Case file · NAS & RAID · SDR-2025-0341
Two Members Gone From a Four-Bay Array.
A four-bay QNAP running RAID 5 had lost two of its disks, and the same description fitted both — both have failed mechanically and they click
. The plan put to us sounded practical enough: repair one of them so the RAID 5 can rebuild
. It was also precisely the plan to argue somebody out of.
Seeing the same thing yourself?
0800 6890668
The translation.
Single parity buys you one absent disk. Two defeats it, because the stripe no longer holds enough information to solve for what is missing. The suggested route made things worse again: repairing one disk and letting the box rebuild would drag a complete read across two tired survivors while laying down new parity across precisely those stripes reconstruction has to read in the state the failure left them. Copies first, or nothing at all.
Kit used on this job.
What happens in a case →| Platform | Its role in this case | Why we use it |
|---|---|---|
| DeepSpar Disk Imager 4 | Donor head stacks fitted to both clickers, then imaged head map first | Head-map-driven imaging, with per-head timeouts, resets and power-cycle control |
| Atola TaskForce 2 | Ran the surviving pair through imaging alongside the bench repairs | Images several drives in parallel — it turns a week on an array into a few days |
| UFS Explorer RAID Recovery | Worked out the stripe layout and built the volume across all four copies | Understands how NAS volume managers really work, instead of a flat array |
In the lab.
Copy all four disks, the survivors included
Nothing was left uncopied, failed or otherwise. A virtual rebuild takes each block from whichever image holds it best, and disks that look healthy inside a degraded array are very often nursing bad sectors of their own. Putting the pair through imaging while the mechanical repairs ran also shortened the job considerably.
Matched heads for both casualties, the worst bands left till last
Each failed member received a matched head stack and was read in head-map order. Damage on both was patchy rather than total, which is the usual picture — most of the surface came off without complaint and the difficulty sat in particular bands. Those bands waited until everything else was safe.
Assemble the array virtually, using only the images
Stripe size, disk order and the direction parity rotates were read out of the on-disk metadata instead of guessed at, and the volume was then put together across the four copies. Any block reading poorly on one image could be taken from another, which is the flexibility a live rebuild will never give you.
The result.
The assembled volume mounted, the shares were checked off against the owners' own inventory, and the QNAP went back into service on a fresh set of disks — restored from copies rather than from optimism.
Similar cases on the index.
Also from NAS & RAID.
Does this sound like your own drive?
The rule holds as in every case above: switch it off, and let a free diagnosis come before any decision.