RAID 10 has a reputation as the safe choice: it is fast, it mirrors everything, and it can keep running after a disk fails. That reputation is why a failed RAID 10 array often comes as a shock. If your server or NAS is reporting a failed or offline array, the decisions you make in the next hour matter more than anything else. A forced rebuild or a re-created array can turn a recoverable problem into a much harder one, so please read this before you touch the disks.
How RAID 10 actually protects your data
RAID 10 pairs disks into mirrors, then stripes the data across those pairs. In a four-disk array you have two mirrored pairs. Each pair holds its own share of every large file, so the array needs at least one working disk from every pair to function.
That leads to the point many people miss. RAID 10 can survive several disk failures, but only if they land in different pairs. Lose both disks in the same pair and that slice of the stripe is gone from the live array, even if every other disk is healthy. In a four-disk array, once one disk has failed, there is a one in three chance that the next disk to fail is its partner.
Why RAID 10 arrays fail
- Both disks in a pair fail. Disks bought together share the same age, workload and often the same manufacturing batch, so they can wear out close together.
- The surviving partner fails during the rebuild. A rebuild reads the whole of the remaining disk, and that is often the moment a weak area or tired heads finally give out.
- The wrong disk is removed. Pulling the healthy half of a degraded pair by mistake takes the array straight offline.
- A failure went unnoticed. One disk can drop out months before anyone sees the alert, leaving that pair with no protection and a stale copy of the data.
- The controller fails or loses its configuration. After a power cut or a controller swap, the disks may be fine while the array definition is missing or damaged.
- A rebuild runs in the wrong direction. If an out-of-date disk is treated as the good copy, newer data on its partner can be overwritten.
What not to do
- Do not force a failed disk back online to get the array mounted. If that disk dropped out earlier, its contents are out of date and the file system will be inconsistent.
- Do not start or restart a rebuild until you know why the array failed and which disks hold the newest data.
- Do not delete and re-create the array, or initialise it. Even when this is offered as a way to bring the volume back, it can overwrite the metadata and data you need.
- Do not move the disks around or change their slots while experimenting.
- Do not run CHKDSK, fsck or any repair tool on the volume. These tools write changes based on what they can see, and on a broken array what they see is wrong.
- Do not keep rebooting. Each restart puts more strain on disks that may already be failing.
What to do instead
- Stop writing to the array. If it is still running in a degraded state and the data is not backed up, copy the most important files elsewhere first, then shut it down cleanly. If it is offline, power it down and leave it down.
- Record what you can see. Photograph the controller or NAS screens, the status lights and any error messages. Save or export the event log if you can reach it.
- Label every disk. Mark each one with its slot number before it leaves the chassis, and include any disk that was removed earlier. An old disk can still be useful.
- Write down the timeline. Which disk failed first, whether a replacement went in, whether a rebuild started, and what happened just before the failure.
- Check your backups. Confirm when the last good backup ran and whether it covers everything before you rely on it.
How RAID 10 data is recovered
The safe approach is to work from copies, never from the original disks. Each disk is imaged first, and any disk with a physical fault is dealt with before imaging, which may mean work in a clean room. The array is then rebuilt virtually from those images, so nothing is written back to your disks.
That virtual rebuild means working out the disk order, the stripe size, the data offset and which disks form each pair. For every pair, the engineer also has to establish which copy is the most recent. This is where RAID 10 can work in your favour. Because each mirror holds two copies, a damaged area on one disk may still be readable on its partner. If that partner dropped out long ago it will be out of date, though, so the timeline you wrote down is genuinely useful.
Whether a particular array can be recovered depends on the condition of the disks and on what has been done to them since the failure, so nobody can give an honest answer until they have been examined. You can read more about how we handle these cases on our server data recovery page and in our guide to what can go wrong with a RAID system.
RAID is not a backup
Once the data is safe, treat the failure as a prompt to review your setup. RAID 10 keeps a system running through a disk failure, but it copies deletions, corruption and ransomware to both halves of every mirror straight away. Keep a separate backup, make sure failure alerts reach someone who will act on them, and replace failed disks promptly so that no pair is left running alone.
RAID 10 array down? Talk to us
Leave the array powered off and call Data Doctor on 020 7394 2529 for a free initial consultation, or bring the disks, labelled in slot order, to us at Unit 1 The Willows, 80 Willow Walk, London SE1 5SY.