BHDR / Specialist work / RAID 10 data recovery

Specialist · RAID 10 array recovery

RAID 10 recovery, Birmingham. One disk per pair can go; two from the same pair ends it.

RAID 10 pairs the disks off, mirrors each pair, then lays a stripe across the pairs, so you get the speed and a second copy in one box. That is why it sits under so many small business servers and four- and six-bay NAS units around Birmingham. The cover is narrower than most owners think: one disk out of each pair can go, but two out of the same pair ends the volume there and then. Every member is imaged before anything else happens, the pairing and stripe order proved off those copies, and the array rebuilt in software, not on the hardware that dropped it.

Nothing recovered, nothing to pay Diagnosed free, then one fixed price From Walsall, Dudley or Coventry, it travels by post

Tell an engineer what it is doing
0800 6890668

Reading the RAID 10 symptoms, one by one.

No match here? Run the triage →

What you see	What is behind it	What to do
One disk failed and the array still works	The mirror is doing its job, but the set now runs with no cover left	Get a copy off it today
A second disk dropped and the volume went offline	Both halves of one pair are down, so the stripe cannot be completed	Power off; leave the disks alone
A four-bay NAS reports the volume crashed	mdadm has dropped members and will not assemble the array	Send the unit whole or all four disks
The controller began a rebuild by itself	It is copying from a mirror that may be failing too	Stop it; add no more disks
Large files come back damaged	The set was reassembled with the pairs or the stripe order wrong	Stop writing; prove it from images
A disk was swapped and the rebuild stalled	The surviving half of that pair has bad sectors of its own	Pull the power; send all members

Sending it by post: send it tracked and fully insured to our intake lab, and we cover the postage coming back; if you want the packing checked before it goes, an engineer will talk it through with you. Every step is written out on the posting page.

Two mirrors underneath, one stripe on top.

What RAID 10 actually isDisks are paired up, each pair holding two copies of the same blocks, and the stripe then runs across the pairs. Reads and writes spread over all of them, and the usable space comes to half of what you bought.

Where the cover runs outOne failure in each pair is survivable, however many pairs there are. Two in the same pair is not, and adding bays makes that coincidence likelier rather than rarer. Speed and resilience are not the same thing.

What has to be proved firstThree questions come before any file: which two disks mirrored one another, what order the pairs sit in across the stripe, and how wide each block is. Bay numbers are a hint, not an answer. All three come off the copies — file system structures, and any record the controller kept of the set — then tested against files we can already recognise.

Four- and six-bay NAS boxesSynology and QNAP units set to RAID 10 run mdadm underneath with ext4 or Btrfs on top. Put one disk into a PC and Windows sees an unformatted volume and offers to initialise it. Say no. Send the whole unit if you would rather, or every disk with its bay marked.

From arrival to the files going home.

Recent jobs on record →

Case opened, and the diagnosis costs nothing Free

Every item gets a case number of its own the day it lands. An engineer then works out what has actually failed and tells you plainly which files have a real chance of coming back and which do not. Only then does a price follow: one figure, fixed, in writing, and it costs you nothing to see it. Nothing is charged until you say yes.

The diagnosis is freeOne price, put in writingNothing agreed

Read-only images of every disk

Each disk is cloned to an image first, read-only, with any unstable member run slow and stepped past the sectors that refuse to answer. Nothing we do writes to the drives you posted.

Read-only copies of each diskNothing written to your disks

Establish pairs and order

Which disk mirrored which falls out of the images, because two members of a pair carry the same blocks at the same offsets. Block width, the order the pairs sit in and the offset the data begins at then come from the file system, and each candidate gets tested.

Mirror pairs identifiedStripe order proved, not guessed

Rebuild it in software and open files

One good half is taken from each pair and the images stitched into one volume again, with the file system mended where it is torn. Then files are opened — a database, a long video, a drawing. A directory that lists neatly on its own proves nothing at all.

Array rebuilt off the imagesFiles opened and checked

Approved by you, then posted back

No invoice is raised until the full list of recovered files has sat in front of you and you have said go ahead. Your data comes back on media we buy new, posted at our expense, and the job is not closed here until every file has opened on your own machine.

