When a drive fails in a RAID 5 or RAID 6 array, its data is not gone — it can be recalculated. Every stripe carries a parity value computed with a bitwise operation called XOR. Read the surviving drives, XOR them together, and the missing drive's contents fall out of the math automatically — the basis of a RAID rebuild.
What Parity Actually Is
XOR (exclusive OR) compares two bits and returns 1 if they differ, 0 if they match. Apply it across every drive in a stripe and you get a parity bit that summarizes all of them at once. Say three drives hold the bits 1, 0, and 1 for a given stripe. XOR them together — 1 XOR 0 XOR 1 — and the parity value is 0, stored on a fourth drive. If the drive holding that middle 0 dies, the controller does not need to guess: it XORs the survivors and the parity value together (1, 0, 1) and the missing bit — 0 — falls straight out. Real stripes are megabytes of data processed this way in bulk, not single bits, but the arithmetic is identical.
RAID 5 vs RAID 6: Single vs Double Parity
RAID 5 spreads one parity block per stripe across the array (not a single dedicated parity drive, contrary to a common misconception), which tolerates exactly one drive failure at a time. RAID 6 computes two independent parity values per stripe — the second typically using a different algorithm (Reed-Solomon) than plain XOR — which lets the array survive two simultaneous drive failures. That extra protection costs more usable capacity and more write overhead, since every write now updates two parity calculations instead of one.
What Happens the Moment a Drive Dies
The array does not lose data the instant a drive fails — it drops into degraded mode. Every read that touches a stripe on the dead drive now has to be reconstructed on the fly by XORing the rest of that stripe, which adds real load to the controller and every remaining drive. Nothing is lost yet, but the safety margin is gone: on RAID 5 there is no redundancy left at all; on RAID 6 there is exactly one failure's worth remaining.
Rebuilding, Step by Step
Once a replacement drive goes in, the controller works through the array stripe by stripe: read the corresponding block from every surviving drive, XOR them (plus the second parity calculation on RAID 6) to reconstruct what the dead drive held, and write that reconstructed block to the new drive. Repeat for every single stripe in the array. That means a rebuild requires reading 100% of the data on every surviving drive — not just the dead drive's old capacity — which is why rebuild time scales with total array capacity, not just the size of one drive.
The Danger Window: Why Drive Choice and Array Size Matter
A RAID 5 rebuild runs with zero redundancy the entire time. If a second drive fails, or if a surviving drive hits an unrecoverable read error while being scanned for reconstruction, an entire stripe — and on RAID 5, potentially the whole array — can be lost. Bigger drives mean more data to read during rebuild, which means a longer exposed window. It's also why the drives doing the rebuilding matter: a drive that cannot sustain steady write throughput drags the whole process out further. See our explainer on CMR vs SMR hard drives for why that distinction specifically matters for RAID membership.
Reducing Rebuild Risk
A few decisions materially shrink the danger window:
- Choose RAID 6 or another dual-parity scheme for larger arrays, so a second failure during rebuild does not mean total loss
- Use CMR-rated NAS drives that hold up under sustained rebuild writes rather than choking on them
- Run periodic array scrubs so latent errors on surviving drives get caught and fixed before a rebuild ever needs to read past them
A multi-bay unit like the QNAP TS-433 paired with Western Digital Red Plus or Seagate IronWolf Pro drives is a reasonable baseline for exactly this reason. None of this replaces backups — parity protects against a drive failure, not against accidental deletion, ransomware, or a fire in the same room as the array.
Frequently Asked Questions
Is RAID a backup?
No. RAID protects against a drive failure so the array keeps working. It does nothing to protect against accidental deletion, file corruption, ransomware, or the array's physical location being destroyed — a separate backup is still necessary.
Why does a rebuild put load on every drive, not just the new one?
Because reconstructing the missing data requires reading the corresponding block from every surviving drive in each stripe before the XOR math can run. The new drive receives the writes, but every other drive does just as much reading.
How is rebuilding RAID 5 different from rebuilding RAID 1?
RAID 1 mirrors data directly, so a rebuild is a straight copy from the surviving mirror — simple and comparatively fast. RAID 5 has to recompute the missing data via parity math across every other member, which involves more total drive I/O for the same amount of data.
Can two drives really fail at the same time in RAID 6?
Yes — surviving two simultaneous failures is RAID 6's entire purpose, achieved with two independent parity calculations per stripe. A third simultaneous failure, however, still exceeds what the array can recover from.
Share this article: Twitter