?  That's the cost since it is part of a raid, and the few sites that I probed wanted $800 per disk in the raid so they can recover files at the file level.  I am not sure if that means the total is $1600  or $2400 yet (2/3 drives or 3/3 drives)

I don't want that kind of recovery, and if I can find a reputable data recovery company that is willing to just mirror the data off the disk onto a new disk (with block read errors set to 00000), then I will have 2 disks of the raid again, and then I can setup a degraded raid and recover my own files. The blocks that could not be read will have checksum errors during my recovery and will be skipped by ZFS, it's good about that, so I may ultimately lose just a few files. Much better scenario indeed. It sounds even plausible, as this failure mode appears to be well known and common for the drive I had.

I've taken disks apart before (ones that were broken already!) so I'm familiar with the internals, but I have never tried to make a bad one work again in this manner.   I have some old disks here on the shelf, so I am planning to build a "clean-box" and test out a head swap on some working disks.  Some people online did this already and had mixed results. Good thing is I don't smoke or drink and have ultra-stable hands.  If I can get the heads off both disks without incident, and put one back, and don't let the heads touch or snap or bend or all the other bad things that can go wrong, then I might have a chance.  But I will certainly practice on about 10 other disks I have here already.

I plan on screwing down the donor drive and the bad drive to the tabletop so they don't move during the process, and screw the jig to the tabletop. Even a bump on the table top during the surgery will kill it, so I am still worried about that possibility and want to be aware of it.  The operating room needs to be rock solid

I've certainly been thinking about how to save the data.  At this stage I just have the disks set aside, no freezer, no serial port attempts. I've duplicated the good drive onto another drive, so I have two copies of the one good drive.  I'll be making another block-by-block copy onto the new server (since it's empty now !) and then I'll still have two good copies when and if I decide to tear into the good drive for it's heads.

If I'm unsuccessful with the practice runs then I won't even bother, and I'll wait until I have the spare cash and take it somewhere that can recover it block by block.

Cheers, and thanks for your comments.  It certainly does make me think twice about what I believe I am capable of and what I am actually going to be capable of.

[BradC]

I remember reading somewhere - putting the HDD to fridge/freezer may make it readable 

Despite what the others are saying, freezing *can* work in a specific instance. Freezing works where the failure in the drive is related to mechanical tolerance of either the spindle bearing, or more usually the head actuator bearing.
In that case, I wrap the drive in several freezer bags, and using long power cables and a USB/ATA(or SATA) adapter) put the whole thing in the freezer and run the recovery in one hit. The more thermal cycles the worse it works.

Like I said, it's a single scenario, it's rather un-common compared to other failures, but in the right case with the right failure it can help immensely.

[wraper]

Likely an urban legend from the days when stiction was a common problem. A drive with stiction can't turn its spindle motor because the heads are cold welded onto the surface.

Still very actual problem for notebook HDDs, personally disassembled such with stuck heads. Though almost non existent for 3.5" HDDs.

[kc8apf]

Freezing doesn't work with new drives.. I'm not even going to try it.

I'll do a head swap when I get time to make a portable "clean-box" .

As for that last post of mine, where I said UDMA_CRC_Error_Count = 36 on one of the new drives,  I swapped cables, swapped controller ports, but nothing new happened and nothing changed after days of additional read/write testing. That number stayed at UDMA_CRC_Error_Count = 36 and the other drive never gained any errors when it was swapped around too.

So I think that was an anomalous event that caused the controller attached to that specific drive to briefly screw up, but I'm not seeing it again.

Today, both drives are in service now, and it's still UDMA_CRC_Error_Count = 36 for that drive and 0 for the other drive.

This is a pretty common problem at my work.  We collect extensive data on every hard drive at installation, during usage, and at failure.  UDMA CRC errors tend to show up at installation and correlate very strongly with being resolved simply by reseating the cable.  It's not necessarily a bad cable or connector but just that it didn't quite seat well.

[Delta]

Excuse my lack of knowledge, but why do you need more than one good disk to recover your data?  I thought the whole point of a RAID array was that there was redundancy between disks?

