Data Recovery Case File · NAS & Network Storage · The Originals Are the Data
Replacing Two Members of a Single-Parity Set Leaves Nothing for It to Rebuild From
His enquiry describes an action taken in good faith and one decision that saves it. A four-drive unit where "two hard drives failed and were replaced with new ones", after which the system reads nothing at all — and, crucially, "I still have the original drives." A set that tolerates one failure cannot survive two, so the new drives have nothing to reconstruct from; the old ones are where his files still are.
| Media | Four-bay network unit in a single-parity configuration — two members replaced following reported failure; original members retained; array no longer presenting content |
| Reported situation | Four-drive domestic network unit · two member drives reported as failed · both replaced with new drives · unit no longer reading files at all · original member drives retained by the owner · reconfiguration of the unit proposed by the owner · content required |
| Fault class | Member replacement beyond parity tolerance — reconstruction impossible from remaining members; original drives constituting the sole source of content |
| Equipment used | Original members identified as the sole source before any unit operation was considered · no reconfiguration or initialisation permitted on the unit · all four original members imaged individually write-blocked · array parameters and member order derived from the members themselves · array assembled offline and filesystem interpreted from the images |
The decode: what the replacement did, and why the originals matter so much
What single parity provides: the ability to lose one member. Information is distributed so that any one drive's contents can be calculated from the other three — real protection, and the commonest arrangement in units of this size.
What it does not provide: tolerance of a second. With two members missing there is not enough information left to reconstruct either, and the mathematics simply runs out.
Why replacing both produced silence rather than a rebuild: a rebuild calculates a missing member from the survivors. Two blank drives and two survivors do not give the unit enough to work with, so it has nothing to present.
Why the replacement was nonetheless the obvious thing to do: the unit reported two failures and replacement is what one does with failed drives. Nothing in the interface explains that replacing the second is different in kind from replacing the first.
Why the originals being retained changes everything: the content never left them. Data written to an array lives on its member drives, and removing a drive from a unit does not erase it — so four original members hold the complete set between them.
Why the reported failures may also be less than they sound: units eject members on timeouts and error thresholds rather than on diagnosis. A drive marked failed is frequently substantially readable, and in a four-drive set even partial readability from a marked member can complete the picture.
Why the request to reconfigure the unit must be declined: that is the one action that would end this. Reconfiguring writes fresh array metadata and, on most units, initialises the members — including any originals refitted for the purpose.
Why the temptation to refit the originals is so strong: it looks like putting things back as they were. A unit shown four drives it does not recognise as a set will offer to make them one, and accepting is destructive.
What is done instead: all four original members imaged individually, and the array assembled offline from those images with the parameters derived from the drives themselves. Nothing is attempted in the unit at all.
Why the new drives are irrelevant to the recovery: they were blank when fitted. Whatever the unit wrote to them since is a partial or failed rebuild, and nothing on them contributes.
On the bench
Original members were identified as the sole source before any unit operation was considered — single parity permitting reconstruction of one member from the remainder but not two, so replacing both leaves insufficient information and the unit presents nothing. Content written to an array resides on its member drives and is not erased by removal, so the four originals hold the complete set. Units eject members on timeout and error thresholds rather than diagnosis. No reconfiguration was permitted.
The outcome
Original members identified as the sole source, no reconfiguration permitted on the unit, and all four imaged individually with the array assembled offline. Free assessment, one fixed written figure including VAT, charged per drive, with 50% of parts and labour upfront and the balance only on successful recovery. The decode: keeping the originals is what saves this. Your set could lose one member and not two — so the new drives have nothing to rebuild from, and your files are still on the drives you took out.
After replacing more than one drive in an array
Keep the original drives and don't refit them to have the unit reconfigured — a unit shown drives it doesn't recognise as a set will offer to make them one, and accepting writes fresh metadata and generally initialises them. Understand what happened: single parity lets the unit calculate one missing member from the others, and with two gone there isn't enough information left, so a rebuild has nothing to work from. Your content never left the originals. And a drive the unit marked failed is often substantially readable, since units eject on timeouts rather than diagnosis.
Keep the originals — call Manchester Data Recovery on 0161 871 0788; originals identified as the sole source, no reconfiguration permitted, all members imaged individually 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.