Taking work now — the first look is freeWhole sets posted in from anywhere in the UK, or handed in at ten drop-off pointsQuicker still, give us a ring:0800 6890668
RARRAID Array Data Recovery 0800 6890668 Price my job

RAID array data recovery · UK-wide, by post · hardware, software and virtualised

The array has failed. The data is still on the drives.

A degraded, failed or mis-rebuilt RAID set is rarely a lost one. What loses the data is what happens next: a rebuild that reads every sector of survivors bought the same day as the drive that died, a foreign configuration cleared to tidy the screen, F2 pressed at an HP 1779 prompt, a stale drive forced online, a pool rewound with -F, chkdsk run on a volume assembled with the wrong stripe. Power down, label the slots, and post every member. We image each one, read the geometry from the controller's metadata or recover it from the parity, assemble the set in software from the images, and repair the file system on a copy of the virtual volume, never on the originals. A set of two to four members is £500 + VAT upwards. Larger sets, 15K SAS sets, storage shelves and enterprise storage start at £1,250 + VAT. The first look is free, the figure is given in writing, and on most jobs no data means no bill.

Put it to an engineer and hear what it would take
0800 6890668

The look is free · a reference for this checkCHECK-····

What kind of array is it?

Choose the closest match. You get a likely cause, the band it falls in, and the one thing not to do tonight.

Free, and nothing to payAnswered inside a working day
£500 + VATA set of two to four members
FreeThe first look
5–10 daysAt the bench, for a small set
UK-wideBy tracked post
Every member imaged; assembled in softwareOne fixed figure in writingMost jobs: no data, no billNothing written to the originalsReturn postage paidICO-registered · ZC173784

Four array jobs closed, and what went home.

Jobs already closed →
UK · RAR-2026-0601JOB LOGGED ✓

A six-drive RAID 5 on a PERC H730, one member failed, a second dropped at 61 per cent of the rebuild, and the virtual disk marked Failed with a firm's file server on it

The owner powered down and sent all six drives, including the first to fail. The two dropped members were imaged, one after a head swap and the other slowly; the DDF metadata gave the order and a 64KB stripe, parity across the images confirmed it, and the set was assembled from the four survivors and the newer failed member, with the first-dropped drive's stripes used where the other could not read. The NTFS volume mounted from the virtual array.

99.98% of the volume recovered8 days here, and back by post
Illustrative example — replace with a genuine case
UK · RAR-2026-0616JOB LOGGED ✓

A DL380 Gen8 with a P420i that stopped at 1779 after a drive that had failed in April was reseated in July, and an engineer who pressed F2 to get the server up

The logical drive came up with April's data in one member's stripes and ran for a day before the corruption was noticed. The RIS metadata on the images identified the reseated drive as three months stale; the five current members were assembled in software and the NTFS volume repaired from its journal, with the day's damage limited to files written after F2.

99.9% of the volume recovered8 days here, and back by post
Illustrative example — replace with a genuine case
UK · RAR-2026-0606JOB LOGGED ✓

A twenty-four-drive RAID 60 in an MD1400 behind a PERC H840, an expander failure that dropped the shelf, and a clear of the foreign configuration by an engineer who wanted the server to boot

Every drive was healthy and no initialisation had run. All twenty-four were imaged, parity and Q computed across candidate groups to sort the members into their two spans and prove each span's geometry, and the span stripe found from the file system's anchors. The set was assembled in software and the file server's volume copied out.

100% of the volume recovered14 days here, and back by post
Illustrative example — replace with a genuine case
UK · RAR-2026-0621JOB LOGGED ✓

A Proxmox host's six-drive RAID-Z2 pool, SUSPENDED after a power cut, then FAULTED with insufficient replicas when two devices came back UNAVAIL, and an owner who had read about -F and stopped short of typing it

The two UNAVAIL devices had damaged labels at the start of the disk; their end labels were intact and read from the images. The pool was opened read-only from the images at the newest uberblock every member agreed on, two transaction groups behind the survivors' newest, and the datasets and zvols copied out. The two lost groups amounted to seconds of writes.

99.99% of the pool recovered8 days here, and back by post
Illustrative example — replace with a genuine case

Begin with the level,
or with what the controller is saying.

