RAID Recovery After a Multi-Disk Failure: What's Actually Recoverable
A RAID array is built to survive a certain number of disk failures without losing data — one for RAID 5, up to two for RAID 6, and varying amounts for RAID 10 depending on which disks fail. Problems start when the array loses more disks than its redundancy level allows, or when a rebuild is attempted on a controller or disk set that isn't actually healthy.
Why the second failure is the dangerous one
Rebuilds are disk-intensive: every remaining member disk is read in full to reconstruct the failed one's data. Disks that were already marginal — especially older drives in an array that's been running for years — frequently fail partway through a rebuild that they would otherwise have survived doing nothing. This is why a degraded array (one disk down, but the array still online) should be treated as urgent: the safest state is to stop writing to it and get it assessed before attempting a rebuild.
The same logic applies to RAID 0, which has no redundancy at all. A single disk failure in RAID 0 doesn't 'damage' the array so much as remove one-third or one-half of every file on it, striped across the remaining disks in fragments.
What tends to be recoverable, and what doesn't
In practice, recovery outcomes depend more on what happened after the failure than on which RAID level was used. Arrays where the failed disk was swapped and a rebuild was allowed to run to completion on a controller that mismatched stripe size or disk order are harder to recover than arrays that were simply left alone and powered down. Physically damaged disks (clicking, not spinning up, or with reallocated sector counts climbing) need imaging at the disk level before any array reconstruction is attempted — reconstructing directly from a failing member disk risks losing what's left of it.
Where the underlying disks are readable (even if the controller no longer recognizes the array), reconstructing the correct disk order, stripe size and parity rotation from the raw data is usually possible without the original controller's configuration. This is why an assessment before any further action — no rebuilds, no re-initializing the array, no formatting — is the single highest-leverage step after a multi-disk RAID failure.
Immediate steps if you're dealing with this now
Stop the rebuild if one is in progress and it hasn't completed. Don't reinitialize the array or run the controller's 'repair' or 'foreign configuration' options. Note the disk order and slot positions before removing any drives, and label them. If a disk is making unusual noise, keep it powered off — every additional power-on increases the risk of further damage.
Dealing with something like this now?
Contact AS for an initial assessment of your recovery situation.