DIY HDD platter swap?

[codeboy2k]

I've been bad.  No backups.

My home server was running in a 3 disk raid and 2 disks failed within 24 hours of each other.  Server is FreeBSD and the disks were in a ZFS RAIDZ pool.  All 3 disks are 2TB Seagate Barracuda 7200.14 / model ST2000DM001/ PN: 1CH164-301 / FW: CC4H

My biggest mistake was not having any backups.  My next biggest mistake was buying 3 disks two years ago at the same time which likely came from the same batch, and thus they might all be prone to a similar failure.  There is approx. 18500 hours on each drive. 

I had 1 drive fail first: the motor spins, then click click click, then the motor spin down.  The BIOS does not see the drive.  If it does see the drive, it sees it as only a 4 GB sized drive. The 2 remaining working drives were always seen as 2000 GB each.  With only 2 drives working, the ZFS RAIDZ pool was still working in a degraded state and I could still read things ok. 

So I picked up 2 new TOSHIBA 2GB DT01ACA200 drives to replace the broken one, and have a spare.  These disks have slightly more logical sectors than the Seagates, so not a problem for a RAIDZ swap.

Before simply just replacing the bad drive and let it re-silver, I wanted to make a backup of the important stuff. So I started backing up the important stuff onto one of the new drives, and before I could complete that task, the 2nd Seagate drive failed in the same way !! click , click, click, then drive spins down, BIOS won't see it. Now I have 2 failed drives and the RAID won't come up.  I didn't get all the data off.

Googling around shows me this is a common failure mode for the 2TB and 3TB 7200.14 Seagate Barracuda's.  Wish I had known that earlier.

There's also a 1.8V LVTTL serial port on the drive and some serial port commands for resetting things on the 7200.11 but warnings that doing the same  commands to the 7200.14 will likely kill it.  There's enough misinformation out there to warrant not trying anything like that just yet.  Some of the commands are to clear the SMART data and especially reset the reallocated sector counts (like it ran out, and cannot start now, but I'm not sure)

I'm thinking of doing the following:

1) a sector by sector copy of the one remaining good Seagate drive onto a Toshiba drive.
2) then trying to DIY a small cleanroom hood (just like like a chemistry hood, but with a positive pressure, HEPA filtered airflow outwards)
3) open the good drive and a bad drive and do a platter swap, close the drive, then try to copy a second Seagate drive onto a second Toshiba.

If that works, I can restart the server with the 2 Toshiba's and get a degraded RAID again.

So, does anyone think I have a chance in hell of being successful (with a DIY clean room like this)?

I would only need the swapped platter to last long enough to dump the sectors, which might take 2 to 3 hours , 4 hours tops.  If I get a degraded RAID back, I just want to get my data off it, then reformat everything and start new.

Does anyone know and willing to share the 7200.14 serial port commands that might breathe some life into the drive?

Thanks!

[SL4P]

1) a sector by sector copy of the one remaining good Seagate drive onto a Toshiba drive.
2) then trying to DIY a small cleanroom hood (just like like a chemistry hood, but with a positive pressure, HEPA filtered airflow outwards)
3) open the good drive and a bad drive and do a platter swap, close the drive, then try to copy a second Seagate drive onto a second Toshiba.

1) Yep - noting that the sector addresses will probably have little relevance to the RAID map
2) Hmmm.  But (at best) you want the HEPA on the input 'side' - otherwise you're sucking warm dirty air in, and on to the disks - then cleaning it to blow out back into the atmosphere!
I'd suggest pulling the air into an airconditioned room (dry air), then HEPA into you clean igloo tent.  But chances of it working is maybe 10% at best with platter alignment, embedded sector mapping, thermal expansin factors and a nunmber of other variables.
3) Your one remaining drive has half of the data (the second drive has the other half), and the third has redundancy data to reconstruct either half.... not a happy place.

I'd take the drives as a complete untouched set to Dr Disk and cross my fingers as well as my
bank account.  Don't spin them up any more in case there is any metal flying around inside the disk cavity!

[Mechanical Menace]

This why I don't use RAID5 or RAIDZ1 anywhere...