[helius]

As was addressed earlier in the thread, multiple drive failures sometimes happen in clusters. Identical drives from the same batch, run for the same number of hours, have a higher probability of failing at the same time; especially if the failure is caused by a firmware bug.

[wraper]

Excuse my lack of knowledge, but why do you need more than one good disk to recover your data?  I thought the whole point of a RAID array was that there was redundancy between disks?

Because he had RAID5 with 3 hard drives, not RAID1.

It requires that all drives but one be present to operate. Upon failure of a single drive, subsequent reads can be calculated from the distributed parity such that no data is lost.[15] RAID 5 requires at least three disks.

[codeboy2k]

As was addressed earlier in the thread, multiple drive failures sometimes happen in clusters. Identical drives from the same batch, run for the same number of hours, have a higher probability of failing at the same time; especially if the failure is caused by a firmware bug.

and I learned my lesson about using 3 identical drives from the same batch in a RAID.  My rebuilt server has just 2 new drives in a RAID1 and these are from Hitachi and Western Digital.  I made sure to buy two different drives this time.

[codeboy2k]

This is a pretty common problem at my work.  We collect extensive data on every hard drive at installation, during usage, and at failure.  UDMA CRC errors tend to show up at installation and correlate very strongly with being resolved simply by reseating the cable.  It's not necessarily a bad cable or connector but just that it didn't quite seat well.

Thanks! I'll just keep an eye on it now. 

What kind of extensive data do you keep for drives?  If I had huge numbers of drives in play I'd probably keep a database for each serial number with smart data snapshots so you can plot it over time, and additional data such as performance data too, i.e. monthly read/write speed tests .  If I was monitoring drives for pre-failure that would be helpful..  Something I think I should have done at home here too, and maybe I will now.

[kc8apf]

Pretty good guess.  At install time, they run the drive through a set of diagnostics that does things such as read/write every sector, verify read and write performance at the inner and outer diameters, etc.  They then record a snapshot of SMART as well as a bunch of vendor-provided statistics.  While in use, SMART and the vendor-provided statistics are recorded every 15m.  Filesystem reported errors (read/write failures mostly) are recorded if they happen.  Any information the SATA controller can provide is logged too.  If those correlate with any patterns we've previously seen for failing drives, the OS is told to stop using the drive.  We then run a more extensive version of the diags that we ran during install.  If it all checks out, which is often the case, the drive gets put back into use.  If anything fails, the drive is erased and RMA'd.

The failure patterns are derived by running regular analysis over the entire set of data for every drive ever deployed.  At this point, that data set is more information about drive behavior than the drive manufacturers have.  They don't have the budget or need to keep that many drives running constantly.  They usually only keep a few 1000 drives of the most recent models running to see when they start to fail and why.

[CatalinaWOW]

I am glad you are at least thinking about the value of your data.  Personal time is very hard to assign a dollar value to.  Those who try to value it the same as their pay scale are making one of two errors, both of which are huge, and are in opposite directions.  One error is assuming that they can convert every hour they have into dollars.  Even those who work hourly consulting jobs have trouble making this true as there are always down times or expenditure limits, and most others cannot even come close.  The other error is that time with their wife, kids, mates and hobbies has no value, or is valued only at their pay rate.  That time can in fact be priceless.

I am guessing that you didn't have these drives full, perhaps only two thirds of a terabyte of data.  Probably a sizeable fraction of that is totally disposable (Utube videos, tunes you never listen to and so on).  But there is likely to be several hundred gigabytes of things you would rather not lose (tunes you DO listen to, web data that has been useful, instruction manuals and so on) and things you would really hate to lose (scans of family photos, downconversions off of video tapes, your own notes and creations, .mp3s from recording sessions or off of old vinyl, clever bits of code and the like).  The things you would rather not lose can be fairly easily replicated, but it takes time measured tens of hours.  The really hate to lose stuff is even slower to get back, and in many cases impossible. 

