A RAID 5 is designed to lose one disk and keep working, which is exactly why so many arrays reach this bench having quietly run degraded for months. Then a second member goes, or a rebuild stalls halfway, and a volume that was never backed up because it was thought to be safe stops existing. This page explains what parity can and cannot do, why the rebuild is the dangerous moment, and how a five-disk set is put back together on copies.
RAID 5 recovery starts at £500 + VAT and rises with the number of members, quoted in writing once the free diagnostic closes two working days after the disks are booked in. Send the member drives labelled with their bay order — not the chassis, not the controller.
Every RAID 5 that reaches this bench is in one of about six conditions, and the condition decides both how much work the job is and how much of it can still be undone. Read down until you recognise yours, then stop touching the array.
One member has dropped out and parity is carrying the load. Every byte is still there and nothing has been lost yet. This is the best moment to act and the moment almost everybody wastes, because the share still opens and nobody feels the emergency. Copy your live data off to somewhere else today, before anything is rebuilt.
Parity in RAID 5 covers exactly one absent disk. Lose a second and the arithmetic runs out, the controller drops the set, and the volume disappears. It is not lost. In most of these the second disk failed only partly, so an image of it plus parity from the rest reconstructs the array off-line, on copies, with your originals untouched.
Rebuilding reads every sector of every surviving member from end to end, which is the heaviest work those disks have done in years. A second marginal drive frequently gives up somewhere in the middle. If a rebuild has stopped at 40 per cent, pull the power rather than restarting it, because a restart begins the same punishing read all over again.
A card that has been replaced, reflashed, cleared or moved into another chassis comes up announcing a foreign configuration, or offers to make a fresh array out of your disks. The layout it has forgotten can be worked out again from the data on the members themselves, so long as nobody accepts the offer to initialise.
Under pressure, at two in the morning, a well-meaning person creates a new array from the same disks to see whether the volume comes back. Sometimes that costs nothing and sometimes it writes a fresh parity set across live data. It is still worth a diagnostic, because how much it cost depends entirely on which initialisation option was chosen.
A healthy array can still lose a volume: a deleted LUN, a formatted partition, a corrupt file system after a power cut, a ransomware run across the shares. No disk needs opening for any of that, the work is done on images all the same, and the price band is the array band rather than the single-disk one.
RAID 5 spreads a volume across three or more disks and, for every stripe of data it writes, it also writes a parity block on one of them. Parity is not a copy. It is the exclusive-or of the other blocks in that stripe, a single value that lets any one missing block be worked out from the ones that remain. The position of the parity block moves from stripe to stripe so no single disk becomes a bottleneck, which is why an array of five 4TB disks gives you 16TB rather than 20TB. The missing quarter is the insurance premium.
The insurance covers precisely one absent disk. With one member gone the controller reconstructs every read on the fly from the survivors, which is what a degraded array is: complete data, no protection left, and a noticeable drop in speed. Take a second member away and each stripe now has two unknowns and one equation. Nothing on this earth resolves that from parity alone, which is why the array goes offline rather than limping on.
That is also why a RAID 5 is not a backup, and it is worth saying plainly because the belief that it is causes most of the calls here. Parity protects against a disk dying. It does nothing at all about a file deleted at half past four, a database corrupted by a bad shutdown, a formatted volume, a fire in the comms cupboard or ransomware working steadily through the shares. All of those are written faithfully to every member at once.
Swapping in a replacement disk and pressing rebuild feels like the safe, obvious response. It is the single most common way a one-disk problem becomes a two-disk disaster, for reasons that are pure arithmetic. To rebuild, the controller must read every sector of every remaining member without a single unrecoverable error. On a five-disk set of large drives that is many terabytes of continuous reading, sustained for hours or days, on disks that were bought in one batch on one day and have spun the same hours in the same warm rack ever since. If one of them is going to fail this year, this is the workload that finds out.
