Data Recovery Case File · NAS & Network Storage · The Disks Remember

The Machine Has Gone and the Array Has Not

His enquiry describes a situation that sounds hopeless and generally is not. Three 500GB drives "that were originally set up as a RAID in a Windows PC. The PC failed and is no more, but I have retained t"he disks. The instinct is that an array without its computer is an array without its configuration — that whatever knew how the disks fitted together died with the machine. In almost all cases it did not. The description of the set is written on the members themselves, which is precisely so that they survive the host, and keeping all three was the right thing to do.

The decode: where the configuration actually lives

Why the disks know what they were: when an array is created, each member has a small block of metadata written to it at a known position — recording that the disk belongs to a set, which member it is, how many members there are, the stripe size, the layout, and a signature identifying the set. This exists so that the array can be reassembled after a motherboard failure, an operating-system reinstall, or a move to different hardware. It is on the disks by design, not by accident.

What that means for him: with all three members retained, the description of the set is retained too. The array can be reconstructed from the disks alone, without the original machine, the original controller or the original installation. That is the entire point of storing it there.

What still needs establishing: what type of array it was, which the metadata will say. Three disks most commonly means either striping without redundancy — where every file is spread across all three and all three are required — or single-parity, where any one member can be missing and reconstructed from the other two. That distinction decides how much tolerance there is if one of the disks turns out to have a fault of its own, which is worth knowing before anything is attempted.

Why the drives are imaged before anything else: these disks are contemporaries. They were bought together, ran together and aged together, so if one is failing the others may not be far behind — and any hypothesis about geometry should be tested against copies rather than against the only originals in existence.

Why geometry is validated rather than trusted: metadata can be stale or partially damaged, particularly on disks that have sat unused. So the parameters it reports are tested by assembling the set and checking whether file content reads coherently — a correct stripe size and member order produces documents that open and media that plays, and an incorrect one produces nothing. That check is what separates a reconstruction from a guess.

The two things that must not happen: no disk should be connected to a machine that offers to initialise it, because accepting writes a fresh partition structure over the start of a member — exactly where the array metadata lives. And nobody should attempt to re-create the array using disk management or a controller utility, because creating an array writes new metadata across all members and destroys the description being relied on. Re-creating and reassembling sound like the same thing and are opposites.

On the bench

All three members were imaged individually write-blocked on the Atola TaskForce 2 before anything was attempted, since contemporaneous disks age together and every hypothesis should be tested against copies. The array metadata was read from each member — recording set membership, member order, stripe size and layout, which is written to the disks precisely so a set survives its host. Geometry was then validated against file content rather than trusted, and the set assembled offline from the images.

The outcome

Every member imaged individually, the configuration read from the disks themselves and the geometry validated against real content before assembly. Free assessment, one fixed written figure including VAT; where a drive has to be opened, 50% of parts and labour is payable upfront with the balance only on success — otherwise no recovery, no fee. The decode, for anyone holding array disks without their computer: the configuration is on the members, not in the machine. Each disk carries metadata recording the set, its position in it, the stripe size and the layout, written there so an array survives a dead motherboard or a reinstall. Keeping all the disks was the right call. Never let anything initialise one, and never re-create the array — creating writes fresh metadata over the description you need.

Our case files are drawn from genuine enquiries received by our laboratory over the past ten years, anonymised to protect client confidentiality. Each one describes the diagnostic and recovery procedure our engineers apply to that fault, using the equipment listed.