NDR / Devices and media / RAID 1 mirrored pairs

Specialist · two-disk mirror sets

RAID 1 data recovery in Nottingham. Send the pair: half of what comes home sits on the disk that never failed.

From £500 + VAT covers RAID, NAS and server work here, and the diagnosis that fixes the figure costs you nothing. A mirror writes every block twice, so in theory either member carries the whole volume. In practice the jobs that reach this bench are the ones theory misses: two disks out of one carton wearing at one rate, a tired controller putting identical corruption on both sides, a re-sync that copies the failing member over the sound one, and the box that dropped a member in the spring and has run on a single disk ever since. Each member is cloned on its own, read-only. The clones are then set against each other block by block, and the volume is built out of whatever each side can still prove it holds, the newer of the two winning wherever they differ. Sets arrive by post from Nottingham, Beeston, Ilkeston, Chesterfield and Melton Mowbray, and your own disks come back unwritten to.

No files back, no bill — on most jobs Free diagnosis first, then a single price, in writing Parcels come in from Derby, Lincoln and Loughborough

Ask an engineer, and the first look is free
0115 8220606

What a RAID 1 mirror's symptom points to.

Not yours? Open the fault finder →

What is happening	What usually causes it	What that means for you
The NAS says the volume is degraded	One member has gone. Whatever is left in the bay now carries the volume with nothing behind it	Copy what you can, then post the pair
Both disks are slowing, and both went in on the same day	Mirrored members log identical hours, so they tend to wear at much the same rate	Take the power off the box now
A fresh disk went in and the readable side emptied	The set re-synced the wrong way round, writing the failing member over the sound one	Stop there, and post the pair
The same files are absent from both members	Every write reaches both sides at once, the bad ones included	Do not allow a re-sync
Nobody noticed the amber light until this week	The mirror has run on a single disk since then, with nothing standing behind it	Post both, not only the dead one
Windows reports the mirror as Failed Redundancy	The set has dropped a member. Very often both disks are perfectly sound	Do not reactivate it and do not import it

How to pack and post it: movement is what finishes a damaged drive, so pack it rigid, cover the parcel for what the files are worth rather than what the disk cost, and put a tracked label on it for the intake lab. We cover the journey back. If you would rather someone checked your packing first, ring before the box is sealed. The whole method is on the packing and postage guide.

A mirror covers one thing. Here is everything it does not.

Two-bay NAS units mirroringA Synology, a QNAP or a plain two-bay enclosure, with the safer of the two settings picked when it was set up. Because Linux software RAID sits underneath, the mirror's description rides on the disks rather than the chassis, so your box can stay on the desk — or come in whole, if opening it is not something you fancy.

Mirrored boot pairs and controllersThe pair a small server boots from, mirrored either by a hardware card or by Intel RST built into the motherboard. Whichever it is, an account of the set gets recorded onto both disks, and that record can be read out of clones with no need for the card itself or the machine it sat in.

Mirrors in Windows, mdadm and ZFSStorage Spaces or a mirrored dynamic volume on Windows; an mdadm pair on Linux; two disks in a ZFS or btrfs mirror inside a NAS. Every one of them writes its layout onto the members, every one will assemble above clones, and every one has a rule of its own for choosing which member to believe — precisely the question once they go wrong.

Why a rebuild can make things worseParity sets let a rebuild calculate the absent disk. A mirror has nothing to calculate: one member is copied across the other, whichever way the card has decided. Decide it wrongly and your only sound copy has gone by teatime. Everything here is done on clones instead, and a wrong turn there costs an afternoon, no more.

From the parcel arriving to the files going home.

Jobs already finished →

Booked in the day it arrives, and the first look is on us Free

Your device picks up a case number the day it arrives, and an engineer then establishes what has genuinely gone wrong — at no cost, and ahead of everything else. Two things come back to you together: a plain account of what can be lifted and what cannot, and one fixed figure in writing. Agree to that figure, or turn it down and pay nothing.

The first look is freeOne fixed figure, in writingNo charge at this point

Each disk cloned by itself

Every member is cloned end to end behind a write blocker, and no file is opened until that finishes. The disk that never failed gets the same care as the one that did, since on a mirror it supplies half of what comes home. Weak ground is handled in short passes, last of all. Not one byte returns to what you posted.

Nothing goes back onto your disksEach member on its own

The two clones set against each other

Where the two sides differ, somebody has to decide which of them to believe. Metadata on each member carries the date of its last write; the file systems held on the clones fill in what metadata cannot. Preference goes to the side that reads and to the side that is newer, and the reasoning behind every such decision is written down for you rather than taken on trust.

The newer of the two establishedSet side by side, never assumed

The volume rebuilt over the clones

Once that is settled the volume is stacked back over the clones and its file system repaired there, never on your disks. Files are then opened rather than tallied — ledgers, shared drives, databases, photographs — since a name in a listing proves precisely nothing. Whatever comes home damaged is listed as damaged.

Every file opened, not countedDamaged files listed as damaged

The file list comes before the bill

The file list reaches you before any invoice does. Say yes and it is billed; say no and it is not — and on most jobs, if nothing comes back there is nothing to pay. Whatever comes off goes home on media bought in for your case, carriage at our end. Nothing here is closed until you have opened those files on your own machine.