All I can say is what have you got to lose? It works with lower density drives if you're careful, should work here at least for long enough to get the data off. Hope you have a steady hand though.

[pickle9000]

When you say change the platters do you mean to swap the pcb to one of the faulty units and then do your backup?

I would maintain the seals and try the pcb swap first.

If you do plan to open the cover you can make a white box (large plastic container cut up), sheet of glass a hepa filter, anti static mat and surgical gloves. Liquid rosin on a lint free cloth (tac cloth) to wipe everything down inside the box.

[codeboy2k]

1) Yep - noting that the sector addresses will probably have little relevance to the RAID map
2) Hmmm.  But (at best) you want the HEPA on the input 'side' - otherwise you're sucking warm dirty air in, and on to the disks - then cleaning it to blow out back into the atmosphere!
I'd suggest pulling the air into an airconditioned room (dry air), then HEPA into you clean igloo tent.  But chances of it working is maybe 10% at best with platter alignment, embedded sector mapping, thermal expansin factors and a nunmber of other variables.
3) Your one remaining drive has half of the data (the second drive has the other half), and the third has redundancy data to reconstruct either half.... not a happy place.

I'd take the drives as a complete untouched set to Dr Disk and cross my fingers as well as my
bank account.  Don't spin them up any more in case there is any metal flying around inside the disk cavity!

1) I don't think it will matter.  If the same data is at the same sector number (LBA) on both drives (even though the Toshiba has MORE sectors) it should still work when restarted.

2) no, I mean blowing outward (positive pressure - HEPA on the inflow air) (not sucking inward over the drives). I think we both mean the same thing.

3) Your one remaining drive has half of the data (the second drive has the other half): that's an interesting way to look at it, and is a good argument for mirror only.  I'm considering reconstructing it as a mirrored server now, actually. or a RAID10 with 4 drives.

It's not worth it to me to go to a repair service. I've already decided that if I lose the remaining data, it will not be the end of the world.  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.

Nothing earth shattering, nothing that absolutely needed to be backed up except my personal documents, which is where I failed myself the most for not backing that stuff up earlier. Everything else can (almost) be recovered from other sources.

[SL4P]

I feel for you - I had a similar event about a year ago - RAID5 across three drives - one went down, followed by the second.
Luckily I had a reasonable backup strategy (raw to DVD), but the online backup SW I was using was proprietary / encrypted by default - so I lost about 15 months of stuff. Some irreplaceable - sigh.
Good luck!

[helius]

Be careful with alignment... When you loosen the screws holding the platters to their hub, you may discover that there is nothing keeping them rotationally aligned. It might be worth making a jig to hold the platters during the time they are off the hub.

[codeboy2k]

When you say change the platters do you mean to swap the pcb to one of the faulty units and then do your backup?

I would maintain the seals and try the pcb swap first.

no, not just a pcb swap, although I can try that first with the known good one.  Do I need to swap the serial EEPROM along with it?  I think so.
However, I don't think swapping the PCB will solve it. It seems more like trouble reading from the platter, and I suspect it's a head failure.

If you do plan to open the cover you can make a white box (large plastic container cut up), sheet of glass a hepa filter, anti static mat and surgical gloves. Liquid rosin on a lint free cloth (tac cloth) to wipe everything down inside the box.

Yes, like this, to open it up in a makeshift clean-hood with HEPA filter and positive pressure airflow, to open the drive and swap the physical platters onto the known good one.

The wipe down with liquid rosin is a good idea.

[codeboy2k]

Be careful with alignment... When you loosen the screws holding the platters to their hub, you may discover that there is nothing keeping them rotationally aligned. It might be worth making a jig to hold the platters during the time they are off the hub.

Seriously?  don't they have pins or screw holes to line them up? a few microns out and they may not  work
Unless you mean just within the relative alignment (like finding hole #1 on both platters of the 6 screw holes in the spindle)

[codeboy2k]

I'm mostly worried about shaky hands and causing head damage on the good drive.  Drives these days have really fragile heads.

And I'm good that I DO NOT have shaky hands   but still, all it takes is one slip..

