A farming business outside Atherstone runs its accounts, its field records and fifteen years of scanned paperwork off one external drive that moves between the farmhouse and a laptop in the office. Somebody unplugged it while a backup was still finishing, which is a thing that happens on every farm and in every office in the country. From then on Windows opened the drive and immediately reported that the file or directory was corrupted and unreadable. The bookkeeper wanted the accounts before the quarter end and had sensibly not run anything at it.
| Media | An external drive that mounts and then reports that the file or directory is corrupted and unreadable, following disconnection during a write. It carries a farm business's accounts, field records and scanned paperwork. |
| What was reported | External drive, unplugged part way through a backup, now reports that the file or directory is corrupted and unreadable whenever it is opened, accounts and scanned records needed before a quarter end, nothing run at it since. |
| Fault class | Filesystem damage caused by an interrupted write. The structures that record where files sit were part way through being updated when the power went, so they no longer agree with one another and Windows refuses to read further. The files themselves were complete and in place. |
| On the bench | Read behind a hardware write blocker so nothing could be written and no in-place repair could be attempted on the original. Imaged on DeepSpar behind PC-3000 UDMA in a single clean pass. The filesystem structures were then parsed and rebuilt from the image in R-Studio, with the damaged directory records reconstructed against the backup copy the volume keeps and the journal replayed to establish what the interrupted operation had been doing. Files were opened by sample across every year before being written to fresh media. |
That message is about a structure, not about a document. Windows reports the file or directory is corrupted when it reads a record that does not make sense, and it says it whether the record describes a spreadsheet or the root of the volume. When it appears the instant the drive opens, the damaged record is near the top of the tree, which is why nothing at all is reachable rather than one folder being awkward.
Interrupted writes damage bookkeeping in a very particular way. Updating a file means changing several things at once: the entry itself, the record of which space is in use, and a journal that keeps track of the change. Pull the power between those steps and they disagree. The filesystem is designed to survive that in the ordinary case, which is why nine unplugged drives out of ten are fine, and this was the tenth.
The volume keeps a second copy of its own index, and that is the way in. These filesystems maintain a backup of their central record, held elsewhere on the disk and updated less often. Where the primary is damaged, the backup usually holds a slightly older but consistent picture, and combining the two rebuilds a tree that is accurate for everything except the handful of files touched in the final seconds.
Repairing in place is the thing to avoid, however tempting the offer. The check-and-repair tool is designed to return a volume to a usable state, which is not the same aim as returning your documents. It will discard records it cannot make sense of and lift orphaned fragments into a folder of numbered pieces, and it does all of that to the only copy you have. Working from an image costs a few hours and leaves the original exactly as it arrived.
The drive profiled healthy: full speed, no reallocated sectors of any consequence and no unstable regions. That matters, because it confirmed there was no hardware fault hiding underneath the filesystem damage and meant the image could be taken in one straightforward pass behind a write blocker.
Working on the image, R-Studio found the primary index damaged across a run of records and the backup copy intact. The tree was rebuilt from both, and the journal was read to establish what the interrupted backup had been doing at the moment the plug came out, which identified the four files that were genuinely mid-write. Everything else was reconstructed by following its own fragment list. A folder from each of the fifteen years was opened before anything was written out.
The accounts, the field records and the scanned paperwork all came back, arranged as they had been on the original rather than flattened into one directory. The four files caught mid-write were the last four of the backup and were duplicates of material that already existed elsewhere on the volume, so nothing was actually lost.
One drive is £300 + VAT. The lid stayed on and nothing was lifted off a circuit, so no fix, no fee applied and the invoice did not follow until the bookkeeper had opened the accounts.
Unplug it and leave it unplugged, because every mount is another chance for something to write to a volume that is already inconsistent. Decline the offer to scan and repair. That tool exists to make a disk usable again, and its idea of usable includes throwing away records it cannot parse and dumping loose fragments into a folder of numbered files, which is a much harder starting point than the damage you have now. Do not run a defragmenter, do not create a new partition, and do not accept an offer to initialise the disk. If the material is a business's accounts, the cheapest thing you can do is stop and get an image taken while the second copy of the index is still intact. Send the unit whole if it is an external in its own enclosure, with the cable, and say what the drive was doing at the moment it stopped, because that single detail often identifies which files are genuinely affected.
What this job cost, and on what terms. This one was charged at £300 + VAT, the figure agreed in writing before any work began. Paying for a job like this one. Nothing here called for the disk to be opened, a board rebuilt or a chip lifted off, which puts the work on the logical side of the line and under no fix, no fee. Files back, invoice follows; nothing back, nothing owed, and the media goes home either way. Four sorts of work sit outside that promise and only those four: DVR jobs, forensic jobs, chip-level work, and electronic or mechanical failure. The rest keep it, firmware and service-area repair included — both are logical jobs whatever their reputation suggests.
If yours is doing the same thing. This drive was read at Oxford Data Recovery on the Oxford Science Park, about fifty-five miles down the M40. Post it tracked and insured, send it by a courier you book yourself, or bring it to the Oxford reception during Mon–Fri 9:00am–5:30pm — there is no collection service and no counter in Coventry. The free look at it takes two working days from booking in and produces one written figure that then holds.
Freephone 0800 689 0668 to talk it through first. The shipping form and what goes in the parcel are on the contact page, and every band is published in full — including which work sits outside no fix, no fee.
These are real enquiries, from households and from businesses in Coventry and out across Warwickshire — Nuneaton and Bedworth, Rugby, Leamington Spa, Warwick, Kenilworth, Solihull, Stratford-upon-Avon, Atherstone and Southam. Names are gone: the customer, the company, the job title, anything at all that would point back at a person. What stays is the reasoning — what turned up, where the fault genuinely sat, which approach it demanded, and which machine on the bench did the work. One of the thirty-six was refused, and it earns a place here as squarely as any of the rest.
Enumerating and reading are two different jobs, and only one of them needs the platters to turn
A broken socket and a broken filesystem look equally fatal from a desk and are neither
Whirring and a tone together describe a mechanism working hard and getting nowhere
A drive Windows calls corrupted and unreadable? Ring 0800 689 0668 before you repair anything. One disk is £300 + VAT and the diagnostic is free.