Your say-so on the file listYour data on new mediaWe cover the postage back

What the first look turns up

The cover is one disk per pair — owners hear four disks and two mirrors and assume any two can fail. They cannot. Two gone from the same pair puts a hole through every stripe, and the volume stops dead at that point.
A rebuild can write over the good half — slot a fresh disk in and the controller starts copying the surviving member onto it. Where that survivor is unwell itself, the copy stalls part way through and the array ends up in a worse state than it began in.
Bay order is not pair order — which disks were mirrored together was the controller's decision, and it is not always bays one and two, three and four. Pairing is proved from the content of the images before a thing is assembled, because getting it wrong scrambles the whole volume.
RAID 10 and RAID 0+1 are not the same — one stripes across mirrors, the other mirrors two stripes. They read alike in a menu and behave differently once a second disk goes, so we work out which of them you actually have.

What we can and cannot promise: where one disk per pair still reads, a RAID 10 normally comes back whole — imaged, pairing proved, stripe order settled, volume assembled, files opened and checked. Where both members of a pair are down, it rests on how much either will surrender under a slow read, and between two failing copies of the same blocks there is often more to be had than either holds alone. Where a rebuild has already run across the survivor, we work from what the images show. The free diagnosis is where you find out which of those you are in.

Recent pages of the casebook.

BH · BHD-2026-8846RECORDED ✓

The four-bay NAS that lost both halves of the same pair

A design firm ran a four-bay NAS as RAID 10 and lost a disk in the spring without ever reading the warning email. The second failure was the other half of that same pair, and the volume went offline. All four were imaged, the two dead members read slowly across two days, and pairing and stripe order proved from the mdadm metadata. Live project folders came back whole; a run of older renders returned with gaps, every one of them named.

Both pairs imaged in full3 days from arrival

Before the parcel goes.

Do this first

Power the server or the NAS down
Mark each disk with the bay position it held
Send all members, dead ones included, none left behind
Say which controller or NAS built the set, and whether a rebuild was started

What not to do

Letting the controller finish its rebuild
Slotting in a fresh disk to set one going
Clearing the configuration to start again
Running file scans on one disk alone

The questions that come up every week.

Two disks have dropped out. Is the RAID 10 finished?

It turns on whether the two sat in the same pair. If they did not, the array is intact in principle and the work is a careful imaging job done before anything else lets go. If they did, the stripe has a hole through it, and the answer rests on what those two members will still give up under a slow read. Two failing copies of identical blocks often cover for each other, and what one will not hand over the other sometimes will. Which position you are in is settled at the free diagnosis, before you have spent anything.

Do you need every drive, or only the pair that failed?

All of them. The stripe runs across the pairs, so half the blocks in any file sit on the pair that is still healthy, and the layout can only be proved with the full set in front of us. Send them as they are: nothing initialised, nothing scanned, no rebuild attempted in the box, and every disk marked with the bay it came out of. A NAS may travel whole if that suits you better. Bench hardware will often read a member the controller has already written off, and how much it reads decides how complete the result is.

How do you work out which drives were mirrored together?

From the content itself, on the copies, never on the disks you sent. Two members of a pair hold the same blocks at the same offsets, so comparing copies makes the pairing plain within minutes. Pair order and block width come out of the file system: partition tables, directory runs, and the point where a long file breaks off on one pair and picks up on the next. Controllers and mdadm usually leave a description of the set on each member. Candidate layouts are then assembled and opened against files we can already recognise, because the wrong answer still produces something that mounts and lists tidily, with every file bigger than a block quietly wrecked.

The controller has started rebuilding on its own. Should we let it finish?

Stop it if you still can. A rebuild reads the surviving member of the broken pair end to end and writes what it finds onto the replacement. Where that survivor has bad sectors of its own, and after a partner dies it often has, the rebuild grinds, retries, and can push a marginal disk over the edge while laying down a half-finished copy. Shut the machine off, leave the disks where they are, and send the lot. What is on them gets read off images instead.

Other pages worth a look here.

When a server dies → Databases & mail → RAID sets & servers → Media we take in →

Nothing more is lost while the power stays off.

Every extra power-up takes something off a drive that is already going. Leave it switched off and let the free diagnosis say what still reads.

0800 6890668

Start with a free diagnosis Post it to the lab

What we take in for recovery

Faults we see

Forensic and CCTV work

How it all works