Data Recovery Case File · NAS & Network Storage · The Refusal Is Protective
Be Glad It Declined
His enquiry treats the refusal as part of the problem, and it is worth turning around. A four-bay array enclosure "which has gone into an error state. It will not automatically rebuild. We have tried to contact the manufacturer for guidance but their support is not very good." He wants it to rebuild and it will not. That refusal may be the most useful thing the unit has done — because an array declining to reconstruct is usually an array that has worked out it cannot do so safely, and forcing it is how these cases are lost.
| Media | Multi-bay hardware array enclosure in an error state — automatic reconstruction not proceeding; vendor guidance unavailable |
| Reported situation | Array entering an error state · automatic rebuild not commencing · manufacturer support approached without useful guidance · members retained in the enclosure · data required |
| Fault class | Array unable to establish a coherent member set — reconstruction contraindicated; vendor-specific metadata requiring format-aware assembly |
| Equipment used | No forced rebuild or array initialisation attempted · all members removed and labelled by bay · each imaged individually write-blocked (Atola TaskForce 2) · vendor-specific array metadata interpreted rather than assumed standard · failure sequence established · volume assembled offline from images |
The decode: what a refusal means, and why forcing it is the danger
What has to be true for a rebuild to be safe: exactly one member missing or failed, and every remaining member readable and in agreement about the array's state. Each disk carries metadata describing the configuration and a sequence number showing how current it is. The controller compares them, and it will only reconstruct when the surviving set is coherent and complete enough to compute the missing member from.
So what a refusal is actually reporting: that this condition is not met. Most often that means more than one member is questionable — a second disk erroring, or dropped out earlier and now stale. Sometimes the metadata disagrees, because members were removed and reinserted in a different order, or one was offline for a period and holds an older view of the array.
Why forcing it is genuinely dangerous: a rebuild writes. It computes the missing member's contents from the others and writes the result across a disk. If the surviving set is not coherent — if a stale member is being treated as current, or a second disk is returning bad data — then what gets computed is wrong, and it is written permanently over the array. A forced rebuild on an incoherent set can destroy data that was entirely recoverable. That is why the refusal deserves to be read as protection rather than obstruction.
Why the sequence matters and must be established: knowing which member failed first, and when, is what allows a stale disk to be excluded from a reconstruction. A disk that dropped out weeks ago holds a snapshot of the array from weeks ago, and mixing it with current members silently produces a volume that mounts and contains corrupted files.
The point specific to hardware enclosures: the metadata is vendor-defined. Connecting the disks to a computer and using general-purpose array software will not read it — a host will offer to initialise them, and accepting that destroys the layout information the assembly depends on. The disks must not be put into a PC to investigate.
On the vendor support being unhelpful: frustrating and not a reason to improvise. The absence of guidance is precisely when the temptation to force something is strongest, and it is the worst moment to act on it.
What preserves everything: removing all members, labelling by bay, and imaging each individually so that the assembly happens offline against copies, where a wrong hypothesis costs nothing.
On the bench
No forced rebuild or array initialisation was attempted — a reconstruction computes the missing member from the others and writes the result permanently, so forcing it on an incoherent set destroys recoverable data. All members were removed and labelled by bay, then imaged individually write-blocked on the Atola TaskForce 2. Vendor-specific array metadata was interpreted rather than assumed standard, the failure sequence established so stale members could be excluded, and the volume assembled offline from the images.
The outcome
The rebuild refused rather than forced, every member imaged individually and the volume assembled offline from copies. 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 whose array will not rebuild: read that as protection. A controller only reconstructs when the surviving members are coherent and in agreement, so a refusal usually means a second disk is questionable or the metadata disagrees. Forcing it writes computed data permanently across the array — and if the set is incoherent, what gets computed is wrong.
Array in an error state that won't rebuild
Don't force it, and don't put the disks into a PC to see what's on them. That refusal is almost certainly protective. A controller will only reconstruct when exactly one member is missing and every remaining disk is readable and agrees about the array's state — each carries metadata with a sequence number showing how current it is. So declining usually means a second member is questionable, or one dropped out earlier and now holds a stale view. Forcing a rebuild computes the missing disk's contents from the others and writes that permanently across the array, so on an incoherent set what gets written is wrong, over data that was recoverable. Hardware enclosures also use vendor-defined metadata that general software can't read, and a host will offer to initialise the disks.
Don't force it — call Newcastle Data Recovery on 0191 406 1051; all members imaged individually write-blocked, vendor metadata interpreted specifically, failure sequence established and the volume assembled offline.
Request a quote online →
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.