Reject key-value batches that cannot commit atomically
A multi-key update survives restart with only half its changes. Add one bounded batch commit marker protocol.
- Focused work estimate
- 4h + prerequisites
- Priority in the scenario
- High
- Engineering practice
- Atomicity · Storage engines
Estimated field mix
- Database engineering60%
- Storage systems40%
Field percentages are editorial estimates of the ticket's engineering focus. They total 100%; they are not measured time, proficiency scores, or ownership evidence.
Review it, then add it to your workspace.
The board opens an editable draft; nothing is saved until you confirm it. Sign-in and workspace permissions apply, and Demo boards remain ephemeral.
Project context
Fictional desktop cache Pebble needs a small embedded key-value engine. Build against generated keys and bytes using an ordinary filesystem adapter; no production database replacement is implied.
Setup prerequisites
- Create a disposable local fixture directory.
- Implement fault injection for partial writes, sync failure and process restart.
Preceding work
Complete these dependencies, or supply their agreed outputs before taking this ticket.
- CKV-101 · Encode local key-value records with explicit length limits
- CKV-102 · Distinguish missing local keys from empty values
- CKV-103 · Build a local key-value index from validated log records
- CKV-104 · Declare local write acknowledgement after the durability barrier
- CKV-105 · Recover a torn key-value record at the end of the log
- CKV-106 · Represent key deletion as an ordered tombstone
Acceptance criteria
- Complete batch applies all records
- Uncommitted batch applies none
- Batch size is capped
Implementation constraints
- Do not claim filesystem transaction support that the adapter lacks.
Verification to include
- Restart after committed batch
- Crash before commit marker
Deliverables
- Batch protocol and fault schedule
Rollout and recovery
Disable batch writes while preserving single-key operations.
Value of the work
For the engineer: Practice storage formats, durability boundaries and corruption handling.
For the team: Inspect whether an engineer can state and test persistence guarantees.
Evidence boundaries
Outcome Evidence: Tests, patches, and runbooks are requested deliverables. They become Outcome Evidence only through a qualified Mission and immutable Evidence IDs.
Ownership Evidence: Independent adaptation must be observed under a declared verification policy and cite immutable Evidence IDs. Completing a planning ticket establishes no Ownership Evidence.