The array won't show up as being mounted, you see "RAW" when attempting to read the volume, or the array's controller indicates that the disk set has failed, but the disks themselves are spinning up normally. That gap describes the entire problem with a corrupted RAID set: the data itself is usually still present, spread intact over each functioning drive, and what has been damaged was the logic used to tie those drives together. This is exactly why the first impulse (rebuild, re-initialise, let the controller repair it) is going to convert a recoverable RAID set to one that cannot be recovered.
A "corrupt" array is a logical problem on a set of otherwise intact hard drives. When you write to those same intact hard drives, you will overwrite whatever was intended for saving.
RAID (Redundant Array of Independent Disks) allows multiple physical hard drives to appear as a single virtual drive by using metadata. The metadata used in RAID includes which drive contains which portion of each stripe, how the drives were ordered and what block size is used. On parity-based configurations such as RAID 5 and RAID 6, the metadata also includes a running checksum. This checksum can be used to recalculate the contents of a missing disk. 'Corruption' occurs when this metadata becomes nonsensical while all the underlying disks remain functional. Corruption may take the form of parity values that do not match the stored data, loss or overwriting of the array metadata, or damage to a file system created on the resulting RAID volume. It is a fundamentally different problem from a dead RAID disk and is treated differently: the member disks are the good news here, not the bad.
There are just a few things that cause most of them to fail. If a rebuild goes wrong after a disk swap, you will have bad parity written across your entire set. The controller (or RAID card) has failed, or its firmware has corrupted the RAID configuration. Your power supply was interrupted during a write and your parity and data went out of sync with each other: this is known as the classic 'write hole'. A disk is removed from the array and marked as failed, then another goes down before someone realises something is wrong. Or you replace a RAID controller and the new RAID controller can't see the old array's metadata. In a NAS (network attached storage) or a server environment, these same failures show up under different names, but they still break the RAID array in the same manner.
This is where "recoverable" turns into "lost". Do not reset (reinitialise) or recreate the array in an attempt to 'correct' the problem: this will overwrite the metadata you need with new information about the array. Don't allow the controller to automatically rebuild onto the original disks. A rebuild generates and writes out new parity information, and when the rebuild begins from the wrong assumptions, it will write over actual data as it progresses. Do not run chkdsk or fsck against a corrupted volume. The tools used for repairing damaged volumes can destroy a misassembled file system. And do not reorder, relabel or swap the disks to see what mounts. Each action will write new output to each member of the set, and there is no other copy.
The bench will do the opposite of the controller. The member disks are each first imaged read-only (a full sector by sector copy) before anything else can damage the original disks. If one of the member disks is actually damaged or "weak", it will be treated gently during this process. After all the member disks have been copied, they are then used to create a virtual version of the entire RAID array. In this step, we determine the correct ordering of the disks, the proper block size, how the parity will rotate across each disk and where on each disk the stripe should begin. Once these items have been determined properly, the stripes of each disk will align correctly to create a valid file system. It is only at this point that your actual data is extracted from the RAID array (using the copy versions), without having to risk damaging the original disks in any way, and the originals stay untouched in case a second pass is needed. A member disk that has partly failed is no longer a blocker: it is just another image to work from.
Array reading as RAW, a rebuild that stalled, or a controller telling you it's all gone? Before you attempt to initialise anything again, stop: the diagnostic will cost you nothing, your quote is fixed and in writing, and an honest answer about the issue from our technicians on 0131 202 0491 will also cost you nothing.