Taking work now — the first look is freeSpinning disks in a machine Quicker still, give us a ring:0203 0868633
ADR Advanced Data Recovery 0203 0868633 Price my job
ADR / Media and machines / Server arrays and RAID sets

Servers · arrays that stopped short

Server data recovery, London. The server will not boot. Nine times in ten the platters never stopped working.

Healthy platters with broken machinery around them: that is what most servers turn out to be once they are on the bench. A controller whose idea of the set no longer matches the bays. A console prompt with two expensive answers. A cache still holding writes that never reached a platter when the mains went. Every member is cloned before anything else happens, all the later work runs on those clones, and your card never comes out of the chassis. Back comes the volume entire — shares, mail stores, guests, databases, the lot. Arrays, NAS units and servers are priced from £500 + VAT, and the free diagnosis settles that figure in writing. Food factories at Park Royal keep traceability records on a small server; a wholesaler on an Enfield estate keeps stock and ledgers on another. On this bench they look much the same.

On most jobs: no data, no bill A free look first, then one figure in writing Media posted in from Croydon, Ealing and Camden

Put it to an engineer — the first look costs nothing
0203 0868633

Servers: what the symptom usually means.

Different fault? Try the finder →
What is going onThe usual reason for itWhere that leaves you
PERC, showing Foreign Configuration DetectedCard and members no longer tell the same story about the set, so it stops instead of guessing. Nothing is gone at this stageLeave the prompt unanswered
1786 on an HP console: Drive Array Recovery NeededA rebuild was started and may have stopped in the middle of itOff at the wall. Clones come first
Amber lights across the bays after a power cutWrites were severed halfway and nobody knows what the cache was still carrying. The set's own description of itself survivesLeave the mains off
It boots, and the data volume is goneEither the card's configuration or the file system above it. Underneath both, the sectors are still thereComes back, usually
The controller itself has given outThe members themselves record which bay, what size of stripe, and where the parity wentA matching card is not needed
Files copied out, but no guest and no database will openA copy that exists is not the same as a copy that is consistent, and only the consistent kind opensThat becomes database recovery work
Packing it and posting it:pad the box until nothing in it can move, base the declared value on the data instead of the hardware, and choose a tracked service to our intake lab. Return postage is at our expense. Would you sooner an engineer checked the packing before the box is sealed? Ring us first. The full account of it sits on theguide to packing and posting.

Members carry the map of the set.

PERC and MegaRAID cardsBoth cards write DDF, an open published layout, into the last sectors a member has. Clones surrender it easily, the set is then rebuilt from those, and your controller spends the whole job where it has always been.
HP and its Smart Array cardsHP does it another way. RIS structures sit at the head of a member, and the parity block is one stripe clear of where most tools go hunting for it. Aim a default preset at an HP set and what you get back is a volume the server never ran.
A prompt with no safe answerThis is the screen to walk away from. Stale metadata can be written by Import straight over stripes that were perfectly good, while Clear wipes the description from each disk and holds nothing in reserve to rebuild it from. Clone first and the prompt loses its teeth.
Writes the cache never got outAs soon as battery-backed memory accepts a write, that write is reported complete, long before a platter ever sees it. Interrupt the mains during a flush and the file system itself breaks, silently, with nothing on screen to tell you. Which is why cloning comes before every other step.

From the box arriving to your files going back.

Work we have closed →
01

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

A case number goes on it the day the parcel lands, and an engineer settles what has truly failed before anything else happens — free of charge, and first in the order of work. Back to you come two things together: a straight note of what is liftable and what is not, plus one figure, fixed and written down. Accept it, or decline and owe us nothing.

Nothing to pay for lookingA single figure, put in writingNothing owed at this stage
02

Nothing is unbolted from the server

Shut the machine down. Take the drives out singly, marking each with the bay it occupied before it goes anywhere. Whatever the console asks, leave it alone. At this end, dedicated imaging hardware works through the members one by one, the rejected ones as well, and nobody opens a file until the last of those clones is complete.

Each member cloned on its ownMembers the controller had dropped included
03

The set is assembled in software

The DDF or RIS records come across with the clones. Read them and the geometry falls out: RAID level, stripe size, the bay each member sat in, and the direction parity travels where there is parity. Software then puts the volume back together from that, and a controller that has given out stops mattering to anybody.

The geometry, off the membersYour card stays in the chassis
04

A clone is not a working volume yet.

Repairs are made to the reassembled clone and never to the drives you sent, whatever the set was carrying — NTFS, ReFS, ext4, XFS or a VMFS datastore. Then it gets tested in earnest: guests booted, shares brought up, databases attached, single files opened and read end to end.

