Data Recovery Case File · NAS & Network Storage · A Mirror Is the Good News

NAS Dead After a Power Cut: Reading the Disks

His enquiry was one of the best-informed in this archive, and it left almost nothing to establish. A Buffalo LinkStation NAS that "will not power up following a power cut in my home. I am confident all of the data is on the drives and that the drives will spin up. I need someone who can access the data on one of the disks — they were set up as RAID 1, I believe. I know the drives run XFS as I spoke to Buffalo directly; they do not provide spares any more so I cannot fix the unit." His ask is exactly right: "can you give me a price to read the disks and copy the data onto a drive that I can read on a Windows machine?" He has diagnosed the appliance, identified the array level, established the filesystem, confirmed spares are unavailable, and specified the deliverable. Two things remain worth saying — and the first is that his configuration is the most fortunate one this situation can have.

The decode: why a mirror helps, the filesystem catch, and the danger to avoid

Why the appliance dying is not the problem: a NAS is a small computer attached to some drive bays, and a power cut is hard on its power supply and board while doing nothing at all to the disks. The data was never in the appliance; it is written on the members, and the unit was only the thing that read them. Spares being unavailable is therefore an inconvenience about the hardware rather than an obstacle to the data — which is exactly the position he has already worked out.

Why a mirror is the best possible configuration here: a mirrored pair writes the same data to both disks, so each member holds a complete, independent copy. That has two consequences worth stating. There is no layout to reconstruct — no stripe order, no block size, no parity rotation to derive — because a mirror member is simply the whole volume. And there is redundancy in the recovery itself: if one disk has developed a weak region, the same data can be read from the other. Compared with the striped arrays elsewhere in this volume, where every member holds half of every file and losing one is catastrophic, he could not have chosen better.

The filesystem catch, which is why he cannot do this himself: the disks are formatted with a Linux filesystem. Windows does not read it natively, and this is where the situation becomes dangerous rather than merely inconvenient: connect one of these disks to a Windows machine and it will appear as an unformatted or unrecognised drive, and Windows will offer to initialise or format it. That offer is presented as the obvious way to make the disk usable, and accepting it on a mirror member would destroy a complete copy of everything. It is the single most likely way for this case to go wrong, and it is worth knowing before the disks are plugged into anything.

The deliverable, done properly: his request — data copied onto something he can read on Windows — is exactly the right specification, and it is a real piece of work rather than a formality. The volume is mounted from images of the members, the contents extracted with their folder structure and dates preserved, and written to a Windows-native filesystem on media sized comfortably above the source. Then the step that matters: it is verified on a Windows machine before handover, because "Windows-readable" is a claim you test on Windows rather than assert.

On the bench

Both members were imaged write-blocked on the Atola TaskForce 2 — both, rather than one, so that a weak region on either could be answered from the other, which is the advantage a mirror provides and which is wasted if only one disk is read. The XFS volume was mounted from the images, where nothing could be written to the originals and no host could offer to initialise anything, and the contents were extracted with folder structure, filenames and dates intact. They were then written to a Windows-native filesystem on fresh media and opened and browsed on a Windows machine before delivery, so the deliverable he specified was confirmed rather than assumed.

The outcome

Both mirror members imaged, the Linux volume read from the copies, and the contents delivered on Windows-readable media verified on Windows. Free assessment, one fixed written figure including VAT; where a drive has to be opened, 50% of parts and labour is payable upfront with the balance only on success — otherwise no recovery, no fee. The decode, for anyone whose NAS has died: the appliance is the reader, not the storage, so a dead unit and unavailable spares are hardware problems rather than data ones; a mirrored pair is the most fortunate configuration for this, because each disk holds a complete copy and there is no layout to reconstruct; but the disks are almost certainly formatted with a Linux filesystem that Windows cannot read, so plugging one into a PC produces an offer to initialise or format it — and accepting that would destroy a full copy of everything.

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.