The lab · finished work
Every job closed at this bench is written up here once the customer has confirmed the result and the names have been stripped out. Each entry points to the guide for that fault, so opening one shows how that kind of recovery is handled.
Describe the fault to an engineer
0800 6890668
Two 8TB disks and two 4TB, one member already down, and the TeraStation rebuild ran overnight only to halt with the array marked failed. Underneath sat Buffalo's own RAID descriptors and an XFS volume the firmware would no longer mount. We imaged all four, worked out member order and offset from the images, and every share opened exactly as it had.
An emergency repair was already queued when the practice rang, the overnight power cut having left the database marked Suspect. Forensic copies were taken first, and they showed the transaction log would still replay. Attaching on that copy worked at the first attempt, which left the morning's bookings twenty minutes behind rather than a week.
The unit had been set to RAID 0 at installation, so its two disks carried one striped volume with no redundancy at all, and when the older disk stopped being seen the whole volume went with it. Both members were imaged first, the failing one slowly. Stripe size, member order and start offset were then worked out from those copies, and the volume opened from the images.
Fusion presents a blade SSD and a 3.5-inch disk as one volume, so a fault on either side takes the whole of it away. Nothing was making a noise; an upgrade had left the CoreStorage metadata inconsistent. The display came off, both halves were imaged in full, and the volume rebuilt from those copies.
The orange bumper did its job; the disk inside did not. The drive went off a desk mid-transfer, and afterwards the light came on but nothing ever mounted. It was left alone after that one attempt, which is what saved it, because the platters stayed as the fall had put them. Inside was a Seagate mechanism, as LaCie units usually are, so a head stack off a matching donor went in and the drive read again. Imaging ran in short sessions across a week, the weakest zones last, and nearly all of the owner's work came home.
Six hundred files, and not one fit to print: the free software had saved only the embedded previews. The card had been left alone since, so it imaged cleanly, and the CR3 originals were reassembled a fragment at a time. The gallery went home at full size, forty frames longer than that tool had managed.
Nobody noticed disk two drop out in June, so when disk four left the array mid-rebuild in August it looked like a double failure. It was not. Imaging found nineteen bad sectors on that fourth drive and nothing worse, so three intact members plus one near-complete fourth rebuilt the stripe, and the filings met their deadline.
A UPS that ran down before the mains returned left the Smart Array asking questions nobody in the practice was willing to answer, and rightly so. Every member was copied first, the RIS metadata read off and the delayed parity accounted for, so the rebuild ran on images alone. The practice-management database opened whole, current to the cache's last committed write.
A student in Maidenhead was handed back a laptop that ran perfectly, on a new and entirely empty disk. The original, written off as dead, had been sitting in a drawer, still clicking. Donor heads went in, and two days later every essay and photograph was waiting on the replacement drive the shop had fitted.
What made this one awkward was the rebuild someone had already attempted on the RAID under the booking system. Every member drive was imaged before anything else was touched, the array was then reassembled virtually, and a verified database went back into service by Thursday lunchtime. Priority queue throughout.
Nine days out from the hand-in date, the drive began to tick. A donor head stack went in, the image was taken at a deliberately slow rate, and the bridge board gave up its encryption keys; the thesis went back complete, references and all, with a week to spare.
The two red lights were survivable; the rebuild attempted afterwards was not, since it overwrote the box's own metadata. Both disks were imaged straight through their bad sectors, and the mirror was reassembled from whichever copy read more cleanly at each point, until the shared volume was whole again.
By breakfast the 512GB SanDisk that had carried a whole wedding the night before was reading as empty. A clone came first; the file system was then rebuilt on that copy, and a deep carve gathered up whatever had been orphaned. Every frame came back, and the photographer kept the client.
A Fusion Drive pairs an SSD with a hard disk, and here the hard disk was failing, taking the logical volume down with it. We imaged both, rebuilt the mapping that ties the two together, and returned a decade of a design studio's work from a Mac the Apple Store had written off.
Its disk had started to click, and the laptop itself was past repair. A donor head stack went in, the image came off slowly, and inside three days she had five years of marking, reports and lesson plans back.
Two heads had failed, and nothing inside had turned in weeks. With a donor head stack fitted, the first imaging pass took 96% off, and the ledgers, invoices and photographs all went back checked.
A firmware update failed part-way and left the laptop's system SSD unreadable. Working in technical mode, we rebuilt its translator tables, and a decade of one photographer's edited catalogues came back whole.
