Specialist · RAID 5 data recovery
RAID 5 survives exactly one failure — and by the time most arrays reach us, that allowance is gone and a rebuild has been tried. The good news the panic hides: 'two drives down' rarely means two drives dead, and everything recoverable is reachable from images without risking a single further write.
Talk it through with an engineer
01273 964902
| The situation | What it means | First move |
|---|---|---|
| One drive failed, array degraded | Running with zero redundancy — every hour of use is unprotected | Power down before the second |
| Rebuild started, then failed | The full-surface read found bad sectors on a survivor — the classic ending | Stop; don't retry the rebuild |
| Two drives showing failed | Often one dead drive plus one with a handful of bad sectors — very different from two corpses | Send both; images decide |
| Drives forced back online | Stale members injected into live stripes | Stop exactly there |
| Array vanished after a controller swap | Metadata mismatch, data intact underneath | Reconstruction territory |
| Someone ran 'RAID recovery' software on the live array | Writes on evidence | Honesty first: send it anyway |
Your device is logged with its own case reference the moment it arrives. An engineer assesses the fault, confirms what's actually recoverable, and you get a fixed price in writing — no diagnosis fee, no obligation, and no paid work until you say go.
Each drive — including both 'failed' ones — is imaged individually with bad sectors mapped and weak zones deferred. From here, nothing that follows can make the situation worse.
Drive order, stripe size and parity rotation are established from the data itself and the controller's on-disk metadata — no working controller needed, no guessing tolerated.
The array is reassembled in software across the images, taking each block from whichever member holds it healthiest — including regions only the 'failed' drives still carry. Then the file system on top is repaired and verified.
Before you pay the recovery fee you approve a full listing of what came back. Your data returns on new media with free return postage, and the case only closes once you've confirmed everything opens on your side.
The number worth actually knowing: consumer drives spec one unrecoverable read error per 1014 bits — about 12.5TB — and enterprise drives 1015. A four-drive 16TB RAID 5 rebuild reads ~48TB from its survivors, so the spec says the risk is real; field data says most drives read far past their floor clean. Both things are true — which is why the professional answer isn't 'never rebuild', it's 'never rebuild before imaging'.
Drive two died in June; nobody noticed until drive four followed in August, mid-rebuild. Images told the truth: drive four had nineteen bad sectors, not a failure. The stripe rebuilt from three-and-a-nearly-whole members, and the practice filed on time with everything.
Very often, yes. 'Failed' by a controller's standards frequently means a few unreadable sectors, not a dead drive — and reconstruction from images can take each block from whichever member reads it best. Two genuinely destroyed drives is the rare case; two dropped drives is the common one.
Not before the members are imaged. A rebuild reads every sector of every survivor — the exact stress that finishes marginal drives — and writes parity as it goes, overwriting what reconstruction would otherwise use. Image first turns the same rebuild into a zero-risk operation.
An unrecoverable read error — a sector a drive's own correction can't rescue, specified at roughly one per 12.5TB read on consumer drives, one per 125TB on enterprise. Respect it; don't be terrorised by it. The spec is a worst-case floor, and modern controllers lose a stripe to a URE, not the array.
Yes — the geometry lives on the drives themselves in the controller's metadata, and where that's damaged it can be derived from the data patterns. Arrays reconstruct on our bench from images alone.
Every restart of a damaged device costs data. Open a case first — the diagnosis is free either way.