Open, and taking cases in · 9am–5:30pm, weekdays In a hurry? Call 0800 6890668
CDR Cardiff Data Recovery 0800 6890668 Book it in
CDR / Work we specialise in / Servers and RAID sets

Servers · broken RAID sets

Cardiff server data recovery. Another server can be ordered in the morning; the data on this one cannot.

Most servers arrive here with every disk still spinning. The failure is in what sits around them — a controller whose notes no longer match the drives, a prompt where either key does harm, a cache holding writes the mains never let it finish. Everything we do happens on copies, your card can stay where it is, and you get the volume back whole: shares, databases, guest machines. A solicitors' office in Cardiff and an engineering firm on Treforest Industrial Estate send us the same parcel.

No files back, no fee on most jobs One fixed price, after a free diagnosis It comes to us by post from Barry, Pontypridd or Newport

Describe the symptoms to an engineer
0800 6890668

What each server symptom is telling you.

Not listed here? Try the triage →
What shows upWhat has gone wrongDo this now
PERC prompt: Foreign Configuration DetectedWhat the disks say about the set no longer matches the card's own copy, so it halts by design — nothing is lost yetAnswer nothing on screen
1786 — Drive Array Recovery Needed (HP)A rebuild has begun and may have stopped halfwayPower it down; copy the disks first
Amber bay lights after a power cutWrites were cut off part-done and nobody knows what the cache still held, though the layout on the members is untouchedDo not bring it back online
The server boots but the data volume has vanishedThe trouble is in the card's own state or in the file system; the data underneath both is intactNormally recoverable
The RAID card has diedBay order, stripe size and parity direction are written on the disks themselvesNo matching card needed
Files came off, but no VM or database will openCopying is not the same as consistency, and only consistent data will runThat is a database recovery job
Posting it to us: post it tracked and fully insured to the intake lab, and the journey back is on us; ring first if you would like an engineer to check how it is packed before you seal the box. Every stage is set out on the posting guide.

The set carries its own blueprint.

Dell PERC & MegaRAIDDDF is an open specification, and both controllers write it into the tail sectors of each member. Lift it off the images and the entire set can be reassembled without the card ever being powered.
The HP Smart ArrayHP chose a different route. Its RIS structures live at the head of every member, and the parity block sits a stripe from where convention puts it. Point a default preset at that and the volume it produces is not yours.
The prompt with two bad answersAnswer this one early and you pay for it. Import can drive out-of-date metadata over stripes that were perfectly good; Clear removes the layout from every disk, with nothing to undo it. Image all the members first and the question loses its teeth.
Data stranded in the cacheThe controller marks a write complete once battery-backed memory holds it, long before any platter sees it. If the mains go while a flush is part-way through, the harm sits inside the file system, invisible. That alone is why we image.

From the parcel to your files coming back.

Cases we have logged →
01

A case number, and a diagnosis at no cost Free

The diagnosis costs you nothing. Every item that arrives is logged under a case number of its own, and an engineer works out what has actually failed — then tells you plainly which files can come back and which cannot. The price follows that: one fixed figure, in writing, and no work begins until you have read it and said go ahead.

Diagnosis at no chargeOne fixed figure in writingNothing owed yet
02

Leave everything as the failure left it

Switch the server off. As you pull each drive, write its bay number onto it before it goes in the box. Whatever the console wants, do not give it an answer. Here every member is copied singly on dedicated imaging hardware, the drives the card had already rejected included. Nothing is read as a file until those copies exist.

Every disk imaged separatelyThe written-off ones too
03

Reassemble the set without the card

The DDF or RIS records are sitting inside those copies. Read them and you have the RAID level, the stripe size, the bay each member came out of and — on RAID 5 and 6 — the direction the parity rotates. With that known, the set is assembled in software, and a dead card no longer counts for anything.