Worse, most rebuilds run while the array is still serving users, so the disks are doing that heavy reading on top of the day job. When a second member drops out mid-rebuild, you now have a partly written replacement, an array that will not assemble, and a set of disks whose contents no longer agree with each other. It is still recoverable in the great majority of cases, but it is a longer job than it needed to be.
The safe order is dull and it works. Get a copy of the live data off the degraded array first, to anything with enough space. Only then consider the rebuild. If the data has any real value and no verified backup exists, image the members before the rebuild rather than after it.
Do not force members back online repeatedly. Each forced assembly writes new metadata and can update the sequence numbers the controller uses to work out which disk holds stale data, and those numbers are frequently what tells a laboratory the correct disk order.
Do not initialise, and do not accept any dialogue offering to create a new array, clear a foreign configuration or write a new configuration. A quick initialise usually touches only the metadata and is survivable. A full initialise writes zeros across the members and is not.
Do not shuffle the disks. Pulling them all out to inspect the labels and putting them back in a different order turns a routine job into a puzzle. If they are already out and the order is lost, say so — it can be worked out from the data — but it costs time that need not be spent.
Do not run a file system repair against the array volume. CHKDSK, fsck and the various commercial repair utilities all assume the storage underneath them is sound. Pointed at an array assembled in the wrong order they will write their conclusions over structures that were perfectly intact, and those writes land on every member at once.
Do not keep a clicking or grinding member powered up in the hope it settles. It will not. A disk making mechanical noise loses recoverable surface every minute it spins, and in a RAID 5 that disk is often the one holding the parity you need.
Nothing is assembled, mounted or repaired on your disks. Every member is imaged individually onto lab storage, using hardware built for drives that stall, drop off the bus mid-read or take thirty seconds to answer for a single sector. Healthy members image quickly. A member with weak areas is read in passes, the good regions banked first and the difficult ones revisited afterwards with shorter timeouts and head-by-head tactics, so that a disk with a limited number of working hours left spends them on the material nobody has yet.
Once the set is imaged, your drives go on a shelf and are not touched again. Everything that follows happens against copies, which means a wrong turn costs an hour of computing rather than your data, and a second attempt is always possible.
An array is a set of rules for turning several disks into one volume, and those rules have to be established before a single file appears. Four things have to be right: the stripe or block size, the order the members sit in, the direction parity rotates and whether it is synchronous or asynchronous, and the offset at which the data area begins on each disk. Get one wrong and the volume either refuses to mount or mounts and hands you files that are quietly full of somebody else's blocks — the failure mode to fear, because it looks like success.
Controller metadata helps and is never taken on trust. Dell PERC, HP Smart Array, LSI and Adaptec cards each write their configuration to a reserved area on the members, Linux software arrays write an md superblock, and there is a common on-disk format that several vendors share. All of it is read, and then the parameters are proved independently against the data itself: file system structures landing where the geometry predicts, parity that actually validates across the stripe, directory records that resolve to real files. When the metadata and the data disagree, the data wins.
Where one member is missing or badly damaged, it is not needed in full. The rules above let the missing blocks be computed from the survivors, so a disk that images at 80 per cent still contributes, and the gaps are filled by parity. That is the whole point of the design, used the way it was meant to be used: once, carefully, on copies, rather than under load on live hardware.
The usual outcome on a two-disk failure is the entire volume with its folder structure and file names intact, because the file system was never damaged in the first place — the array simply could not be assembled. Deleted files, a formatted volume or a ransomware run are more variable, and the diagnostic will say which of those you have before a figure is agreed.
Three situations genuinely end badly and it is fairer to name them here than to discover them at your expense. A rebuild that ran to completion onto a fresh disk with the array assembled in the wrong order overwrites as it goes. A full initialise, as opposed to a quick one, writes zeros over the data area. And blocks that were never written, on a thin-provisioned volume, contain nothing to find. Everything else is a question of how long it takes rather than whether it is possible.
