RAID systems — what they are and why they fail
RAID stands for Redundant Array of Independent Disks: an array of ordinary drives working together. Such systems are used by organisations with particular storage requirements — high reliability, better performance, or a combination of the two. They are built from standard hard disks on various interfaces (SATA, SAS, NVMe), configured so that they operate as one.
The most common mistake is to assume RAID is a backup. It is not. RAID protects against the failure of a single disk. It does not protect against accidental deletion, ransomware, a controller failure, fire or theft. A RAID array still needs a separate backup.
Recovering an array calls for a deep familiarity with the different RAID configurations and with how various controllers behave. Every controller manufacturer has its own stripe order, block size and parity placement — and without reconstructing those parameters correctly, the array cannot be reassembled.
The common configurations
RAID 0 — striping
Two or more disks share the load between them, so the array as a whole is faster than any single disk. The drawback is that there is no redundancy at all: one failed disk brings the whole array down. Recovery means testing each disk separately, recovering the failed ones, and only then reassembling the stripes according to the original controller's layout.
RAID 1 — mirroring
Each disk is mirrored by an identical one, and every write goes to both simultaneously. If one fails, the system carries on. Recovery is often comparatively straightforward — but there is a well-known trap: sometimes the controller fails to notice that one disk has stopped updating, or it raises an alert that nobody acts on. Data is then genuinely being written to only one disk, and when that one fails it turns out the 'second' copy is months out of date.
RAID 5 — striping with parity
A configuration combining good performance with redundancy: if any one disk stops working, the system keeps running. The minimum is three disks. Recovery is needed when more than one disk has failed, when the controller itself has failed and corrupted data in the process, or in the very common case where one disk failed unnoticed and the problem only surfaces when a second one goes.
RAID 6 — double parity
Similar to RAID 5, but it survives two simultaneous disk failures. The minimum is four disks. Recovery is called for when the controller is damaged, or when more than two disks fail together — something that happens more often than you would expect after an electrical fault, or through a manufacturing flaw in a batch of identical drives bought at the same time.
RAID 10 and combined layouts
There are further configurations such as RAID 10 and RAID 01, along with vendor-specific layouts. All of them require each disk to be examined individually, the failed members to be recovered, and the right tools to be used to reassemble the drives according to the original controller's configuration. At the end, the logical structures on the reassembled array have to be verified.
What to do when an array goes down
- Do not start an automatic rebuild before there is an image of the current state — a failed rebuild overwrites data
- Do not re-initialise the array or change the controller configuration
- Label the disk order in the bays before removing anything
- Note the controller model and configuration if known — it shortens the recovery considerably