Rebuild Risk Sheet
How long the window is, how much gets read, and what the datasheet allows over that read.
BFCBrilliance · bfcbrilliance.com/tools/raid-rebuild-time-calculator
The numbers
- REBUILD TIME
- drive TB x 1,000,000 ÷ MB/s ÷ 3600
- RAID 5 / 6 read
- (n − 1) drives, end to end
- RAID 10 read
- ONE drive — the mirror partner
- Consumer / NAS spec
- 1 error per 10^14 bits = 12.5 TB
- Enterprise spec
- 1 error per 10^15 bits = 125 TB
- 8 TB at 100 MB/s
- about 22 hours, idle array
- Busy array
- often a third the speed — treble the window
- ⚠ The spec is a MAXIMUM
- real drives beat it — this is a worst case
- ⚠ RAID 6 protects
- the rebuild window itself
Layouts compared
| Level | Drives | TB each | Read TB | Hours | Spec risk % |
|---|---|---|---|---|---|
This array
- Machine
- Date
- RAID level
- Drives
- Drive size
- Rebuild speed
- REBUILD HOURS
- DATA READ (TB)
- Drive class
- Errors spec allows
- SPEC RISK %
- Controller behaviour on a URE
- Backup verified before starting?
When a drive fails
- BACKUP VERIFIED before starting the rebuild — this is the riskiest moment
- Replacement drive from a different batch where possible
- Rebuild prioritised over array performance if the controller allows it
- Other drives' SMART data checked — a second one may be close behind
- Rebuild window understood as the period with no redundancy left
- Controller behaviour on a read error known IN ADVANCE
- Array left as idle as practical for the duration
- Second failure plan decided before it happens, not during
The read is the risk, not the write
Rebuilding one 8 TB drive in a six-drive parity array means reading 40 TB off the survivors — five times what gets written. RAID 10 reads one drive instead, because it copies the mirror partner. That difference never appears in a usable-capacity comparison and it is the strongest practical argument for mirroring on large arrays.
Treat the probability as a ceiling
The datasheet figure is a maximum the manufacturer guarantees not to exceed, not a measured failure rate, and real drives do considerably better. Rebuilds succeed far more often than this arithmetic implies. Use the number to COMPARE levels, drive classes and array sizes — not to predict your Tuesday.