Even if the budget doesn't have room for the expense right now you might be better off in the long run putting these disks on a shelf until you can save the money.

[codeboy2k]

I'm definitely one who cringes every time I hear someone say their off-time (or personal time) is worth as much as their billable time.   I simply know that's not true.   I usually have arguments with my wife about this ALL THE TIME! because she thinks my personal time is worth my billable time, so if I spend a week in a woodshop building furniture because I enjoy it, than in her eye I am losing money because I'm not being productive. She would want me to buy the furniture and use my time to make money.  We argue this a lot, much to my chagrin.  But I agree with you, that it's wrong for me to say it has zero value.

As to the contents of the disk, you nailed it, it is exactly all that you said.  From my second post in this thread :

It's a personal loss, representing years of work, more than anything else. 15 Years of scanned personal documents, (tax data, tax returns, retirement savings, etc), my family photos (scanned from print or slides), approx. 600GB of datasheets, scanned my bookshelfs, ISO copies of my subscription CD's which I don't have anymore, my online books and magazines collection, DVD and music CD collections..  etc.

So I certainly hope to get it back, but if I can't then I'll be able to recreate most of it too.

[Chris C]

Likely an urban legend from the days when stiction was a common problem.

It's not an urban legend.  I've recovered two "click death" drives by chilling them, one PATA, one SATA.  Both 3.5", and both continued to operate only as long as they were cold.  Multiple cycles of chilling and recovery were necessary.  With high priority files first just in case, but I was able to achieve 100% recovery on both.

As for why it works, I don't know.  But limiting the possibilities to stiction, or even to something mechanical, seems myopic.  It could be that there's a cracked solder joint, IC wire bond, or some other intermittent that could be pushed temporarily back into good contact by thermal contraction.  Or because something electronic has gone out of spec, and electronics work better with less noise when cold, allowing it to edge just enough back into spec to function.

[madires]

When you search youtube for "Scott Moulton" you'll find several of his Defcon presentations about harddisk recovery.

[codeboy2k]

I was able to recover an old QUANTUM hard drive (anyone remember them?) back in the 90's via freezing, so it does work, with some drives.

I was told repeatedly and read over and over again in many forums that freezing modern high density TB level drives just does not work, and may cause more damage making it completely unrecoverable. 

So I really don't even want to try it.

Besides, I don't think it's the kind of problem that freezing may fix.  There is no stiction, the drive spins, trys to read a few things on the platters, the heads move, click 2 or 3 times , retract then the motor spins down.

When you search youtube for "Scott Moulton" you'll find several of his Defcon presentations about harddisk recovery.

Thanks, I'll do that!

[Chris C]

I was told repeatedly and read over and over again in many forums that freezing modern high density TB level drives just does not work, and may cause more damage making it completely unrecoverable. 

So I really don't even want to try it.

Besides, I don't think it's the kind of problem that freezing may fix.  There is no stiction, the drive spins, trys to read a few things on the platters, the heads move, click 2 or 3 times , retract then the motor spins down.

Both drives I recovered did the exact same thing.  Spin up, seek and click, seek and click, seek and click, then shut down.  The second was a Seagate 1.5TB, SATA2.

Funny how the world is full of consumer products with bad lead-free solder joints that open up when they reach a certain temperature, sometimes room temperature.  And technicians who identify a failed IC by hitting it with freeze spray, bringing it back to life.  Yet when it comes to chilling a hard drive, people always say it's only good for certain mechanical issues like stiction, or it's an urban myth.  Hard drives contain electronics and solder, do they not?

Though I can imagine how condensation on the platters, especially if it freezes, could cause damage.  I doubt the internals of my drive ever got down to actual freezing point, as I only had it in the freezer for 30 minutes at a time.  I wish now I'd taken some temperature measurements.  But it was enough for 10-15 minutes of reliable operation.  Took many cycles and pretty much all day.  Putting an ice pack on the drive while operating didn't noticeably help (I tried).  Oddly, on the very last recovery cycle, the drive seemed to start working for good.  Ran it for an hour straight without issue, during which time I double checked my recovery.

