RAID has made just one promise. The array will survive when a disk fails. Most of Glasgow's server rooms believe that this means more than it does. Parity gives you enough time to replace a disk. Parity is not a second copy of data. Once RAID's promise runs out in the middle of the rebuild process, what you have left are maths problems spread across several damaged drives. Optimism is the worst possible tool for solving a maths problem.
Nearly all of the RAID disasters that we have deconstructed began as recoverable events: one disk was lost; the array had been reduced to a degraded state; the data was still alive. It was determination that turned them into disasters: a rebuild that was halted and started again; a second disk that failed under the load the rebuild placed on it; swapping out controllers "just to see"; accepting an initialisation at 2am. With each attempt, information is written onto the very same disks from which the solution will need to be calculated.
The simple rule of thumb for RAID recovery is short enough to remember under pressure: one failure, one attempt, then stop touching it. An array that stops being touched almost always keeps its data, and an array that keeps being rescued rarely does.
The process of recovering a real RAID does not rely upon a RAID controller or work on the member disks. Instead every single disk in the RAID, including "the failed" (often only tired) disks, is completely imaged. The weak points of the disks will be read last with extreme care. Then the RAID will be rebuilt using these images to recreate the original virtual configuration. This includes determining the sequence of the disks, size of each block, rotation, and parity based solely on the pattern inherent within the data's own characteristics rather than relying upon the RAID configuration that may have been destroyed. Any stale member disks are identified through their timestamp and removed prior to their potentially poisoning the results.
Nothing gets backed up until the virtual disk is mounted and passes a file system check. That’s why we don’t gamble with the original disks: they represent the evidence of what went wrong.
Striped RAID 0 without any backup; mirrors that quietly ran on one leg for months; then RAID 5 when the second drive failed; RAID 6 when the third failed; nested 10s; every piece of hardware and software that manages a RAID set (Dell PERC, HP Smart Array, mdadm, Windows Storage Spaces); they also can fail. A controller that cannot find an array does not say anything about the drives behind it. Remove each disk from your server or enclosure; bring them in and/or post them to us; do so in any order, regardless if labelled or not, as the imaging process will not be impacted by this and the maths works out the order for itself.
The price for multi-disk work starts at £500 + VAT and is quoted to the job, as each individual member disk will need to be imaged prior to touching the array. This means that the more disks you have, the longer and therefore more careful the time we spend on the job. All work begins with a free diagnostic, and all work will conclude as quoted; this means that you will receive written confirmation of the price quote before a single sector of your data is accessed. On most jobs, there is nothing to pay unless your data returns. We do not charge by the hour, nor do we assess an evaluation fee or take a percentage of what the files are worth.
If you want a “layman's” explanation of the different levels (and how each level fails), then the guide to RAID 0, 1, 5, 6 and 10 is the honest version and has been written from experience working on the actual arrays that come in to us.
No. That is the textbook worst-case scenario; however, as "failed" disks are generally weak, you should be able to recover an image of at least one of those disks. If you recover at least one good image using one of the weak members, you will likely have the required number of images to recreate the data via the parity maths. The loss is typically realised once you have attempted to rebuild the array on top of the weak disk and therefore lost access to it.
When the rebuild process writes parity across the array, there may be a point where it cannot read some piece of data and the rebuild will stop (mid-thought), resulting in members that do not agree on their last known values. Do not restart the rebuild process. Each drive is imaged, and we use the images to recreate the state before the rebuild started. A second rebuild will destroy the remainder of the readable data on those drives.
We do not care about this. Because each of these configurations (order, stripe size, parity rotation and offset) leaves evidence in the data itself (in addition to their effects on how the data is written), we simply solve for each of these from our images as part of the process. Labels are good for morale rather than maths; we will take whatever order the drives come out of the box in.
Yes and usually without getting a replacement controller card to use. The data does not reside on the controller. We will image the individual pieces (members) that make up the array and then rebuild the RAID in software. This method has the advantage of avoiding the risk associated with finding a matching controller card and being able to trust it would pick up where the old card left off and not attempt to initialise the array.
Before you post or drop off your array please get a quote from us by phone or using our form. Disks must be removed from servers or enclosures first (label each one with its bay); send the disks only. Then there are two honest routes, no vans involved. Hand the disks in at Tay House, 300 Bath Street (right at Charing Cross; just two minutes off the M8), Monday to Friday, 9am–5:30pm. Appointment not required. Or wrap it well and send it via insured tracked post from Paisley, Ayrshire, the Highlands or anywhere else in the United Kingdom. Either way the work is done in-house by our own engineers with documentation as part of the chain of custody and your data never leaves the UK.
Arrays arrive from all corners of the country for processing through our data recovery UK service. We take the full disk set by insured courier chosen by you after you receive a quote, on exactly the same terms as walk-ins: free assessment; written quoted price; most jobs without recovery have no fee.
Remove the disks, resist a second attempt, then ask for a quote. Get the whole set of disks on their way to Bath Street either personally or via an insured courier of your choice. We take sets of disks in once they are sent to us (we never collect them), and we recover from copies made of the original disks. RAID data recovery for Glasgow, where parity is considered evidence, rather than a guarantee.