[Psi]

I've never seen anyone successfully recover a HDD by swapping either the PCB or the platters with a good drive.
It probably depends on the brand, some might work, but i think most drives have their low level configuration split between flash memory on the PCB and data on the platters.
So swapping things around just corrupts stuff.  I can see it working if the motor is dead though, (swapping both the pcb and platters).

Does the drive only start clicking when you read data from it?  ie, when you start copying some file/folder. Or does it click pretty much as soon as you power it on?

If they only start doing the click click click when you read something then there is a chance they can be recovered.
First i would figure out what folder /file is causing the click click click by a few trial an error copes of folders. eg, get off whatever data you can first. If the drive is part of a raid you probably cant do this. Since it has to be mounted as the whole raid volume the controller will keep dumping it from the raid as soon is you add it. So move to the next step..

Run Spinrite on the drive/drives

Spinrite does a think called dynostat recovery which disables SMART and then tries to read the bad sector many times by instructing the drive to read some other random data on the disk and then to read the bad area in raw mode.
The result is the head coming at the bad sector from different directions and with repeated attempts from slightly different alignments as well.
With many iterations it will build up an average value for each bit in the bad area until it finds combination that passes the drives checksum with a good probability. At that point it re enables SMART and writes back the correct data to the bad sector. Since the drive will have this sector marked as 'reallocation pending' as soon as it sees a write it reallocates that sector somewhere good.

I've had good success using spinrite to fix drives that start clicking when a file/folder is accessed.
There was a time when there was good money to be made buying dead Ipod classic's from ebay (the type that have a real HDD inside). All they needed was a spinrite run on them and they were good to go.

[codeboy2k]

They can't be accessed anymore for any kind of sector copying (or spinrite).

They both now spin up when power is applied, click 2 or 3 times (I can't remember .. I think it's 3 clicks) then the motor shuts down and it never reports itself as being available to the BIOS.  So the drive does not report any READY status.

Googling for this and reading forums seems to tell me two things
1) the drive can't read the firmware track, so it can't boot
2) or, the remapped sector counter is full, so the drive won't start up.

For #2, there are some serial port commands to send to the drive to clear the smart data, and it can possibly be booted up in a bad state, but likely able to read some data.   I have not been able to find the proper serial port commands for a 7200.14 drive.

Also, I think I need to first hook up the serial port to the drive and watch the embedded boot, and see if it reports anything.

I'll do that first and copy it to here if anyone's interested.

[pickle9000]

I have done board swaps on older drives (500gig and smaller) successfully but only for data recovery and not all where a success. I think that platter removal would be a fail. Even keyed the tolerances are way too small, do it last.

[SeanB]

New drives are not going to survive the swap, as the alignment is critical, you need to align the whole stack to within 1nm, which is the data track width. In the factory this is done with a machine that moves the heads from the top through a hole in the case ( those little covers on the case) and writes the alignment data onto the clamped stack of disk platters. Loosen the bolts and tighten them again without doing anything will destroy this, just from the metal relaxing.

Try doing the controller swap and swap the eprom on the controller, which stores the head alignment info for the drive, and how it is aligned inside.

It will be the only thing you can do, other than sit with an electron microscope to read the magnetic data off the platters bit by bit.

[codeboy2k]

New drives are not going to survive the swap, as the alignment is critical, you need to align the whole stack to within 1nm, which is the data track width. In the factory this is done with a machine that moves the heads from the top through a hole in the case ( those little covers on the case) and writes the alignment data onto the clamped stack of disk platters. Loosen the bolts and tighten them again without doing anything will destroy this, just from the metal relaxing.

This is the information that decides it for me. Thanks. Not gonna waste my time then. Yes, now I seem to recall Free Electon wrote about this alignment procedure too, that's done from the taped hole on the side of the disk drive. 

Try doing the controller swap and swap the eprom on the controller, which stores the head alignment info for the drive, and how it is aligned inside.

