Devices · RAID arrays & servers
Most RAID disasters aren't the first drive failing — they're what happens next: the rebuild that stresses a second marginal drive to death, the forced-online gamble, the wrong disk pulled. We work only on images, so the array can't get worse in our hands. Brighton servers and RAID sets travel as bare disks labelled by bay, never as a loaded chassis.
Talk it through with an engineer
01273 964902
| The message | What it means | First thing we do |
|---|---|---|
| Foreign Configuration Found (Dell PERC) | The controller sees array metadata that doesn't match its memory | Neither Import nor Clear — image |
| 1786 — Drive Array Recovery Needed (HP) | A rebuild is needed or was interrupted; redundancy is compromised | Stop-and-image event |
| 1784 — Drive Array Drive Failure (HP) | A physical member has failed | Don't hot-swap and pray |
| 1788 — Slot X Drive Array Reports Incorrect Drive Replacement (HP) | Wrong bay for the new drive, or the wrong member came out | Order matters — stop |
| Virtual Drive is Degraded / Offline | Running without redundancy, or not running at all | Power down the array |
| 1720 — SMART drive detects imminent failure (HP) | A member is pre-failing inside a live array | Plan the imaging now |
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 the 'failed' ones — is imaged individually on hardware tools. All work happens on copies; your original disks are never written to.
Drive order, block size and parity rotation are worked out from the data itself, and the array is reassembled in software across the images — no controller required, no rebuild risked.
File systems, virtual machines and databases are repaired on the re-assembly, and verified against a listing before anything is signed off.
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 real number behind the RAID 5 warnings: consumer SATA drives are specified at roughly one unrecoverable read error per 1014 bits — about every 12.5TB read. A big RAID 5 rebuild reads more than that from its survivors, which is why single-parity arrays fail their own rebuilds. It's a datasheet figure, not folklore — though how doomed that makes RAID 5 is still argued.
Production data trapped on a server that had eaten its second drive mid-rebuild on a Friday. All four disks imaged, the stripe reconstructed from three-and-a-bit good images, and the factory reopened on Monday morning.
Yes, send every member. 'Failed' drives often hold the freshest copy of some regions, and re-assembly chooses the healthiest source for each block across all of them.
No — the correct order, block size and parity rotation are recoverable from the data patterns on the disks themselves. It's routine analysis, not guesswork.
Yes. Once the array is rebuilt we recover VMDK, VHDX and database stores as first-class targets and verify they mount, not just that they exist.
Multi-drive jobs typically run four to seven working days; genuine business-down emergencies can be worked continuously — say so when you call and we'll plan the intake around it.
Every restart of a damaged device costs data. Open a case first — the diagnosis is free either way.