Nothing is charged until you agree the figureNew media, bought in for your jobWe pay to send it home

What comes in most often

Nobody ever sold you a mirror as a backup — each write lands on both disks at the instant it happens, so a deletion, a bad write or a ransomware incident gets copied over just as faithfully. The single event covered is one disk failing. Worth having, and the whole of it.
Disks bought together give out together — one batch, one shelf, one cabinet, one set of running hours. Inside a few weeks of the first member going, the second is commonly on its way as well, so the grace period a degraded mirror buys is much shorter than owners assume. The survivor in the bay is not automatically the healthier disk either.
Which way a re-sync runs decides everything — one member is written across the other and the card settles the direction on logic of its own. Nearly every bad job that reaches this bench began with a sound side being overwritten by a failing one, or a newer side by a stale one. Where a box has offered to rebuild, refuse it and ring us before anything further is pressed.
Post the disk the box condemned — very often that is the member carrying the newer data, and very often it reads well enough to be worth cloning anyway. Hold it back and half of what could be done has gone before we start; occasionally the whole job goes with it.

The honest limits: storage is all this bench does. Mirrors are this page's subject — a pair inside a network box, a pair in a desktop enclosure, a pair on a hardware card, a pair on Intel RST, a mirrored Windows volume, an mdadm or a ZFS pair — arriving degraded, written over, or re-synced in the wrong direction. Read both sides and the volume is rebuilt above the clones with your folders arranged as you kept them. Blocks that have gone on both members cannot be conjured by anyone: where identical ground has given out twice, certain files come home with holes in them, and each of those is named for you in the report. Overwritten data does not survive either. Encryption will want its key from you. Phones and tablets are not taken. Pricing starts at £500 + VAT, fixed in writing after a free diagnosis. Where nothing is recovered, most jobs carry no fee at all.

One of these, from start to finish.

NG · NTG-2026-1996ON THE LOG ✓

A two-bay box that had run on one disk since the spring

An engineering firm in Ilkeston rang when the office box stopped presenting its data volume. Months earlier a member had dropped out behind a door nobody opens, with an amber light the only sign of it, and the survivor had been running the volume single-handed since — and was starting to fail itself. Each disk was imaged alone, the worn one over two nights in short passes. Blocks that disk refused were pulled off its older partner; blocks the older partner had missed came from the newer side. Twenty years of drawings, plus the job ledger, were opened on the bench before the parcel went back. Three days, door to door.

The two members cloned one at a time3 days between arrival and return

What helps, and what does damage.

Worth doing first

Cut the power to the box and leave it off
Post every member, the dead one included
Number each disk to its bay before pulling it
Tell us whether a rebuild or a re-sync has run

What makes it harder

Letting it re-sync onto a replacement disk
Reactivating a Windows mirror marked Failed
Turning repair utilities loose on either member
Posting one disk and keeping the other at home

Answers before you spend anything.

One disk is still running. Why do you want the other?

Two versions of one volume is what a mirror leaves you holding, and they have almost never stayed identical. Blocks one side refuses will often come off the other cleanly. A member can also be weeks out of date, having quietly left the set before anybody looked at the front panel. Put both clones on a bench and each block can be sourced from the side that still reads it, at the date you actually want; no striped set offers that choice. So: the pair, bay numbers marked, and the disk your box has condemned as well.

The box wants to rebuild onto a fresh disk. Should I let it?

No — and of everything on this page, it is the decision that does the most damage. What a rebuild does is copy one member across the other in a direction the controller settles for itself. Should the disk you were left with turn out to be the failing one, or merely the older one, your good side is what gets written over, and no second version then survives anywhere. Kill the power, leave every disk in its bay and have clones taken first. After that, nothing anyone presses can undo them.

I deleted a folder and it vanished from both disks. Is a mirror not a backup?

It is not, and buying one as though it were is the commonest misunderstanding this bench meets. Mirroring duplicates each write onto both members in the same instant, so a corrupt write, a deletion, or an encryption run by ransomware gets copied across as faithfully as anything else. One disk failing is the single event covered. Whatever has not yet been written over the top can be recovered, which is the reason a box wants unplugging on the day rather than at the weekend.

What do I send, and what will it cost?

RAID, NAS and server jobs begin at £500 + VAT. The diagnosis ahead of that is free, and one fixed figure, in writing, is what you agree to before any chargeable work begins. Include every member of the set, bay number marked on each, the disk the card condemned as well, and tell us what has already been attempted. Would you rather not open the chassis? Then post the network box as it stands. Disks pulled out of a server chassis or a desktop enclosure travel bare, in antistatic bags. Nothing is collected here and there is no counter to call at. Post it insured, with tracking, to the intake lab; the address sits on our contact page.

Start here if the same thing is happening to yours.

When the server will not start → Asus and HP laptops → Server and RAID work → The full device list →

Switched off, a drive cannot get any worse.

The look costs nothing. What comes back to you is a file list — what opened, what did not — plus one figure to finish the job, in writing, before anything chargeable begins. Most jobs carry no fee at all unless the data comes back. Leave the drive switched off until you have that list.

0115 8220606

Send it for a free look Posting it to Manchester

Disks, cards and whole machines

Start with what it is doing

Footage, forensics and reports

The price, and what happens next