It will be the only thing you can do, other than sit with an electron microscope to read the magnetic data off the platters bit by bit.

ok.  I think I'll first try to get the serial port wired up and capture the boot log, then maybe there's some serial port commands I can do.

if swapping the platters is going to be a problem, perhaps I can swap the head stack? i.e leave the platter in the bad disk screwed down, but swap the good heads and the good controller (and the EEPROM from the bad controller) and maybe that will have better luck?  Yes, I know the heads are REALLY delicate and fragile and cannot bounce or touch together, they will be destroyed.   I opted for a platter swap first because I know a head swap is much harder.
But if you think that a platter swap is almost guaranteed NOT to work, then I'll start with a head swap. 

And it's almost an exercise to see if it will work, simply because I'm ready to lose the data, but would like to try to save it still.

Thanks for all the suggestions so far.

[Psi]

New drives are not going to survive the swap, as the alignment is critical, you need to align the whole stack to within 1nm, which is the data track width.

Modern drives have data packed so dense that the data is read as the head moves over (and past) the track it's looking for. The head doesn't simply move to track X, stop and read the data because there's not enough resolution in the head control DAC by an order of magnitude or two.

Old hdds and floppy disks used to seek to a specific track. They had a stepper motors to control the head whereas modern drives do head control using a electromagnet allowing them to move position with infinite steps but not much absolute accuracy.

So the drive should be able to tolerate platter alignment. It has to home-in on the track anyway so it shouldn't matter where it is.

Atleast, this is how i was told they work.

[wraper]

So the drive should be able to tolerate platter alignment, it has to home-in on the track so it shouldn't matter where it is.

In your dreams. Even those who professionally recover the data have a very big trouble to move the platters, even with special tools. I'm not even sure is this is possible at all with newest HDDs, though could be done about 10-7 years ago but not at home.

[wraper]

At best head block could be tried to be replaced and work afterwards, but no way platters.

[codeboy2k]

So given that the heads do not move with absolute accuracy, but the platters MUST be aligned, do you guys think it would tolerate a head swap better?

[codeboy2k]

At best head block could be tried to be replaced and work afterwards, but no way platters.

Ha.  you answered my question while I was typing it.

All in all it's going to be a learning experience. I'll keep the thread informed.

[amyk]

Yes, a head swap would have much better chances, although the chances are still quite small...

The problem isn't with rotational alignment as there is no "index position" (sector numbers are encoded along with data), but with radial alignment (runout). The head servo can track runout to a certain extent, but too far and it won't be able to follow the track.

[wraper]

And this was made with 750 GB HDD.

Please note, this process only works with 7200.10 model Seagate hard drives. Do not attempt on any other model Seagate drive, or any other brand hard drive...especially Western Digital.

In addition, having proper software for imaging is crucial. After the platter swap in this video, this is absolutely no way this drive would have been accessible on a computer, no matter how it was attached. The drive had many areas where it really struggled to image completely, and our Deepspar Disk Imager is one of the few tools on the market that intuitively knows how to handle unreadable sectors without pushing the hard drive to its limit.

[Richard Crowley]

Swapping platters is way beyond anything that could be considered remotely promising.Even if you could get the proper clean environment (which seems VERY doubtful) and get the platter(s) into proper mechanical alignment, I strongly suspect that the tracks are self-aligned (i.e. the tracks are defined by the actual heads in the drive and not to some "standard".  So swapping platters into a different drive may never "align" with the receiving head stack.

Swapping boards seems like the only kind of field-procedure that has any kind of reasonable expectation of success.  And even that method may be less promising with newer drive technology where so many factors are firmware and saved in NVRAM.

If the data is really valuable, I would send it off to a professional recovery company who know how to do it right. This hardly seems like DIY territory if you at all serious about actually recovering the data.

[codeboy2k]

It's not valuable which is why I'm willing to try to DIY it.   Just enough to get it to run for about 4 hours so I can image the drive, that's my goal.

[wraper]

I strongly suspect that the tracks are self-aligned (i.e. the tracks are defined by the actual heads in the drive and not to some "standard".  So swapping platters into a different drive may never "align" with the receiving head stack.

That's actually true. AFAIK, when swapping heads, multiple donor HDDs can be scrapped just to find the head block which is closest by it's parameters to the faulty one.