NAS & RAID · job record · RDG-2025-1491
One Bay Failed. Nine Months Later, a Second.
A four-bay Synology had been living on a shelf over the grinding bench of a boatyard workshop at the marina in Caversham, its fan pulling dust off that bench for years. Bay 2 had been flagging bad sectors since the turn of the year; rather than ring anybody, somebody had turned the warning emails off. Then an order file would not open one morning, the unit was shut down and restarted, and it came back up reading volume crashed
. Somebody had left an instruction on the lid in marker pen: put a new disk in and let the thing rebuild itself
. That was the one instruction we did not follow.
Same fault on yours? Tell us what it is doing.
0800 6890668
What actually caused it.
RAID 5 holds exactly one disk in reserve and not a fraction more. The parity is thin, a little of it in every stripe, and it will reconstruct one absent member and nothing beyond that. With two members gone the arithmetic has no unknown it can solve. Switching the unit on for a look would therefore have cost twice: two exhausted disks put back to work, and a failed volume left free to write fresh metadata over the structures the recovery would need. It stayed unplugged. We refused the marker-pen instruction in writing and gave the reason — feeding a failing disk back into a live set simply adds load to the members still carrying it. Nothing was powered. Four disks were copied, and every step afterwards ran on the copies. NAS and RAID work is published from £500 + VAT, and the assessment that sizes a job costs nothing.
What this one needed on the bench.
How every job runs →| Bench tools | Its part in this job | Why it is here |
|---|---|---|
| DeepSpar Disk Imager 4 | New head stacks into both failed members, and the more recent failure imaged first | It maps the heads before anything is taken off, keeps to one platter side per pass, and stays in charge of the resets, the timeouts and the power |
| Atola TaskForce 2 | The two sound disks copied side by side while the failed pair waited | Takes every member of a set at once, so nothing sits waiting behind another disk |
| UFS Explorer RAID Recovery | mdadm superblocks read off first, then the volume rebuilt across four images | Puts the array back together, then unpicks whatever volume manager the NAS laid over it |
What the bench did.
The surviving pair were tired too
Carrying a degraded array for months leaves its own marks, and by now both surviving members had pending sectors. The TaskForce took that pair concurrently while the failures waited, which saved several days and, more usefully, meant no member of this set was declared healthy without having been read from one end to the other.
What the event log turned up
A donor head stack went into each failed disk and both began to read. Then the event log changed the shape of the job. The noisy bay had left the array back in January, so its contents were nine months out of date — a picture of the volume from long before the trouble. Everything recent lived on the disk that had failed most lately, and there the unreadable areas fell into bands that could be worked around.
Three images in, and January left out
Member order, stripe width and the direction parity turned were all read out of structures the array itself had written; none of it was assumed. Assembly then used the three recent images and excluded the turn-of-the-year disk altogether, since blocks written months apart can blend into corruption neat enough to pass a consistency check while remaining wrong. In any event, what the newest disk withheld proved to be sitting in unallocated space.
What went back.
The volume mounted at the first attempt. Drawings, job cards and the yard's order books were all checked against its own numbering before new media and replacement disks were posted back. That Synology now sits in the office rather than over the bench, and the alert emails are switched on again. That figure was agreed in writing before a single disk was opened.
If this sounds like yours, start here.
More RAID & NAS work, written up.
If that sounds like yours, stop there.
Post it in as it is, switched off. Looking costs nothing, and you get one fixed figure in writing before any chargeable work begins — £250 + VAT for a card or stick, £300 + VAT for a single drive, from £500 + VAT for RAID and NAS. On most jobs, nothing is charged unless your data comes back.