NAS, RAID and Backup Storage
RAID rebuild URE risk.
The chance of at least one unrecoverable read error (URE) during a rebuild is 1 - (1 - p)^bits, where p is the drive error rate and bits is everything the rebuild must read. Four 8 TB drives in RAID 5 at 1 error in 10^14 bits read 192 trillion bits, which gives about an 85% chance of at least one error on paper. Runs in your browser; nothing is sent or stored.
Your array
Read error risk
Fill the form and press Calculate.
Worked example
Four 8 TB drives in RAID 5 after one fails: the rebuild must read the 3 survivors, 3 x 8 TB x 8 x 10^12 = 192 trillion bits. At 1 error in 10^14 bits that is 1.92 expected errors and an 85.3% chance of at least one. At 1 in 10^15 the chance falls to 17.5%, and at 1 in 10^16 to 1.9%. A mirror that only reads one 8 TB drive (64 trillion bits) at 1 in 10^14 has a 47.3% chance.
How it works
Treat each bit read as an independent trial that fails with probability p. Over B bits the chance of at least one error is 1 - (1 - p)^B. B is the number of surviving drives that must be read in full x drive size x 8 bits per byte (1 TB = 8 x 10^12 bits).
This is a worst-case reading of the datasheet. The published figure is a limit the drive is specified to stay under ("less than one bit in 10^14"), so many drives do better, and scrubbing the array regularly can find and repair errors before a rebuild. It also assumes errors are independent. Treat the result as a reason to choose RAID 6 or mirrors for big drives and to keep a real backup, not as a prediction.
Estimates only. RAID is not a backup: it keeps an array running through a drive failure but does not protect against deletion, malware, fire or theft. Real arrays and drives vary, so use this for planning and check your own drive datasheets.
Sources and basis (opened 2026-10-08):
- RAID (Wikipedia): Describes unrecoverable read errors during rebuild, quotes typical rates of less than one bit in 10^15 for enterprise-class drives and less than one bit in 10^14 for desktop-class drives (the article marks these figures as disputed), says parity schemes such as RAID 5 are particularly prone to UREs, that RAID 6 provides redundancy that allows double-drive failures, that mirroring lowers the risk, and that scrubbing can reduce it.
- Hard disk drive (Wikipedia): States that 2018 specifications for consumer SATA hard drives give one uncorrected bit read error in every 10^14 bits, and for enterprise SAS drives one in every 10^16 bits.
FAQ
What is a URE in RAID?
An unrecoverable read error is a sector the drive cannot read even after its own error correction. During a RAID 5 rebuild it can stop the rebuild or corrupt the reconstructed data, because there is no spare parity left to repair it.
Does an 85% URE chance mean RAID 5 fails 85% of the time?
No. The datasheet rate is an upper limit that drives are specified to stay under, so real drives often do better, and scrubbing helps. It is still a good reason to prefer RAID 6 or mirrors with large drives.
Why is a mirror less exposed than RAID 5?
A mirror rebuild reads one partner drive, not every surviving drive, so far fewer bits are read. The 4 x 8 TB example drops from 192 trillion bits to 64 trillion bits.
Which URE rate should I enter?
Use the non-recoverable read errors per bits read line on your drive datasheet. Consumer drives commonly list 1 in 10^14, and enterprise-class drives list 1 in 10^15 or 10^16.
Is RAID a substitute for backup?
No. A read error during a rebuild is one more reason to keep separate backup copies, following the 3-2-1 rule.