I see you live in Canada, though I don't know exactly where.  Average temp and relative humidity right now for Ontario is 68°F and 64%, putting your dew point at 54°F.  Chill the drive to just short of that (or whatever your actual conditions are), and you will have no issues with condensation.

Swapping platters is a pretty extreme idea, so here's another one.  You might be able to partially dessicate the drive, allowing you lower temperatures still.  Will it be sitting around for some days/weeks, while you decide what to do?  Put it in a sealed container with silica gel or calcium chloride.  Dessication could be accelerated by pumping air out of the enclosure to some pressure the drive can tolerate, wait for a few hours, backfill with dry nitrogen, wait again, repeat a few times.

Or is it possible your drive is hermetically sealed?  I've heard mention of some new ones being sealed, and filled with helium.

Yep, I remember the Quantums.  Especially the aptly named "Fireball".

EDIT:  I seem to remember something about condensation tending not to form distinct droplets on very clean surfaces.  I can't think of anything cleaner than a hard drive platter, so maybe short-term, condensation isn't so big an issue as one might imagine.  Even if decent-sized droplets did form, wouldn't they be flung off the platters by centripetal force, before they achieve the required velocity to cause collision damage to the head?  Just musing at this point.

[codeboy2k]

yes, the drives will be sitting around for a while..

I have a vacuum food bag sealer  and dessicant.. maybe I can throw in a dessicant pouch and vacuum seal the harddrive and let it dry out.

yes, I think the biggest worry is condensation when you bring it out of the freezer .. so definitely need to keep it above the dewpoint.

I need 4 hours of operation to image the entire disk.. that would still be my goal.. just make an image.

From what's been said here, I understand that platter swaps are difficult because of alignment issues, and that a head swap is more likely to be successful, so that's the route I would take if I get there.  I'll not move the platters out of alignment.

[kc8apf]

For 2TB and up, the primary failure more is head failure due to physical damage. Impurities on the platters stick up like mountains and the head fly height varies enough that the head occasionally runs into the peaks. I've got electron microscope shots of this somewhere from a pile of drives that my team asked the vendor to do a full teardown and analysis on.

Head or media failure lines up exactly with the described symptoms (spin up, heads released, a few seeks, park heads, spin down).  It could be a few of the firmware sectors went bad though it would need to be across the redundant copies. Head failure causes the self tests to fail.

For modern Seagate drives, if the firmware gets damaged, the drive will go to a boot loader mode that reports a special model and serial number.

[max666]

...
Funny how the world is full of consumer products with bad lead-free solder joints that open up when they reach a certain temperature, sometimes room temperature.  And technicians who identify a failed IC by hitting it with freeze spray, bringing it back to life.  Yet when it comes to chilling a hard drive, people always say it's only good for certain mechanical issues like stiction, or it's an urban myth.  Hard drives contain electronics and solder, do they not?
...
Yep, I remember the Quantums.  Especially the aptly named "Fireball".

I still don't think a broken solder joint is very high on the list of most likely drive failure causes.

Man I wish I hadn't used my Quantum Fireball for lethal magnetic field strength testing. The Fireball I have, had been in daily office use for many many years, without any issues (easily 10 years).
But I wanted to know if I could quick erase a HDD by swiping it with a magnet (D70mm x H40mm N42). Long story short, nothing magnetically I tried resulted in any bad sectors.
So I opened her up and let her spin, and to my surprise the thing worked perfectly fine without the lid, admittedly this wasn't a very high density drive (I believe 30 GB or so).
Now you won't believe me, but I even touched the spinning surface very lightly with the screwdriver and even that didn't produce any bad sectors. I had to push the screwdriver hard enough to scratch to surface in order to get bad sectors. (Disclaimer: the excessive scratiching seen in the picture is just me going to town, it didn't take that much scratching to produce the first bad sectors)

[kc8apf]

I still don't think a broken solder joint is very high on the list of most likely drive failure causes.

