hesela.devStable storage
Stable storage preserves data through a specified set of failures and supports recovery without losing the data it has committed.
For NFSv4.1, RFC 8881 section 1.7 covers repeated power interruptions, including closely spaced failures, operating-system failures and failures of components other than the storage medium. Failure of that medium is excluded. [1]
The implementation can use protected intermediate storage with recovery software; stable does not require every byte acknowledged as stable to be on its final disk surface. [1]
SNIA separately identifies single non-media hardware failures, repeated software crashes and a required period without external power. State the applicable contract and retention interval; the dictionary does not assign one universal number of hours. [2]
A persistence domain identifies protected components; stable storage specifies the recovery guarantee. Neither label alone certifies a particular appliance.
The 1979 Lampson-Sturgis construction uses two copies outside a shared modeled decay set, ordered verified writes and recovery cleanup. Undetected corruption and overly frequent decay are excluded from its assumptions; two copies alone do not establish its guarantee. [4]
Relations
- related
- NFS COMMIT, Persistence domain, Power loss protection, Non-volatile random-access memory, Data durability
Citations
- Noveck and Lever (editors). RFC 8881, NFSv4.1, section 1.7: Stable Storage.
- SNIA Dictionary: stable storage; failure classes and powerless-retention scope.
- Butler W. Lampson and Howard E. Sturgis. Crash Recovery in a Distributed Data Storage System. working paper, 1 June 1979, sections 4.1 and 5.1 (pp. 6-7, 10-12).
Further reading
Sources & caveats
- Primary-source definition review. Appliance compliance and recovery behavior are not measured; no universal retention duration is inferred.