An HPE ProLiant running RAID 10 suffered three disk failures at once — the kind of event that layout is supposed to survive, right up until it doesn't. Whether the data lived or died came down to which disks failed. We imaged every member, repaired the failed ones, and reconstructed the array.
The ProLiant hosted the company's databases and shared files on a RAID 10, and had been running with warnings that went unactioned when three of its disks failed together and the array collapsed. Backups were incomplete. Faced with multiple simultaneous failures, the only safe move was to stop, remove every disk, label each with its slot, and bring them in — an on-controller rebuild attempt with three failed members could have destroyed what was still recoverable.
RAID 10 combines mirroring and striping: disks are paired into mirrors, and data is striped across those pairs. It can survive losing a disk from each mirror — even several disks at once — but there's a catch. If both members of the same mirrored pair fail, the stripes held only on that pair are gone. So three failures might be survivable or catastrophic depending entirely on how they fall across the pairs. The first job was to establish exactly which disks belonged to which mirror, and whether any single pair had lost both members.
Each disk was imaged. The surviving members were cloned in full; the three failed drives were assessed and repaired in controlled clean-air conditions — head and electronics faults addressed — then imaged adaptively to recover as much as possible. Mapping the pairs against the images showed that no single mirror had lost both members beyond recovery: every pair had at least one readable copy, once the failed drives were imaged, which meant the full stripe could be reassembled.
From the images, the RAID 10 was rebuilt: the mirrored pairs reassembled, then the stripe across them reconstructed with the correct stripe size, order and offset, confirmed against known file-system structures. For each pair, whichever member read most cleanly supplied the data. That reassembled the volume, and the databases and shared files were recovered with their structure intact.
The databases were validated and files opened across the shares to confirm they were intact, then everything was returned on fresh storage. About 98% came back — a strong result for a three-disk failure, and one that hinged on repairing the failed drives well enough to image them. We stressed the point plainly: a resilient array is not a substitute for acting on failure warnings or keeping a real backup — the redundancy had quietly been used up before anyone noticed.
Clean-air physical repair of failed members · adaptive imaging · RAID 10 mirror-pair and stripe reconstruction with PC-3000 RAID and Atola Insight · database validation. Physical work and imaging carried out in-house in Belfast.
Send it to us for a free, no-obligation diagnostic. We’ll tell you what can be recovered and put a fixed price in writing before any work starts — and on most jobs, if we can’t get your data back, there’s nothing to pay. Post your device in, or drop it to us by appointment.
Often, yes — it depends on how the failures fall across the mirrored pairs. As long as each pair has at least one member that can be imaged, the array can usually be reconstructed. Stop, remove all the disks labelled with their slots, and send them in — don't attempt a rebuild.
RAID 10 tolerates losing a disk from each mirror, but not both members of the same pair. Warnings often go unnoticed, so redundancy gets used up quietly — by the time an array collapses, several disks may already have failed.
RAID and server recoveries start from £500 plus VAT, with a fixed written quote after a free diagnostic. Physical repairs carry a deposit toward parts and bench time.