The array has run in a degraded mode for months prior to the second drive dying halfway through the rebuild, or perhaps the controller has died, taking the configuration with it. RAID 0, 1, 5, 6 and 10 sets from various servers, workstations and NAS enclosures across South Wales are rebuilt here, always images first, and urgent server data recovery for business clients is carried out before their turn comes up.
All RAID array jobs are given a free diagnostic prior to any of our work being carried out. Each job will be provided with one single fixed price in writing for all work involved in recovering your data, before any work begins.
In most cases you only pay if we recover your data. This is true of all our services except for failure of electrical/mechanical components, chip level repairs, DVRs and forensic analysis. Our full price list can be seen at the cost of data recovery page.
Each RAID array job starts by matching the symptom to what has gone wrong, the fault. These fifteen faults represent nearly every scenario we come across at the bench.
One loss is survived by the array; at this point, the rebuild then kills off the oldest survivor, which is the defining disaster.
If a rebuild does not complete, the rebuild process has corrupted the parity it depended on. Working from the images, we map out the boundary of where the corruption occurred and base the recovery on that boundary.
After a controller card dies, it takes the configuration with it, including the geometry of the array. Therefore, we derive the geometry of the array based solely on the individual drives.
A member goes out of date, then its old data returns to the set. The file system then destroys itself.
A new array is created over the top of the previous one. Stop before the initialisation process completes.
Two or three members fail at the same time. Each is repaired and imaged separately.
Multiple drives 'fail' in blocks as a result of bad backplane wiring. Diagnosis prior to panic is always best.
The maths of the array works perfectly. The issue lies at the NTFS/ext4/VMFS level above it. A rebuild at the parser level is the fix.
A definition removed in error is recovered, as long as the underlying blocks survive.
Dead cache batteries plus sudden power loss desynchronise parity from data, reconciled at image level.
The H710s and P420s fail while holding the configuration (definition), solved using the drives; no matching card required.
A member drive has failed heads, so it's repaired on the bench before being placed into the imaging queue.
Matched-batch ST4000VN006s in small business arrays are ageing in unison, so the members are all imaged in parallel before the next one fails.
When software arrays such as mdadm or Storage Spaces fail, the data structures for managing the array (superblocks and pool metadata) can be lost and need to be reconstructed from the raw members that were part of the original array.
A member that is remapping sectors every day is quietly failing; if we catch this failure while capturing an image, the set will be saved.
Rule number one will not bend: every member drive is captured sector for sector prior to attempting any reconstruction, and the original drives will never again be spinning together as an array. The geometry (member order, stripe size, rotation, offset) will be determined from the images themselves and the volume will be built virtually. The file system is then walked on top of this virtual build. Standard weekly workloads include stale members in a RAID 5, double failures in a RAID 6, nested 10s and controllers with memory loss.
Forcing an offline drive back in, rebuilding onto a failing disk, and letting someone take a final shot at it: these three turn recoverable arrays into a dead loss more surely than any hardware problem. If the array is down: power it off, label each drive with its bay, then call. And say 'business-critical'; this will bump your diagnostic to the top of the line when the drives arrive.
There is one law for RAID recovery: image everything first, and the bench enforces it:
Each of your disk drives is imaged at the same time using dedicated equipment before we begin analysing them, so when it comes to rebuilding, no original drive ever gets reused.
It runs against the images to break down the geometry (order of members, block size, rotation, offsets), then virtually mounts the entire array.
The failed members of a disk group will have to undergo the full hard drive treatment first (heads, firmware, electronics), as they will then be able to be imaged alongside the other working members.
After checking parity and entropy to confirm the solved geometry, it can then extract a file system from the entire reconstructed volume.
NTFS, ReFS, ext4, XFS, BTRFS and VMFS file systems are read natively, as well as the virtual machine disks stored on top.
Member images are captured read-only; the drives you have sent to us will go back to you just like you sent them.
Hardware card, software layer, motherboard chipset: the geometry is written across the member drives, and that's where we read it from. Controllers such as PERC and Smart Array have a habit of dying while still holding the array definition, and matched-batch NAS drives (IronWolf ST4000VN006s especially) are well known to wear out in formation, which is exactly why members are imaged in parallel the day they arrive.
Shut down the server or array before you remove anything, and label each drive with its slot as it comes out of the system, as the original sequence is necessary to rebuild. Once you have your quote then post the labelled drives only; controllers and chassis stay with you, and a quick phone photo of how the bays were configured will also be very helpful.
First get a quote over the telephone or use our free on-line diagnostic. The majority of customers then have their media sent to us via post, this is tracked and insured.
Do you want to send a drive from your computer, laptop, MacBook, iMac, CCTV / DVR or server? Please do not send the complete unit of computer or device with a hard disk drive (HDD) or solid state drive (SSD) inside. Please remove it and only send the drive itself. For a server, please remove all disks and label them by their bay. We will also not perform removal of internal drives from devices in order to access data. Also, we will not extract data from devices where storage is soldered onto the motherboard (such as T2 and Apple Silicon Macs and some thin laptops). Additionally, we do not recover data from phones and tablets, 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)
Clearly mark the parcel with the name Bristol Data Recovery, and when your media has been diagnosed we will call you immediately.
If unsure about what to post, please call us first on 0800 689 0668 (the free phone number), or complete a free online diagnostic.
Diagnostic provided at no charge. Fixed quote. In most cases you only pay if we recover your data. Start now or call the freephone.