The array had been running degraded for months, then the second disk died twenty minutes into the rebuild, or the controller failed and took its configuration with it. RAID array recovery is a key part of the work on this bench. RAID 0, 1, 5, 6 and 10 sets from servers, workstations and NAS devices across Swindon and the M4 corridor are imaged first and then rebuilt virtually, and cases where a business is down go to the front of the queue, including urgent RAID server data recovery for Wiltshire businesses.
Free diagnostic included on every RAID array job. Every client receives a single fixed quote in writing before any work is done on their array.
In most cases you only pay if we recover your data. Electronic and mechanical failures and chip-level work are 50% upfront, with the balance only if we recover your data, while BitLocker, ransomware, DVR and forensic jobs are paid upfront once you accept the fixed quote. You'll find full pricing on the data recovery cost page.
Every RAID array job starts by matching the symptoms to the actual fault. These twenty scenarios account for almost everything brought to the bench.
An array is designed to lose one member. It is the rebuild afterwards which kills the weakest surviving member, and that is the defining disaster of RAID array recovery.
If a rebuild stops halfway, the result poisons the parity that the rebuild depended on, and the extent of the damage is worked out from the images.
If a controller card fails, it also takes the configuration with it, so the geometry is worked out again from the individual members, without needing a controller card.
A member holding months-old data is forced back into the array, and the file system tears itself apart because of the mismatch.
A new array has been created on top of the existing one. Stop it as soon as possible, before the initialisation process finishes, and plenty will survive.
Several disks can fail from a single power spike, so each disk is repaired and imaged separately before the calculations start.
Whole banks of disks can appear to have "failed" at the same time when the actual problem is damaged wiring in the chassis. So we diagnose before raising the alarm.
The stripe maths is spot on; however, the file system (NTFS, ext4 or VMFS) above it isn't, and rebuilding at the parser level wins it back.
When array definitions are accidentally deleted, they can be restored, because the blocks those definitions described are still intact.
Flat cache batteries combined with a power cut put parity out of step with the data. Parity and data are reconciled at the image level.
An H710 or P420 card can fail with the array definition still stored on it, but since the layout can be recovered from the members, no replacement card is necessary.
A member with dead heads goes to the head-swap bench first and then joins the queue with the other members for imaging.
ST8000VN004s from the same batch installed in small business arrays will age together, so imaging them in parallel lets us stay ahead of the second disk failure.
Arrays built in software can lose superblocks and pool metadata. These are rebuilt structure by structure from the individual raw members.
A member that remaps sectors every day is failing in slow motion, and catching it when it is imaged saves the whole set.
Losing both sides of a single mirror defeats the redundancy, but lab imaging often brings one side back far enough to complete the stripe.
The auto-rebuild lands on a drive that has been sitting spinning idle as a hot spare for years, slowly dying the whole time.
Clicking "import foreign configuration" against the incorrect metadata will scramble your set, and the actual layout then has to be recovered from the individual members.
Blues and Barracudas have no limit on error recovery, so they drop out of arrays under heavy load, which is why we image each one patiently before the array is rebuilt.
A power cut can corrupt pool metadata on hypervisor builds. We work our way back through the transaction history using full member images.
Before anyone starts guessing what went wrong, we capture all of the information on every single member drive, and we don't allow the original drives to be part of another array after they are captured. Next, from those captures we work out the geometry of the array (member order, stripe size, rotation and parity offset) and then create a virtual version of the array on top. After that, the file system can be examined by walking through that virtual array. Stale members in a RAID 5, double losses in RAID 6, nested RAID 10 sets and controllers that forget everything are just about normal for us in RAID array recovery, and that's an ordinary week here. If the controller card itself dies, however, RAID controller repair matters less than the owner fears, because the data is recovered from images taken of the individual member drives rather than from the controller.
There are a few things people do that turn what was once a recoverable array into scrap more often than the actual failure did, namely forcing an offline member back into the set, rebuilding onto a disk that is already dying and letting someone have one last experiment. If your array is down, shut it off and mark each of the drives with its bay location. Then call us. If you tell our staff that the business has come to a complete stop, your assessment will move to the front of the queue and should be finished within 2 working days of us receiving your drives.
There is one principle that RAID array recovery can't do without, which is to always image everything first, and this bench sticks strictly to that principle:
We image all of the member drives at the same time on dedicated hardware before analysing anything. Once that's done, the original drives never take part in any rebuild.
Works on the images to establish member order, stripe size, rotation and offsets, then mounts the assembled array virtually.
When one member of the array fails, it gets the full hard-drive treatment (heads, firmware and electronics) so that it can be imaged alongside the healthy ones.
The solved geometry is confirmed with entropy and parity checks before a single file is lifted from the assembled volume.
NTFS, ReFS, ext4, XFS, BTRFS and VMFS are all opened natively, including virtual machine disks stored on top.
Member imaging is physically enforced as read-only, so your drives come back to you untouched.
Hardware card, software layer or chipset RAID: the geometry lives on the member drives, and the members are where it gets read from. PERC and Smart Array cards tend to die with the array definition still on board, matched-batch NAS drives such as the ST8000VN004 tend to fail together, and desktop-grade members drop out of sets under load. They're three different roads to the same bench, and all end in parallel imaging once the set is booked in.
Power the server or shelf down before touching anything, then mark each member with its slot as it comes out, because the reconstruction relies on knowing the original order. Only the labelled drives travel; the controller and chassis stay with you, and a quick phone photo of the bays as they stood is worth taking. Get a quote first, by phone (0800 689 0668) or with the free online diagnostic, before anything is posted.
Most customers get a quote first, by phone or with the free online diagnostic, and then post or courier their media to us.
Sending a drive from a computer, laptop, MacBook, iMac, CCTV / DVR or server? Please remove the internal hard drive or SSD and send us just the drive (from a server, every disk, each labelled with its bay), as we don't offer a drive-removal service. We don't recover storage soldered to a motherboard (for example T2 and Apple Silicon Macs and some thin laptops), only drives that can be removed and sent to us. If you're not sure which kind you have, ring us first.
↓ Print the booking-in & shipping form (PDF)
Mark the package for the attention of Oxford Data Recovery and we'll ring you as soon as we've diagnosed your media.
Not sure what to send? Ring 0800 689 0668 first or use the free online diagnostic.
Free diagnostic, fixed quote, and in most cases you only pay if we recover your data. Start now or call the freephone.