Escaping Version Skew: Formalizing Compatibility in a World of Partial Rollouts
summary
A concrete method to avoid rollout breakage: detect breaking schema changes automatically, generate strict writer types and union reader types, and gate merges with a jsoncompat subsumption checker that uses static proofs plus fuzzing to produce witness payloads. CI rejects unsafe edits, engineers stamp versions and the system measures the old-read tail so risky rollbacks are visible. An incident where mismatched cache writes caused a 15% error spike motivates the flow and the stamp/union pattern for safe gradual rollouts.
AI-written summary · about
chapters (14)
- 0:00Version skew challenges
- 1:03Redis cache mismatch incident
- 3:34Breaking changes limits
- 5:18Time in system state
- 7:01Protobufs limitations
- 8:46Agents need boundaries
- 12:12Reader/writer type safety
- 15:44Six-step schema evolution
- 19:58JSON schema checker
- 23:50Compatibility checker demo
- 27:43Incremental compatibility checks
- 30:30Where to apply tooling
- 31:13SRE playbook for schemas
- 32:00Q&A
related talks
25mConsistent Updates for Scalable Microservices
40mBackwards Incompatible Tales
01h 13mChange Is Inevitable: Versioning Event-Driven Systems
41mSurviving Continuous Deployment in Distributed Systems
51mPatterns & Playbooks to Make Every Commit Production Ready
28mMaking the Impossible Impossible: Improving Reliability by...
43mPatterns of Legacy Displacement
30mPutting the Semantics into Semantic Versioning