Data Recovery Case File · NAS & Network Storage · Not Proportional

Losing a Quarter of the Array Is Not Losing a Quarter of the Files

His enquiry asks for a partial outcome that this configuration does not offer. One failed disk in a four-drive stripe: "the faulty drive does not show in the firmware. I would be very interested in the probability of mending this drive to allow the array to work again, or the salvage of some of the files." His instinct to mend the drive is exactly right — and the alternative he offers as a fallback is not available, because striping does not lose files, it loses parts of all of them.

The decode: what a stripe does to a file

How the data is arranged: every file is cut into pieces and written across all four drives in rotation — a chunk here, the next chunk there, and so on. No drive holds a complete file of any size beyond the smallest, because the arrangement exists to have four devices working simultaneously.

What losing one member actually costs: not a quarter of the files. A quarter of every file, distributed evenly through its length. A document is missing every fourth chunk from beginning to end; a photograph is missing every fourth chunk; a video is missing every fourth chunk. Nothing is left whole.

Why "salvage some of the files" is the natural expectation and the wrong model: people reasonably picture the contents as spread across the drives like books across four shelves, where losing a shelf loses those books. The actual arrangement is more like every book having every fourth page on a different shelf — and losing one shelf damages the entire library rather than a quarter of it.

What does survive: files small enough to fit entirely within one chunk. Those sit whole on a single drive and come back intact — which is why a carving scan across the surviving members returns small items and almost nothing anybody cares about. It is the same pattern this archive sees whenever a scan is run against one member of a distributed set.

So why his first instinct is correct: the failed member has to be made readable. There is no route around it, because a quarter of the content only exists there. Repairing that drive is not one option among several — it is the whole of the work.

Why "does not show in the firmware" is the useful detail: a drive invisible at that level has not completed its own start-up, which places the fault at the board, the service area, or the mechanism — all of which are addressable, and none of which involve the content.

Why four drives is worse than two: the probability of the volume failing is roughly the probability of any one drive failing, multiplied by four. A stripe becomes more fragile as it grows, which is precisely the opposite of how storage arrangements are usually assumed to behave.

On the bench

The failed member was restored to a readable state as the determining step — a stripe distributing every file across all members in rotation, so a quarter of the content exists nowhere else and no route around that drive exists. All four members were imaged individually write-blocked, stripe parameters determined from set metadata and confirmed by content analysis, and the volume assembled offline from the images.

The outcome

The failed member made readable first, all four imaged, and the volume assembled offline from the copies. Free assessment, one fixed written figure including VAT, 50% of parts and labour upfront with the balance only on successful recovery. The decode: your first instinct is the right one and your fallback is not available. A stripe cuts every file into pieces across all four drives, so losing one costs a quarter of every file rather than a quarter of the files — nothing is left whole except items small enough to fit in a single chunk.

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.