Data Recovery Case File · NAS & Network Storage · The Copy That Should Have Held
Both Members Showing Partitions Means the Data Survived and the Controller Disagrees
His enquiry describes the failure a mirror exists to prevent. A two-disk mirror on a server where "one disk failed and now the virtual disk won't come online" — and, importantly, "on a disk reader both disks show partition information but we cannot access the file data." That last observation is the case: the content is present on both members, and what has failed is the controller's willingness to present it.
| Media | Two-disk mirrored array of 1TB on a server platform — one member reported failed; logical volume not presenting; both members showing partition structures on direct connection |
| Reported situation | Two-disk mirrored array in service · one member reported as failed by the controller · logical volume not coming online · both members connected to a separate reader · partition structures visible on both · file content not accessible · recovery, cost and turnaround queried |
| Fault class | Array metadata state preventing presentation — member content substantially intact on both disks; filesystem accessible once assembled outside the controller |
| Equipment used | Controller refusal distinguished from data loss before any action at the controller · no rebuild, import or initialisation permitted · both members imaged individually write-blocked under strict per-sector timeouts · mirror metadata read and the healthier member identified by comparison · volume assembled offline from the images and the filesystem interpreted there |
The decode: why a mirror can fail to present despite holding two copies
What a mirror does in normal operation: writes everything twice. Each member holds a complete copy, which is the entire point of the arrangement.
Why losing one should therefore be survivable: the other is complete. A mirror is designed to continue on a single member.
Why it did not continue here: the controller records state. Alongside the data, each member carries metadata describing the array and its condition, and the controller decides from that whether the set is usable.
What commonly goes wrong with it: the state becomes contradictory. A member marked failed while the other is marked out of date leaves the controller with no member it considers authoritative, so it presents nothing.
Why that is a refusal rather than a loss: the data is untouched. The controller is declining to present content that is physically present on both disks.
Why his observation confirms exactly that: partitions are visible. Structures readable on a direct connection prove the members hold a formed volume.
Why the file content is still not reachable that way: the volume is not assembled. A host connected directly sees a member of an array rather than the array, and the arrangement has to be interpreted before a filesystem appears.
Why one member is likely better than the other: they diverged. Once a member was marked failed, writes continued to one and not the other, so the surviving member is more current and is identified by comparison.
Now the action to refuse absolutely: importing, rebuilding or forcing the array online. A rebuild copies one member over the other, and if the controller chooses the stale one the current copy is destroyed.
Why that risk is the reason for imaging first: the choice becomes reversible. With both members imaged, either can be assembled and compared without consequence.
On the bench
Controller refusal was distinguished from data loss before any action at the controller — a mirror writing every block to both members, so each holds a complete copy, while metadata describing array condition determines whether the controller presents the set. Contradictory state leaves no member considered authoritative and nothing is presented, though partitions visible on direct connection prove a formed volume persists. Both members were imaged and the healthier identified by comparison before offline assembly.
The outcome
Controller refusal distinguished from data loss, no rebuild or import permitted, and both members imaged before the volume was assembled offline. Free assessment, one fixed written figure including VAT, charged per drive, with 50% of parts and labour upfront where a drive has to be opened. The decode: both disks showing partitions is the finding. The content survived on both members, and what failed is the controller's record of which one to trust — which is a refusal, not a loss.
A mirror that will not come online after losing a member
Refuse to import, rebuild or force the array online. A rebuild copies one member over the other, and if the controller picks the stale one it destroys the current copy — which is the real risk here, because once a member was marked failed the two diverged. Read your own observation as good news: partitions visible on a direct connection prove both disks still hold a formed volume, so what's failed is the controller's record of which member to trust rather than the data itself.
Don't rebuild it — call Cambridge Data Recovery on 01223 655015; controller refusal distinguished from data loss, both members imaged first, healthier member identified by comparison and 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.