A D-Link NAS reported a failed drive. The client fitted a replacement and started a rebuild, and the unit estimated four months. They let it run, it failed partway, and afterwards nothing on the array would open at all.
The NAS had been freezing during large transfers for some time — stalling mid-copy, unresponsive until the whole unit was reset. Then it reported one of its two mirrored drives as corrupt.
The client did what the interface invites you to do: fitted a new disk and started a rebuild. The unit estimated 120 days. They let it run anyway, it failed partway through, and from that point the array would not mount at all — including the original disk that had been in the unit since the beginning.
This is the whole case in one paragraph, and it is worth being precise about the arithmetic.
A two-disk mirror rebuild is a straight copy. Read every block from the surviving member, write it to the replacement. On healthy hardware a modern drive sustains well over a hundred megabytes a second, so a few terabytes takes hours — and the kernel deliberately throttles resync anyway so the array stays usable while it runs.
Now work backwards from 120 days. Copying two terabytes in that time is roughly two hundred kilobytes a second. That is not a slow disk. That is a disk spending almost all of its time not reading — sitting in internal retry on sectors it cannot resolve, returning data only in the gaps between failures.
So the unit was not estimating a long job. It was measuring, and reporting, that it could barely read the source at all.
The stalling during large transfers had been the same fault announcing itself for weeks, and it went unrecognised because of how it presents.
When a drive hits a sector it cannot read, it retries internally — and a desktop-class drive will spend a long time doing so, tens of seconds on a single sector, because in a single-drive machine that persistence is correct behaviour. Drives sold for NAS use ship with that timeout deliberately capped at a few seconds, so the array can get on with reconstructing the sector from redundancy instead of waiting.
A desktop drive in a NAS therefore produces exactly what the client described: the whole unit hangs, unresponsive, until the drive either succeeds or gives up. It feels like a network problem or a firmware problem. It is a disk stalling, and the NAS waiting politely.
So the array was never one failed drive and one good one. It was one failed drive and one deteriorating drive — and the rebuild put the deteriorating one under continuous full-surface read stress for as long as it was permitted to run.
All three arrived: both originals and the replacement.
The disk the NAS had marked corrupt had a head fault and was genuinely failed. The surviving mirror had extensive reallocated sectors with a pending count still climbing — readable, but slowly and not indefinitely. The replacement held a partial aborted copy: some blocks written, most not, no coherent file system.
That last one mattered more than it looks. The rebuild had been writing to the new disk, not to the survivor, so nothing had been destroyed by the attempt. Had this been a parity array where a rebuild writes across members, the same sequence would have been considerably more expensive.
The recovery approach follows from what a mirror is, and it is the reason this returned 99% rather than something much worse.
The surviving member was imaged first and imaged carefully — a fast initial sweep at large block sizes to bank everything that read immediately, then repeated passes over the gaps with longer timeouts and rest intervals. On a disk in that condition the order is everything: secure the bulk before it deteriorates further.
The failed member was then opened under filtered air, fitted with a matched donor head assembly, and imaged too. Recovering a disk the NAS had already written off might look like duplicated effort. It is the opposite.
In a mirror, both members hold identical content at identical addresses. So with two images in hand you can merge them address by address, taking whichever member returned readable data for each block. A sector unreadable on one is frequently perfectly readable on the other, because the two disks failed in different places for different reasons. Recovering both gave two independent attempts at every block on the volume rather than one.
The file system was rebuilt from the merged result, with each set filling the other’s gaps.
About 99%. The shortfall was confined to the small number of addresses unreadable on both disks simultaneously — which is precisely what having two copies is supposed to prevent, and did, almost entirely.
The transferable point is short. If a mirror quotes days or weeks to rebuild a few terabytes, it is not being slow — it is telling you it cannot read the source properly, which means the surviving disk is failing too. Stop, and get the data copied off before replacing anything. And if a NAS freezes during large transfers, suspect a disk before you suspect the network. NAS and RAID recovery is from £500 +VAT after a free 48-hour diagnostic.
Send the device over with a short account of what happened to it. The diagnostic costs nothing and commits you to nothing, and the figure that follows is fixed in writing before anything is opened.
Nothing can begin until the device is on the bench, which makes this the only step that needs anything from you. Pack it properly, put your details in with it, and send it over. What follows is a free diagnostic and a written figure, in that order.
Posting it? Use something tracked and insured — whatever is on the drive is worth considerably more than the postage. Bringing it in? Weekdays, 9am to 5:30pm, and it still wants packing as above for the journey.
Tell us what the device is, what it is doing, and anything already tried — an engineer will come back with a realistic view of the odds and a price band, before you commit to posting anything.
We’ll be in touch shortly. For anything urgent, call 0161 871 0788.
No. A mirror rebuild is essentially a copy, and a few terabytes on healthy hardware takes hours. An estimate in days or months means the unit cannot read the source at speed — which means the surviving disk is failing as well. Stop it.
Frequently, yes, though it reads like a network fault. A drive retrying unreadable sectors stalls everything waiting on it, so the unit appears to hang. If freezing coincides with large reads, check the disks rather than the network.
Usually less than you fear, on a mirror. The rebuild writes to the new disk rather than to the survivor, so the original data is generally still there. What it does do is run the surviving disk hard for the whole attempt, which is why stopping early matters.
Because both hold the same data, so a sector unreadable on one can often be read from the other. Imaging both gives two chances at every block, and in this case that is what took the result from partial to near-complete.
Copy the data off first. Replacing prompts a rebuild, and a rebuild is the heaviest read load the array ever experiences — applied to disks the same age as the one that just failed.
From £500 plus VAT, fixed in writing after a free 48-hour diagnostic. Where a member needs clean-air work there is a 50% deposit with the balance due only on success.
Start with an instant online quote, or call and talk it through with us first. You'll have a clear, fixed price before any work begins.