Ten O'Clock Monday, and Four Disks in a Box.

Ten o'clock on the Monday was when the school extension tender closed, and the Truro practice had most of it written already. On the Thursday the QNAP disappeared off their network. By Friday there was nothing left in it at all: the box is dead, the disks look fine. Every drive was out of the chassis before anybody rang us, labelled by bay as they came out. They wanted a fixed figure and an answer to one question: will it be back for Monday.

The customer checked and agreed No names published

Sounds like yours? Talk it through.
0800 6890668

Why it happens.

Weekend jobs are won on the timetable. Nothing was taken out of the process; instead the stages were made to overlap. Every disk had an imager of its own and all four began together, so no drive sat in a queue behind another. Array geometry was settled from whichever copy finished first while the slower members were still reading. Nothing on the bench sits idle. An obvious short cut was available: rebuild straight onto the practice's own four disks and do without copies. That saves a few hours. It also bets a tender on the hope that all four disks read perfectly first time, and had one of them softened halfway through, there would have been nothing to fall back on.

The kit this job called for.

How the job runs →

The kit	What it did here	Why we keep it
Atola TaskForce 2	Four images taken simultaneously — that is what bought the weekend back	Copies the whole array in one run, so no member waits its turn on the bench
PC-3000 Portable PRO	Assessed each disk in place, for condition and firmware state	Manufacturer-level access, from a case standing beside the imagers
UFS Explorer RAID Recovery	Derived the disk order, the stripe width and the parity rotation from the completed copies	Rebuilds the set, then works back through whatever volume layers the NAS added on top

How it ran.

Four disks, four imagers, nobody queuing

The box arrived at midday. Within the hour every disk was sitting on an imager and reading, and none of them came off until the last sector had crossed. Feed a set through one machine at a time and three days of array becomes a week of work. Run them alongside each other and a weekend remains a weekend.

Settle the geometry while the copying runs on

Metadata was taken off whichever images landed first, and it produced a short list of possible layouts: which disk sat in which bay, how wide the stripe ran, which way parity turned. Every candidate was then tried against actual directory structures on the volume. No option was taken on trust merely because controllers of that make usually behave in a particular way. The layout was down on paper before the last copy closed.

Assemble it in software, and check the tender before anything else

Put together in software from the four copies, the volume mounted at the first attempt and the filesystem came up clean. Checking followed the practice's own priorities rather than the drive's: drawings, then specification, then appendices, then cost plan. Every document in the pack was confirmed by telephone as it surfaced, with a great many folders still to come.

How it ended up.

Verified files reached the office before Sunday was over, and on Monday the practice filed its tender from its own drawings. It went that way because the stages overlapped, and because anybody who rang got a straight answer about where the job stood. Nothing was left out in order to make the date.

In short: A deadline is met by overlapping the stages, never by leaving one out. Any lab offering to skip the copying is gambling with your files, not its own.