Send the member drives on their own. The chassis, the rails, the controller card and the power supplies stay with you, because none of them are used at this end and they turn a small parcel into a freight problem. Label each disk with the bay it came out of — 1, 2, 3, 4 in the order they sat in the front of the unit — and photograph the front of the chassis before you pull anything, so the labels can be checked against reality. That photograph occasionally saves a day.
Pack them so nothing can move: individually wrapped, in a rigid box, with padding between them. Include the shipping and booking-in form, a note of the controller model and, if you know it, the array type and the order in which members dropped out. Royal Mail Special Delivery is tracked and insured and is what most businesses use. If you would rather hand them over, the bench takes drop-offs at the Cambridge address, Mon–Fri 9:00am–5:30pm — about 70 miles from Leicester, down the M1 to J19 and east on the A14, roughly an hour and a half, and the building sits two minutes off Junction 32 with parking outside the door. What does not exist anywhere in this network is a collection service or a counter in Leicester.
Array work starts at £500 + VAT and rises with the number of members, because each disk earns an image of its own and a five-disk set is genuinely more work than a three-disk one. That covers RAID of any level, NAS boxes, SAN volumes and servers. A single disk or SSD outside an array is £300 + VAT, memory cards and USB sticks are £250 + VAT, a CCTV or DVR recorder disk is £400 + VAT, and an encrypted volume where you can supply the key is also £400 + VAT. Forensic work with a written report is £800 + VAT, and a forensic binary image with deleted-file extraction and no report is £400 + VAT. Ransomware on an array is priced as ordinary array work from £500 + VAT: it is a recovery job, not a forensic instruction, and quoting it as one would be dishonest.
The diagnostic costs nothing and closes two working days after the disks are booked in. The figure that follows is fixed, in writing, before anyone is authorised to start. Logical recoveries carry no fix, no fee. Sitting outside that guarantee are electronic failures, mechanical failures, chip-level work, DVR jobs and forensic jobs, and where a member has failed physically half the quoted figure is due before the invasive stage begins.
Most of the arrays that come here from Leicestershire are unglamorous and business-critical in equal measure: an accounts server in an office off the ring road, a design studio's file store in the Cultural Quarter, a warehouse system at one of the Lutterworth distribution estates, a research group's data at one of the universities. The distance makes no difference to the price or to the queue.
A degraded array still holds everything. Powering it down costs you an afternoon of access; a rebuild onto a tired disk can cost you the volume.
Multi-disk jobs travel as a set of bare drives in one box, and they travel by post far more often than by car. Tracked and insured is the sensible choice for a parcel holding four disks of company data, and something handed over in Leicester before the last collection is normally logged in at Cambridge the next working day.
The general rule is the drive travels and the machine stays behind — out of the laptop, out of the tower, out of the iMac, out of the recorder under the counter. This bench does not dismantle equipment, and a repair shop will do it while you wait. Three things are the other way round, and getting them wrong costs you the recovery: an external drive stays sealed in its own case, a NAS comes as a complete unit, and a WD My Passport or My Book travels whole with its cable, because on those the encryption key is held on the bridge board rather than on the disk — separate the two and the data becomes unreadable even to us. A Fusion Mac needs both of its drives, each labelled. The one thing nobody can work round is flash soldered onto the mainboard, as on Apple Silicon machines: if it will not come off, there is nothing to post.
↓ Print the shipping & booking-in form (PDF)
Address it to Cambridge Data Recovery. It is about seventy miles from Leicester if you fancy driving it — M1 south to Junction 19, then the A14 east — and the lab is two minutes off Junction 32 with parking at the door. Posting costs you a stamp and a day instead. Whichever you choose, you hear from us the moment it is booked in, and the free diagnostic closes two working days after that.
Not certain what belongs in the box? Ring 0800 689 0668 before you tape it up, or let the free online diagnostic ask the questions for you.