MD mismatch count
Linux MD mismatch_cnt counts sectors rewritten, or that would be rewritten during check, when checking array redundancy.
Granularity: The accounting unit can exceed the actual damaged region. The mdadm 4.2 manual illustrates a failed 64 KiB comparison contributing 128 sectors, not proof that all 128 sectors differ. [2]
Interpretation: A mirror mismatch can involve unused or swap space rather than corrupted live file data. A mismatch alone does not identify the correct copy. Repair may make copies agree without independently establishing application correctness. [2]
Coverage: Retain the operation, checked range, progress and completion alongside the counter. A zero value alone does not establish a completed full-array scan; check is not guaranteed read-only for every RAID level. [1]
Relations
- related
- Data scrubbing, RAID 1, RAID 5, RAID 6, Parity inconsistency, Parity pollution
Citations
- Linux 6.18 documentation: RAID arrays, sync_action, mismatch_cnt and sync limits
- mdadm 4.2 md(4): scrubbing and mismatches
Sources & caveats
- Documented Linux MD behavior; not a hardware measurement or a conversion from counter values to lost bytes.