Man I wish I hadn't used my Quantum Fireball for lethal magnetic field strength testing. The Fireball I have, had been in daily office use for many many years, without any issues (easily 10 years).
But I wanted to know if I could quick erase a HDD by swiping it with a magnet (D70mm x H40mm N42). Long story short, nothing magnetically I tried resulted in any bad sectors.
So I opened her up and let her spin, and to my surprise the thing worked perfectly fine without the lid, admittedly this wasn't a very high density drive (I believe 30 GB or so).
Now you won't believe me, but I even touched the spinning surface very lightly with the screwdriver and even that didn't produce any bad sectors. I had to push the screwdriver hard enough to scratch to surface in order to get bad sectors. (Disclaimer: the excessive scratiching seen in the picture is just me going to town, it didn't take that much scratching to produce the first bad sectors)

The platters are composed of many layers.  The outermost layers are a lubricant and then a protective coating.  The actual magnetic layer comes next.

[codeboy2k]

For 2TB and up, the primary failure more is head failure due to physical damage. Impurities on the platters stick up like mountains and the head fly height varies enough that the head occasionally runs into the peaks. I've got electron microscope shots of this somewhere from a pile of drives that my team asked the vendor to do a full teardown and analysis on.

Head or media failure lines up exactly with the described symptoms (spin up, heads released, a few seeks, park heads, spin down).  It could be a few of the firmware sectors went bad though it would need to be across the redundant copies. Head failure causes the self tests to fail.

For modern Seagate drives, if the firmware gets damaged, the drive will go to a boot loader mode that reports a special model and serial number.

And that's why I want to try a head swap in a "clean box". But if the media is damaged and it can't load the firmware, the bootloader mode should still allow it to load firmware from the serial port, correct?   I would need to have the properly formatted firmware and the correct software download tools (which I don't have!).  I think I need to get a serial port up and see what it says ..

[knks]

Another urban legend I heard - how to easily create a "clean room" for safe HDD opening. Go to your bathroom, turn on hot water in your shower until the bathroom is full of steam, turn off the water and wait for steam to condensate (on the walls, not the HDD). All the dust will be removed by steam.

Is it myth or not?

[helius]

This bathroom technique is used by photography hobbyists. It does reduce airborne dust, but the dust will still be present on the walls and counters so is not a "clean room" in any sense. Dust also sheds from your skin and clothes.

[codeboy2k]

Interesting technique, but I don't think I'll try it at home  

I'm still planning to make a simple clean box using two identical plastic tubs from IKEA or some other store, and duct tape them together, a large one on the bottom for the work space, one smaller one (cut down in height) on the top for the filtration and laminar flow box.  I'll put a HEPA filter feeding into the bottom box, and a smaller 5 micron filter feeding into the top box. A blower fan will be in the top box, sucking air from outside through the top 5-micron filter and pushing it into the bottom box through the HEPA filter.  So the bottom box will have positive pressure and the top box will have negative pressure.

I'll cut an opening into the bottom box for my hands, and positive clean air will exit via this opening, keeping dust out.   I'll use clean room gloves (I  requested a sample from a distributor) and I'll use some non-linting cloth to cover my arms.  I am considering using this conical 1 micron PE filter bag from Amazon as arm covers:  1 Micron Singed Polyester Filter

I'll cut a hole in the small end of that conical filter for my hands to exit, and wear the gloves over top.  Basically making arm covers that can block 1 micron particles.  This should keep skin cells and small dust particles on my body out of the clean box work area and anything that did get in would get (hopefully) blown out the front via the positive pressure.

The whole setup might be less than $100.00 and I'll do my best to make sure I have a proper clean environment to work in.

[CatalinaWOW]

You might want to find some information on clean room startup.  I am no expert, and don't see anything fundamentally wrong with your approach, but do know that if there is any chance of mechanical jarring to your boxes and or any surfaces inside them (like the outside of the disk drives) those surfaces need to be clean or they will shed particles.  Full size clean rooms spend a lot of time cleaning surfaces at startup as part of the solution to this problem. 

Your solutions might include things like letting your box run for hours after loosening, but before removing screws and things like that.