A six-disk RAID 5 went offline after a power cut, with one member failed and only two of the remaining five recognised. The client used the controller’s built-in recovery function. It reported success — and then Windows showed six unallocated disks and offered to format them.
A blackout took the server down. On restart the array was offline: one drive had failed outright, and of the five remaining the controller recognised only two as belonging to the set.
The client went into the RAID management utility and used its recovery option. It reported success. When Windows loaded, all six disks showed as unallocated space and it offered to format them — which they declined, and rang instead.
This is the trap, and it catches experienced people because the wording is genuinely misleading.
On most controllers, the options labelled recover, restore, repair array or reconstruct do not examine your disks and fix what is wrong with them. They re-create the array definition — writing fresh metadata across the members describing a new array.
From the controller’s point of view that is a successful outcome. There is now a valid array where previously there was a broken one, and it will say so.
From your point of view it is the opposite of what you wanted. The new definition rarely matches the old one exactly — member order, stripe size, the offset at which data begins, whether parity rotates left or right — and the moment it is written, the record of how your data was actually arranged has gone.
Which is exactly why the disks appeared as unallocated. The controller had assembled a technically valid array using the wrong parameters, so the file system that should have been at the start of the volume was no longer where anything expected it to be.
There is no undo. If a controller offers to recover, restore or re-create an array holding data you need, the safest response is to leave the utility entirely and power the machine down.
Because re-creating an array writes metadata, not data.
Several hundred kilobytes of descriptors were replaced at fixed positions on each member. The remaining terabytes — the actual files, the file-system structures, the parity blocks — were untouched. The map had been overwritten; the territory had not.
So the job became one of disregarding the controller entirely and working out the original geometry from the data itself. That is routine when nothing else has been written since.
What would have made it far worse is what the client was invited to do next. Formatting writes across the volume. Letting a rebuild run under the new definition is worse still: it recomputes parity according to the wrong arrangement and writes the results across every member, which converts a metadata problem into a data problem, permanently.
All six drives were imaged read-only. The failed member had a head fault and was opened under filtered air and given a matched donor assembly before imaging — in a RAID 5 with one member already absent, the array has no redundancy left, so that disk’s contribution is not optional.
The original geometry was then derived from the data rather than from any metadata. Candidate combinations of stripe size, member order, parity rotation and data offset were applied to the images, and each tested two ways: whether recognisable file-system structures resolved at the offsets they should occupy, and whether parity computed correctly across stripes sampled from across the whole volume.
That second test is what makes the answer certain rather than plausible. Parity either resolves across the set or it does not, and a wrong arrangement fails immediately and everywhere.
The descriptors the controller had written were disregarded entirely. With the array assembled virtually from the images, the file system was rebuilt and the business data extracted and verified.
Everything. Nine days, most of it imaging six drives and repairing the failed member rather than the reconstruction itself.
Two decisions shaped this job. Using the controller’s recovery option cost time and money and turned a straightforward reassembly into a derivation exercise. Declining the format prompt afterwards is why there was still a job to do at all. RAID and server recovery is from £500 +VAT after a free 48-hour diagnostic.
Post it or bring it in. Either way the diagnostic costs nothing and commits you to nothing, and you get one figure in writing before a single screw is turned.
Nothing can begin until the device is on the bench, which makes this the only step that needs anything from you. Pack it properly, put your details in with it, and send it over. What follows is a free diagnostic and a written figure, in that order.
Posting it? Use something tracked and insured — whatever is on the drive is worth considerably more than the postage. Bringing it in? Weekdays, 9am to 5:30pm, and it still wants packing as above for the journey.
Tell us what the device is, what it is doing, and anything already tried — an engineer will come back with a realistic view of the odds and a price band, before you commit to posting anything.
We’ll get back to you soon. Anything pressing, call 0161 871 0788.
Not on an array holding data you need. On most controllers those options re-create the array definition rather than repairing it — writing new metadata over the record of how your data was actually arranged. The controller reports success; you lose the map.
Almost certainly not. Re-creating an array replaces a small amount of descriptor metadata, not the terabytes of actual data beneath it. The files, the file system and the parity are usually untouched — what has gone is the description of how they fit together.
No. That is the single most damaging thing available at that point. Formatting writes across the volume rather than a corner of it, and it destroys the structures a recovery would use to rebuild the array.
Yes — stripe size, member order, parity rotation and offset are all derived by analysing the disks themselves. Candidate arrangements are tested until reassembly produces coherent file system structures rather than noise, and the wrong ones fail immediately.
It can be recovered and it is worth doing. RAID 5 tolerates one missing member, so the array can often be assembled without it — but repairing that drive and imaging it as well gives a complete set and removes any dependence on parity to fill gaps.
From £500 plus VAT, fixed in writing after a free 48-hour diagnostic. Where a member needs clean-air work there is a 50% deposit with the balance due only on success.
Kick off with an instant online quote, or ring us and talk it through first. Either way you’ll know a clear, fixed price before any work starts.