Both disks aged together
Members are usually the same model, bought together, doing identical work since day one. They wear at the same rate, so the second failure often follows the first by weeks.
Common
It survives a disk dying. It does not survive a deletion, a corruption, a ransomware run or a rebuild started on a member that was also failing — because whatever happens above the array happens identically to both copies.
Where one member cannot read an address, the other frequently can. The work is merging both images into one complete picture — not rebuilding onto a replacement while the survivor is under load.
The protection is real, and narrower than people assume.
In a RAID 1 every write is sent to both members, so each disk holds a complete, independently readable copy of the volume. Lose one and the other carries on without interruption. That is genuinely valuable and it is the entire scope of the protection.
What it does not protect against is anything happening above the array. Deletion is a file system operation, so it is reflected on both members instantly. So is a corrupt file, a ransomware encryption run, and an accidental format. The mirror faithfully duplicates the damage, because duplicating is what it does.
Which is the distinction worth keeping: redundancy protects against hardware. Only backups protect against people and software.
Drawn from the mirrors that reach us.
Members are usually the same model, bought together, doing identical work since day one. They wear at the same rate, so the second failure often follows the first by weeks.
The classic. One member fails, a replacement goes in, and the rebuild reads every sector of the survivor — which was failing too. Watch the estimate: hours is normal, days means the source cannot be read.
A write reaching one member and not the other leaves the copies disagreeing. Without scrubbing nobody notices until the array is assembled from the wrong one.
Reflected on both instantly. No RAID level helps here, which is why snapshots and off-device backups exist.
The technical reason a mirror recovers well even when both members are damaged.
In a failing mirror the two disks rarely fail in the same places. A sector unreadable on member one is frequently perfectly readable on member two, because the bad regions developed independently.
So both are imaged — including one the controller has already marked failed, and including one needing donor heads first — and the images merged, each filling the other’s gaps. Where they disagree, the more internally consistent version is chosen by testing against the file system’s own structures rather than picking arbitrarily.
That is routinely the difference between a partial and a near-complete recovery, and it is why sending only the ‘good’ disk is a false economy.
In order.
A degraded mirror is still serving files. That window is the cheapest recovery available and it closes without warning.
A two-disk mirror rebuild is a straight copy. Hours is normal. Days or weeks means the surviving disk cannot be read at speed — so it is failing too, and the rebuild will run it at full stress throughout.
Two copies means two chances at every sector, and it regularly takes a result from partial to complete.
Against one disk failing, yes. Against deletion, corruption, ransomware or an accidental format, no — those happen above the array and are written to both members identically. Redundancy protects against hardware; only backups protect against everything else.
No. A two-disk mirror rebuild is essentially a copy, and a few terabytes on healthy hardware takes hours. An estimate in days means the unit cannot read the source at speed, which means the surviving disk is failing too.
Because they are usually the same model, bought at the same time, and have done identical work since installation. They wear at the same rate, so the second failure frequently follows the first within weeks.
Yes. The two rarely fail in the same places, so a sector unreadable on one is often readable on the other. Merging both images regularly takes a result from partial to near-complete.
That happens when a write reached one member and not the other. The more internally consistent version is chosen by testing against the file system’s own structures rather than picking one arbitrarily.
From £500 plus VAT after a free 48-hour diagnostic, with a 50% deposit where a member needs physical drive-level work.