The mirror that wouldn’t take its medicine.

The unit arrived with both original disks and the unhelpful newcomer, everything labelled. The pattern in the story was the diagnosis half-made: a two-bay mirror survives one failure by design, so the data’s guardian was the surviving disk — and every failed rebuild had meant hours of that lone survivor thrashing at full load, re-reading its entire surface while the box tried to copy it onto a drive that kept baulking. The volume dropping on attempt four was the survivor waving a flag.

The newcomer was slightly smaller than the original once the box did its arithmetic — a mirror can’t rebuild onto fewer sectors than it holds — and it was an SMR drive besides, a recording technology that writes in a slow shuffle the NAS kept mistaking for failure. Meanwhile the survivor had developed a small cluster of slow sectors under the repeated full-surface hammering. Every further rebuild attempt was wearing out the only disk that mattered.

The surviving disk was imaged first and gently, its slow patch read last with long timeouts, before anything else was even examined. From that image the ext4 data volume mounted directly — a mirror’s member is a complete copy, which is the whole point of mirrors — and the family’s entire library was extracted and verified. The failed original and the unsuitable replacement contributed nothing, and were never asked to.

Photos, documents, the lot — verified and returned on a new external at the from-£500 + VAT multi-disk band quoted in writing up front. The advice that went home with it: when a mirror loses a disk, buy the replacement by exact model or a vetted equal — same capacity or larger, conventional recording — and if the first rebuild fails, stop. Each retry spends the health of the one drive still holding everything.