Geometry taken from the membersWorks without the original card
04

Copied is not enough — it has to open

Whichever file system was riding on top — NTFS or ReFS, XFS, ext4, a VMFS datastore — gets its repairs on the rebuilt copy. Your own drives are never written to. Then comes the test that settles it: guests boot, databases attach, shares mount, and every file inside them opens.

Repairs happen only on imagesTested until it boots and mounts
05

You approve it, then it goes back

Nothing is billed while you are still deciding. A complete listing of what came off the drive reaches you first, and the invoice only follows your yes. Files travel back on media bought new for the job, return postage ours, and we do not close it here until every one of them opens on the machine you will use.

The file list is yours to approveBack on media bought newThe postage back is on us

What we usually find first

  • Which end of the disk holds the map — HP puts RIS at the start of each member; PERC and MegaRAID write DDF into the final sectors. Search the wrong end and the geometry becomes guesswork.
  • The drive the card threw out still counts — a member ejected a month back may carry the most recent version of some regions, so every image is set against the others block by block before any one of them is believed.
  • Default settings mis-read Smart Array sets — HP shifts the parity block one stripe from where a generic tool goes looking, so that tool reassembles a volume nobody ever had. The rebuild has to be done HP's way.
  • The cache battery is asked about early — whether write-back data made it down to the platters is what governs the state the file system arrives in, and we establish that before giving anyone an expectation.

What the two keys actually do: Import takes the foreign metadata at its word and lays it over the configuration the card is running; Clear strips that map from every member, permanently. Either is one keypress away, and the screen offers no way back — but with full images already made, neither can cost you anything. Imaging is not a first precaution. It is the whole method.

Straight out of the casebook.

CF · CDF-2026-0816LOGGED ✓

The battery ran out, and the ProLiant wanted an answer

With the power off long enough to flatten the cache battery, the controller came up asking the practice to choose between options none of them could safely answer. Nothing was answered. Each member went to a copy first, the RIS metadata was lifted off those copies, the parity delay was worked into the sums, and the set went back together on images alone. The database behind the practice software opened complete, as far as the last write the cache got down.

Database opened cleanly5 days after arrival

Before you send it off.

Start with these

  • Get a photo of every screen the controller shows
  • Number each drive to match the bay it sat in
  • Send every disk in the set, the failed-looking ones included
  • Tell us what was on it — SQL, Exchange, VMs, file shares

What to avoid

  • Choosing Import or Clear on that screen
  • Slotting in a spare and starting a rebuild
  • Updating the RAID card's firmware at this stage
  • Switching the server on again before the disks are copied

What we get asked, week in, week out.

The server is dead. Have we lost everything on it?

Rarely. Most of the time the disks themselves are unharmed and the damage sits a layer up — in the controller's own notes, in the map of the set, or in a file system knocked out of shape. Keep the power off, answer nothing on the console, and you are still in a good position.

'Foreign Configuration Detected' — what is the card telling me?

That the arrangement it found on the disks is not the arrangement it expects. A replacement board, a drive returned to the wrong bay, or a card starting to fail will all produce it. Halting there is deliberate, and it is protecting you. Both Import and Clear can destroy data; with full copies on the bench, neither one can touch what matters.

Do we need to send the RAID card as well?

No. The layout is written on the members themselves — DDF, an open standard, on PERC and MegaRAID; HP uses its own RIS blocks. We read that off the bench images and rebuild the set in software, so no matching controller is ever needed.

Do databases and virtual machines run again, or just files?

Both — but treat it as two jobs. One is lifting the VMDK or the MDF clear of the volume we have rebuilt. The other is getting a hypervisor to mount that VMDK, or SQL Server to attach that MDF; the database recovery page covers that stage.

Pages people often read next.

RAID 5 rebuilds → Crypto wallets → Mail and databases → Devices we take in →

An unplugged drive keeps whatever it still holds.

A failing drive has only so many starts left in it. Spend none of them. Keep the power off, and the free diagnosis will tell you what is still readable.

0800 6890668

Get it checked Where to post it