RDR / What we take in / Servers and their RAID sets

Servers · RAID sets that have stopped

Server data recovery in Reading. The server has stopped; the disks inside it are very probably fine.

In nearly every server that arrives at the Reading bench, the drives are the healthy part. What has failed is the machinery around them: a controller whose idea of the set has drifted away from the disks actually in the bays, a console prompt whose two buttons are both expensive, a cache still holding writes that the mains cut short. Every member is copied before anything else is attempted, and all the repair work then happens on those copies, so your controller can stay bolted in the chassis. Back comes the whole volume — mail stores, file shares, virtual machines, databases and all. From £500 + VAT, one figure in writing after a free diagnosis. A distributor at Theale and a research spin-out on the science park at Shinfield send in the same chassis.

Nothing recovered? On most jobs, nothing to pay The diagnosis is free; then one fixed figure in writing Parcels arrive from Newbury, Woodley and Theale

An engineer answers, and looking is free
0800 6890668

Match the server symptom to the fault.

Different fault? Begin at the triage →

What you are seeing	What is likely wrong	What happens next
On the PERC: Foreign Configuration Detected	What the members say about the set and what the card remembers have stopped agreeing, so it halts rather than guess. Nothing is lost at this point	Give the console no answer
HP: 1786 — Drive Array Recovery Needed	A rebuild has started and may have stopped part-way	Power it off. Copies before anything else
Amber bay lights after a mains failure	Writes were cut off half-finished, and what the cache still held is anybody's guess. The array's own description is intact	Do not switch it back on
It boots, but the data volume is missing	Configuration on the card, or the file system above it — either way the sectors underneath are intact	Normally comes back
The RAID controller itself has failed	Every member records the bay order, the stripe size and the way parity rotates	No identical controller needed
Files copied off, but no VM or database will run	A copy that exists is not a copy that is coherent, and only the coherent kind will start	That turns into a database recovery job

Packing and posting: pack it so nothing shifts, and put a value on the parcel that reflects the data rather than the hardware, then send it tracked to the intake lab. We cover the return postage. If you would rather an engineer talked the packing through before the box is taped up, ring first. Everything is laid out on the packing and postage page.

The members carry the map of the server's array.

MegaRAID and PERC cardsDDF is an open format, and both cards write it into the final sectors of every member. The images give it up readily, and the set can be reassembled with your controller still bolted into the server.

HP's Smart Array controllersHP takes another approach: RIS structures at the head of each member, and a parity block sitting one stripe off the position most tools assume. Point a default preset at that and you build a volume the server never held.

The prompt where both answers cost youAnswer this one at your peril. Import can lay stale metadata over stripes that were entirely sound, and Clear takes the description off every disk with nothing held back to put it there again. Take the images first and this prompt loses all power to hurt you.

Writes still sitting in the cacheAs soon as battery-backed memory has the write, the write is reported done — long before a platter sees it. Lose mains part-way through a flush and what breaks is the file system itself, invisibly, with nothing on screen to tell you. Reason enough to image before anything else.

From your parcel to your files coming back.

Jobs we have closed →

Logged the day it lands, and the look costs nothing Free

A case number goes on your device the day it reaches us, and an engineer then finds the real fault — free, and before anything else happens. You are told plainly what will come back and what will not, with one fixed figure in writing beside it. Nothing chargeable starts until you have read that and said yes.

The look costs you nothingA fixed price, in writingStill nothing to pay

Leave the machine exactly as you found it

Power the server off first. Then the drives come out one at a time, and each gets its bay number written on it before it goes anywhere. Answer nothing the console asks. Here, every member goes onto purpose-built imaging hardware and is copied on its own — including the ones the card had already refused — and not one file is opened until the final copy completes.

Each disk imaged one by oneThe rejected members too

Reconstruct the array in software

The images carry the DDF or RIS records with them. Pull those out and the geometry is yours: stripe size, RAID level, which member sat in which bay and, on a parity set, the direction the parity rotates. That is everything needed to assemble the array in software, at which point a card that has failed stops mattering to anybody.

Geometry read off the membersYour controller stays in the machine

A copy is not yet a working volume.