Only the clones are repairedBooted and mounted as proof
05

You see the file list before you pay

What was recovered is listed for you first, and only then does a bill exist. Approve the list and it is invoiced; turn it down and it is not — and where nothing has come back, most jobs carry no charge whatever. Recovered data travels home on fresh media bought in for your job, with the postage at our end. Your case is not closed until you have read those files on a machine of your own.

No charge until you accept the figureFresh media, supplied with the jobThe post home is ours

What arrives most often

  • The map is not in the same place on every make — HP puts RIS structures at the front of a member; DDF, which is what PERC and MegaRAID write, lives at the far end. Read from the wrong end of a member and every figure you take away is guesswork.
  • A member the card threw out still matters — a disk dropped weeks ago can be the only one where a particular region still reads, which is why the clones are compared against each other byte for byte. Nothing on this bench goes through on trust.
  • Standard presets misread Smart Array — an HP set keeps its parity block a stripe away from where ordinary tools expect to find it. Assemble on those defaults and you get a volume your server never ran. HP's own scheme has to be followed instead.
  • The cache battery gets asked about early on — whether write-back was still holding writes that no platter ever received decides the true state of the file system, and that is settled before you are given anything to expect.

What those two answers really do: Import reads the foreign description as gospel and lays it over whatever configuration the card is holding. Clear goes the other way, scrubbing that description off each member with nothing kept in reserve to rebuild it. One keypress does either, and the screen warns you of neither. Hold full clones and no reply can rob you of a file. On work of this kind, cloning is not a stage — the rest of the job stands on top of it. The band for arrays, NAS units and servers opens at £500 + VAT, written down once the free look is over, and where no data comes back most jobs carry no charge.

One job, followed all the way through.

LDN · ADR-2026-3366JOB LOGGED ✓

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

The cache battery had been emptied by a long power cut. With the mains restored, the controller asked the office something nobody there could answer safely, so it went unanswered. Not one member disk was touched directly. Every disk was cloned instead, and the array's own metadata was read from those clones, which fixed the parity order. The rebuild ran across the copies, and what came up at the end was the practice database entire, the write still sitting in the cache included.

That databaseopened up whole5days here, and back to you

What helps, and what harms.

Do this much first

  • Photograph the controller screen before anything else
  • Write the bay number on each drive
  • Send every member, the ones that look fine included
  • Tell us what it ran: SQL, Exchange, shares, guests

What sets us back

  • Pressing Import, or pressing Clear
  • Slotting a spare in and letting a rebuild run
  • Putting new firmware on the card at this stage
  • Powering the machine up again before clones exist

Questions answered before you commit.

Our server stops before the volume shows. Is it all gone?

Seldom. Three things sit between you and your files, and the platters are the bottom one of the three. Above them is the description each member keeps of the set; above that, the card's own copy of it; above both, a file system. Damage almost always lands in the upper two, and both are reconstructable. Two habits protect the position you are in: mains off, and no reply given to any prompt on the console.

It says Foreign Configuration Detected. What does that mean?

It means what the members say about themselves no longer matches what the card expects to find. A replacement card, a member returned to the wrong bay, a controller going quietly unreliable — any of those brings the prompt up. The halt is correct behaviour and not itself a fault. Both replies write to your disks, so have the clones made first. Once they exist, nothing anybody presses on that screen can cost you a file.

Must the card come too, or only the drives?

Only the drives. Each member is carrying the description of the set. PERC and MegaRAID use DDF, a published layout anyone can parse; HP uses RIS structures of its own. Both come out of the clones, and the reassembly happens in software from there — meaning your card need never leave the chassis and the machine can be rebuilt around it whenever suits you.

Will the guests and the databases actually run again?

As a rule, yes, though count it as two separate jobs. Job one ends when the reassembled volume gives up an intact VMDK or an intact MDF. Job two is getting that VMDK mounted by a hypervisor, or that MDF attached by SQL Server, and the database recovery page here covers what is involved.

Begin here if yours is doing the same thing.

When a RAID 5 rebuild fails → A mirrored RAID 1 pair → Toshiba laptop drives → Every device we take →

Nothing gets worse while the power is off.

Looking at it is free. Back comes a list of what opened and what did not, together with a single price to finish, set down in writing while you are still free to say no. On most jobs an invoice only follows the data. Until that list reaches you, leave the drive switched off.

0203 0868633

Get it assessed Sending it to Guildford