A design agency in Hammersmith runs everything off a two-disk striped enclosure on the studio network: live jobs, artwork, the archive of everything delivered in the past six years. On Monday it mounted as usual, at the right size, with an empty root. No error, no warning, no missing disk. Just a volume with nothing in it.
| Media | A two-disk striped set in a desktop enclosure, holding a design studio's live jobs and six years of delivered artwork. The volume mounts at its correct size and presents an empty root directory. |
| What was reported | Two-disk external RAID enclosure, mounts on the studio machine at the correct capacity, no files or folders visible at all, no drive reported as failed, six years of artwork wanted, recovery sought. |
| Fault class | Logical, above the array. Both members were healthy and the stripe geometry was intact, which is why the volume mounted at the right size. What had been damaged was the filesystem's master directory, so the volume could be measured and not described. |
| On the bench | Both members removed from the enclosure and attached separately behind hardware write blockers on PC-3000 UDMA, profiled and imaged individually on DeepSpar with no rebuild permitted at any point, the stripe reassembled virtually from the images with geometry confirmed against filesystem structures, the damaged master directory rebuilt in R-Studio with UFS Explorer used as a second reading, artwork carved by content signature where entries were unrecoverable, everything validated and written to fresh media. |
Mounting at the correct size tells you a great deal. To present a volume of the right capacity, the enclosure has to have read both disks, agreed on the stripe order and assembled the whole thing. All of that worked. The failure is one layer higher, in the filesystem sitting on top of the assembled volume.
An empty root is a directory problem, not a data problem. The master directory is the map. Damage it and every file becomes unreferenced simultaneously, which is why a studio's entire history disappears in one moment rather than in pieces. The artwork itself is untouched, spread across both members exactly where it was written.
Striping means neither disk is useful on its own. Every file is cut into blocks that alternate between the two members, so a single disk holds fragments of everything and the whole of nothing. Both have to be imaged, and the reassembly has to be right before the filesystem can even be examined.
The dangerous option was the one the enclosure offered. Boxes like this present a repair or a rebuild function when a volume looks wrong, and both write to the members. On a set whose disks are healthy and whose directory is damaged, that is the fastest available way to turn a recoverable job into a much worse one. They did not use it, which is why this went well.
Both disks came out of the enclosure and were worked separately, each behind its own hardware write blocker, and neither was ever put back into a box that might have offered to help. Both profiled as healthy with short defect lists and full read speed, and both were imaged on DeepSpar.
The stripe was reassembled virtually from the two images, with the block size and member order confirmed by checking that filesystem structures resolved across the stripe boundaries rather than by trusting the enclosure's labelling. R-Studio then rebuilt the damaged master directory from the surviving records and the volume's own backup structures, with UFS Explorer used as an independent second reading to check the result. Artwork whose entries could not be rebuilt was carved by content signature, and files from every year were opened before anything was written to fresh media.
The whole archive came back with its folder structure intact, which for a studio that files by client and by job number mattered nearly as much as the files themselves. It went back on a single large drive so it can be read without an enclosure.
Multi-disk sets are quoted from £500 + VAT, and this was one: two members, imaged separately, reassembled in software. No disk was opened and no part was replaced, so the work was logical throughout and no fix, no fee applied.
Unmount it, power the enclosure down and stop there. Do not accept a repair offer, do not start a rebuild and do not remove a disk to test it in something else, because all three of those write to members whose contents are currently intact. Do not run a filesystem check either: on a striped set a check that guesses wrongly can discard directory records across both disks at once. Label the bay order if the disks have to come out, though the order is recoverable without labels. Send every member, including any that look fine, because the volume only exists when they are read together.
These files are written up from genuine enquiries taken from households and businesses in Bracknell, across Berkshire and Surrey and into London, with names, firms and anything else identifying stripped out. Each one sets down how the fault was reasoned through, the method it actually needed and the equipment it was worked on. One of them is a job that was turned away, which belongs here as much as the rest of them.
Examining it is free and takes 2 working days from arrival, the figure goes to you in writing and does not move afterwards, and a logical fault carries no fix, no fee. Start online or ring the freephone.