TN-R5-003 — Latent Sector Errors and Why Rebuilds Surface Them
0. Purpose and Scope
This note explains why unreadable sectors appear during RAID 5 rebuilds even when drives previously appeared error-free.
1. Definition of Latent Sector Errors (LSEs)
LSEs are unreadable sectors that:
- Exist prior to failure
- Are never accessed during normal operation
- Surface only under rebuild I/O patterns
2. Why Rebuilds Trigger LSE Exposure
Rebuilds force:
- Sequential reads across the entire disk surface
- Access to previously cold data regions
- Reconstruction of every stripe involving the missing member
This guarantees LSE discovery.
3. Impact on Parity Reconstruction
When an LSE is encountered:
- A stripe cannot be reconstructed deterministically
- Controllers substitute zeros, guesses, or partial data
- Parity is rewritten using incorrect inputs
This permanently alters parity trust.
4. Failure Outcome
A single unreadable sector at the wrong stripe location can:
- Stall rebuilds
- Complete rebuilds with silent corruption
- Collapse recovery paths entirely