By manufacturer
Dell PowerEdge HPE ProLiant Lenovo ThinkSystem Fujitsu Primergy Supermicro IBM System x / Power Acer AltosRAID 1: mirror. Recovery is often possible from a single healthy side.
RAID 1 (mirror) duplicates data across two identical disks: the failure of one alone loses no data, but more complex situations — successive failures, split-brain after a controller failure, a desynchronised mirror — call for targeted work. We work on the clones, never on the originals.
When RAID 1 runs into trouble.
- Double failure: both disks in the mirror fail, close together in time. Physical- and logical-level recovery on at least one of the two copies.
- Split-brain after a controller failure: following a controller fault, one of the two copies is arbitrarily marked as "valid" and subsequent writes diverge from the other. Recovering the correct state requires timeline analysis.
- Silent bit rot: one of the two copies has corruption on specific sectors, undetected by SMART. A byte-by-byte comparison of the two copies is required.
Differential analysis + recovery.
- Bit-by-bit cloning of both copies.
- Differential analysis to identify which copy is the most recent / most reliable / has only localised corruption.
- Filesystem recovery from the better copy, with targeted repair of the corrupted areas from the complementary copy where available.
The questions we get most often.
Is RAID 1 really "always safe"?
Safer than RAID 0, but not immune. Correlated events (a power anomaly that damages both disks, vibration, a faulty batch, a ransomware attack that encrypts whatever the system sees without distinguishing whether it comes from one copy or two) bypass the mirror's redundancy. RAID 1 must always be paired with a separate backup.
Split-brain: how do you work out which copy is the "right" one?
Forensic analysis: filesystem metadata timestamps, the filesystem journal, application logs held on the volume, and any sequence number fields in the data. In ambiguous cases we present both possibilities to the customer and leave the final call to whoever knows the application context.