Whatever the array was carrying — NTFS, ReFS, ext4, XFS, a VMFS datastore — is repaired on the reassembled copy and never on the drives you sent us. Then it is tested in earnest — guests booted, shares mounted, databases attached, and individual files opened and read.

Repairs only ever on copiesChecked by mounting and booting

The list first, then the bill, then the files

You see the file list first, then decide. Nothing is invoiced until you have said yes, and on most jobs there is no fee at all if the data does not come back. What we pull off goes onto media bought in for your job, and the postage home is ours. The case stays open on the bench until you tell us the files open on your own machine.

Nothing billed until you say yesNew media, bought for the jobPostage home is on us

What goes wrong most often

No two makes store the map in the same spot — on an HP member RIS sits at the front, while DDF, which PERC and MegaRAID write, lies right down in the last sectors of the disk. Look at the wrong end and every figure you derive is a guess.
The disk a card discarded still counts — a member thrown out weeks ago may hold the only readable copy of some regions, which is why the images are set against one another, byte for byte. Nothing here is accepted on trust.
Default presets get Smart Array wrong — HP places its parity block a stripe off the position the usual tools assume, so the volume those tools hand back is one your server never had. It has to be assembled to HP's own scheme.
The cache battery is an early question — whether any write held in write-back ever made it to a platter settles what state the file system is actually in, and it is answered before anybody is handed an expectation to hold.

What those two buttons really do: Import treats the foreign description as the true one and writes it over whichever configuration the card holds now. Clear scrubs that description from every member and keeps nothing to restore it from. Each of them is a single keypress away, and nothing on screen warns you. With full copies in hand, though, neither of them can take a file off you. Imaging is not one stage among several on a job like this. Everything else rests on it.

How one of these actually went.

RG · RDG-2026-1966ON FILE ✓

A flat cache battery, and a practice database that opened whole

A long outage left the cache battery with nothing in it. Power returned, and the controller asked the practice a question no one on the staff could safely answer, so it went unanswered. None of the member disks was touched: each was cloned first, and the array's own metadata was lifted from the clones. That gave the parity order, the set was rebuilt over the copies, and the practice database opened whole — including the final write the cache had been holding.

The live database opened whole5 days from arrival to return

What helps, and what makes it worse.

First things to do

Photograph every screen the card puts up
Number each drive to the bay it came out of
Send every disk, including the apparently finished ones
Say what it ran: SQL, Exchange, guests, shares

What makes it worse

Answering Import or Clear at that prompt
Dropping in a spare and starting a rebuild
Updating the card's firmware at this point
Powering the server up before copies exist

Answers before you commit to anything.

Our server will not start. Have we lost the lot?

Rarely. The disks themselves are usually sound, and what has failed sits a layer above them — the controller's record of the set, the metadata written on the members, or a damaged file system. Keep the power off and give the console no answers, and you will still be in that same good position when the drives reach a bench.

'Foreign Configuration Detected' — what does that really mean?

It means the metadata written on the members no longer describes the array the controller expects to find. Change the card, reseat a member in the wrong bay, or let a failing controller carry on, and up comes the prompt. Stopping there is the card behaving properly, not a fault. Import and Clear both write to your disks, so take images before you touch either. With images on the shelf, neither answer can cost you a file.

Do you need the RAID controller as well as the disks?

No. Every member holds its own account of the array. PERC and MegaRAID write DDF, a published format; HP writes RIS structures of its own instead. Either one is read straight out of the images, and the set goes back together in software — so your card stays in the machine, and the server can be rebuilt around it afterwards.

Will our databases and virtual machines run again afterwards?

Usually, though this is two pieces of work rather than one. First the reassembled volume has to give up the VMDK, or the MDF, in one piece. Then a hypervisor has to be persuaded to mount the one, or SQL Server to attach the other. The database recovery page on this site covers that second half.

If this sounds like yours, start here.

When a RAID 5 rebuild goes wrong → RAID 1 mirrors → Fujitsu laptops → Every device we handle →

Nothing gets worse while it sits switched off.

You pay nothing for the look. What comes back is a list — the files that opened, the ones that did not, and one fixed figure to finish the job. Leave it switched off until you have that in writing.

0800 6890668

Get it looked at Posting it to Oxford

Drives, cards and whole machines

Start from the fault

Evidence, footage and reports

What it costs, and what happens