Production stopped when a Friday rebuild claimed a second member and locked the live data away. Each disk was imaged separately; the stripe was reassembled from the three that read cleanly plus whatever the fourth would give up, and the line was running by Monday morning.
The tax return was the deadline, and the stick carrying the CIS records and seven years of invoices had been stood on in a site office until its connector came away. Monolithic construction left the test pads as the only way in, so the controller pinout was mapped here first. Everything filed on time.
Nothing was switched on, which is what saved it: the drive was sealed into a bag while still damp and posted the same day. At the bench the contacts were cleaned, a donor board was fitted carrying the ROM off the original, and the disk was imaged in full inside the week. Twenty years of family photographs came home.
It was clicking by the first test on the bench, after going onto a hard floor while a wedding shoot was packed away. Three years of client work sat on it. A donor head stack went in, and two days of imaging left not one gallery missing, the morning's shoot among them.
Because an overnight firmware update shifted the TPM measurements, the workstation woke asking for a recovery key nobody had ever written down. The Recovery Key ID led us to the practice's escrow in Entra ID, and repair-bde then rebuilt the damaged BitLocker metadata onto a clean disk. Not one project file was lost.
Nothing was encrypted where it lay. Each file was copied, the copy locked and the original deleted, so what mattered was still in free space waiting to be carved, and a NAS snapshot they had overlooked filled the gaps. Trading resumed inside the week, nothing paid and no one answered.
One evening it came out of the socket mid-write, and by morning Windows would only offer to format it. The whole disk was imaged before anything else, and the boot sector and damaged MFT mirror were repaired on that copy alone. Fifteen years of client files went home with the folder structure still standing.
Nothing had been written to the disk after the mistake, and that is what saved it. Two identical 2TB disks sat in the one tower, and the fitter quick-formatted the family archive rather than the spare. The original file table survived, so every folder returned under its own name.
Shift and Delete on the wrong folder took a year of bookkeeping off an office PC, bypassing the bin entirely. An image was taken the same afternoon, the file table rebuilt from that copy, and every name and date came back with it.
An overnight outage took the board with it: the TVS diode shorted, which is precisely what that component exists to do. By morning there was no spin and nothing in BIOS. The ROM moved to a donor board and the disk imaged in full the day it was booked in.
Five years of EPOS data sat on a disk that ticked roughly once a second when it came in. Two donor head assemblies went in, the image was taken slowly over several days, and the shop's VAT return went off on time with the ledger complete.
Overnight, the clip an insurer needed disappeared from a cab camera. The card was imaged first, and the image showed a deletion rather than an overwrite — which is the whole difference. Repaired and hashed, the footage went back within the week, and it answered the insurer and the HR file at once.
By the time the insurer asked for proof of two vans meeting in a yard, the recorder had looped well past the date. The only copy had been an export a manager made and then deleted, and nothing of it survived elsewhere. That file was carved back off the disk, hashed, and set out in a short report; the claim closed without a hearing.
Four of the five lost days were still described in the recorder's index, timestamps intact, and the fifth came back through carving as playable video with no labels. What saved it was switching the recorder off that same afternoon, after a member of staff deleted a batch of 'old clips' — one of which held a theft.
By the ninth day the recorder had been overwriting itself for days, and only then did the insurer ask. An image was taken, the recorder's index decoded, and the break-in lifted from sectors the loop had not reached. Two exports went back, one to watch, one for the claim.
The uploads finished twenty days before the resignation letter arrived, which is what made the sequence so plain: documents pulled together in one place, a private cloud account signed in to, then the transfer itself. A Henley-on-Thames employer had only a rival's pitch deck to go on. It closed on undertakings.
Two companies, two sets of accounts that would not reconcile, and one folder missing between them. Neither version had come through the image untouched, and the timeline said as much in plain terms. Settlement followed within a month.
A departing employee set a wiping tool running the evening before the laptop went back. It managed about fifteen minutes, and its own log came through in the image: when it began and which files it never reached. That was enough to set out the sequence.
The client was certain the files had been taken deliberately. The evidence pointed somewhere duller: a sync client set up wrongly had cleared them from the shared drive, and the greater part was recoverable. That answer closed the dispute faster than an accusation could have.
45 cases, anonymised and verified, each set out in full: what arrived, what depended on it, what we did with it, and what went home.
Each of these began the same way yours would: a diagnosis that costs nothing.