Or begin with the controller or the host →

Every level and layout

RAID 5One parity, one failure, and the rebuild that finds the second→RAID 6P and Q, two failures, and the third that ends it→RAID 10Mirror pairs striped: which drives were pairs, which copy is newest→RAID 0No parity: order and stripe from the data itself→RAID 50 and RAID 60Spans in shelves; per-span geometry, then the stripe across them→RAID-Z1, Z2 and Z3ZFS: labels, uberblocks and the rewind that belongs on images→Storage SpacesSlabs across columns, and the states Microsoft names→

Servers and virtual machines, a layer at a time.

The whole section →

Servers and virtual machines

VMware data recoveryESXi datastores, VMFS and vSAN: the set, the datastore, then each machine→ESXi datastore and VMFSInactive, not mounted or reformatted: the metadata repaired on a copy→Virtual machines, VMDK and VHDXBroken snapshot chains and deleted machines; every guest opened→Hyper-V data recoveryCluster shared volumes, ReFS, checkpoints and Storage Spaces Direct→Server data recoveryPowerEdge, ProLiant, ThinkSystem: Windows or Linux, databases checked→

The two things that decide a job

How many members still read, and what was done before the parcel.

The level, the controller and the size of the set matter less than two things. The first is whether enough members can be imaged, including the one that failed first, which holds every stripe a stalled rebuild never reached. The second is what happened between the failure and the parcel: the rebuild, the clear, the F2, the force online, the -F, the recreate, the chkdsk. Each writes to the drives; on images each would have been reversible. The pages on this site say, in each vendor's own words, which button that is.

order · stripe · rotation · offset · stale member

The geometry, and where it is read from

A hardware controller writes the set's description to every member, DDF near the end on PERC, MegaRAID and Intel, RIS on Smart Array; mdadm keeps a superblock in one of four places; ZFS keeps four labels; Storage Spaces a pool database. Where it survives, the bench reads it. Where it was cleared or overwritten, parity across the members at the same offset should XOR to zero on a consistent stripe, which proves the level and finds the stale member; where parity lands says the rotation; the file system's own structures give the stripe size, the order and the start.

  • Metadata first, parity second, file system third
  • RAID 6 proved by P and Q together
  • RAID 0 from the data alone
  • Nothing written to any image
How a set is rebuilt from images →
rebuild · clear · F2 · force online · -F · --create · chkdsk

The buttons, in the vendors' own words

HPE's 1779 prompt offers F2 to accept data loss and to re-enable logical drive(s). Dell defines a puncture as rebuild with errors. Synology says a crashed pool can no longer be repaired by yourself. QNAP says Recover RAID does not help in the event of disk failure. Microsoft defines No redundancy as having lost data because too many drives failed, and in 2020 told owners of RAW parity spaces not to run chkdsk. Each of those is quoted on its own page, with the document number, and each page says what to do tonight instead.

  • Twelve messages, each with the button to leave alone
  • Advisory numbers and dates where they exist
  • The first-response page in one place
  • A set of two to four from £500 + VAT
What to do tonight →

What happens after it arrives.

All seven stages in order →

  1. Send the form

    Tell us the controller, the level, how many members, what it says and what was tried. Back by email come the posting address, the nearest drop-off point, packing notes and a booking sheet to print. The address is not published; it comes by email in reply to the form, and the booking sheet inside the parcel is what matches it to your enquiry when it is opened.

  2. Post every member, or hand them in

    Each drive with its slot written on it, in an anti-static bag in its own padding; the controller only if it was replaced or holds a key. Tracked post from anywhere in the UK, or handed in at the nearest of ten drop-off points; say where you are on the form and it comes by email with the posting details. Nobody collects, and we pay the postage home whichever way the job turns out.

  3. The free look

    Every member on the imager its interface needs, never on a controller: the metadata read from each, the level, the order, the event counts. An engineer settles what happened to the set and how much can honestly be read back. That stage carries no charge, and you may stop at it owing nothing.

  4. One price, written down

    Findings and price arrive together, and nothing chargeable begins until you have agreed to the figure.

  5. You get the files back

    The set assembled from the images, the file system repaired on a copy of the virtual volume, the file list approved by you, and the data home on fresh media. 5–10 days at the bench for a set of two to four, 10–15 days at the bench for larger sets. Your case stays open until you have opened the files yourself.

It has failed, and the rebuild button is not the answer tonight.

A rebuild reads every sector of every survivor on a set whose drives were bought together, and a cleared configuration deletes the set's description from drives that were fine. Power down, label the slots, send the form, and back comes an honest account of how much still reads, with the price of getting it home. That account is free, and so is the parcel returning.

0800 6890668

Price my jobPacking and posting