Specialist · Synology NAS work
That one alarming line stands in for three unrelated faults — members dropping out of the RAID, LVM losing track of its map, or the btrfs tree going bad — which look the same on screen and want opposite handling. So the move that is right in every case is the dull one: switch the DiskStation off and stay away from Repair.
Speak to an engineer about it
0800 6890668
| The problem | What that tells you | First step |
|---|---|---|
| DSM reports Volume Crashed | One alert covering three possible culprits: RAID, LVM or btrfs | Switch off and leave Repair alone |
| DSM reports Storage Pool Degraded | A disk has dropped; the pool still runs, but with no protection left | Each further hour adds risk |
| The DiskStation enclosure has died | Your data sits on the disks, not in the chassis | Among the better outcomes |
| Rebuilding onto a replacement disk stalled, or 'crashed' | The long read of a rebuild exposed another weak disk | Stop now; clones will tell |
| Disks put in a PC that now asks to format them | Normal — Windows cannot read Linux NAS layouts | Refuse every prompt |
| A ransom demand sitting where your shares were | One of the NAS-targeting families got in | Photograph it, shut down, read our ransomware page |
As soon as it lands with us, your device gets its own case number. An engineer works out the fault, says what can genuinely be pulled off, then puts one fixed price in writing — no charge for diagnosis, no obligation, and no chargeable work until you approve it.
Every disk is cloned with its bad sectors charted. The DiskStation never gets a chance to resume anything, and the earlier btrfs roots stay in the state the crash left them in.
Working on the clones: the slices mapped, each mdadm array reassembled, the LVM map read, then the ext4 or btrfs volume mounted read-only — one layer at a time, in the sequence that works.
A crash at btrfs, LVM or RAID level each gets its own fix, applied to the reconstruction — btrfs wound back to an older sound tree when the newest is broken — and the shares checked before sign-off.
You sign off a complete list of the recovered files first; only then does the recovery fee fall due. Everything returns on fresh media, postage paid, and the job stays open until you confirm the files open at your end.
The structural point that settles these jobs: an SHR volume built from unequal disks isn't a single array — it's a stack of mdadm arrays over per-disk slices, pooled by LVM under one filesystem. Unwind those layers out of sequence and you find nothing; unwind them properly and volumes the DiskStation called crashed come back routinely.
A pair of 4TB disks, a pair of 8TB, one dead member, and an overnight DSM rebuild that finished in Volume Crashed. Beneath that sat three distinct mdadm arrays: two intact, the third degraded but still derivable. Reassembled in sequence off the box, with the btrfs tree wound back a single generation, every share came back.
Very unlikely. Crashed only says DSM can't put the stack back together — perhaps the RAID lost members, perhaps LVM metadata broke, perhaps the btrfs tree corrupted. Three separate faults, and mostly recoverable ones; the message simply can't say which. Leave it powered off and all three stay open.
No. Underneath, SHR is ordinary Linux: mdadm software RAID over partition slices, pooled by LVM, then formatted as btrfs or ext4. From drive images it rebuilds on a normal workstation; the trouble only starts with tools or people expecting a single flat array.
Not until every member disk has been imaged. A rebuild pulls each sector off the tired survivors — precisely the strain that triggers a second failure — and it can write over the old filesystem roots that recovery leans on. Image them first and the rebuild carries no risk.
Often in part, sometimes in full, depending which strain hit — campaigns aimed at NAS boxes exposed to the internet go back years, so we check snapshots, replication targets and remnants in free space. Photograph the note, shut the unit down, and start with the routes on our ransomware page.
Powering a damaged device up again costs you data. Start a case first; diagnosis is free whatever you decide.