Building Reliable Infrastructure in Rust
summary
After years chasing production breakages the talk shows concrete techniques: property and model‑based testing with prop test and quickcheck that auto‑shrink failing traces, Sled embedded database examples that exercised crash and restart, and fault injection using the fail crate, Jepsen, and TLA+ plus simulator work. It cites empirical findings (92% of critical failures leave code traces; 58% would be caught by simple tests), explains lineage‑driven fault injection, and demonstrates deterministic simulators, clocks, RNGs, and replayable failing inputs that actually find bugs.
AI-written summary · about
chapters (11)
- 0:00Intro
- 3:30Your Brain
- 6:52PropTest
- 10:38ModelBased Testing
- 14:18Linearize Ability
- 17:21Fault Injection
- 22:28Jepsen
- 24:09ToA
- 25:33Network Simulation
- 28:41Lineage Driven Fault Injection
- 30:42Miscellaneous
related talks
42mJepsen 9: A Fsyncing Feeling
37mHow three engineers at Antithesis shipped a production database in 14 months - safely!
34mToo Big to Fail
53mBuilding Distributed, Multi-Tenant Databases with FoundationDB and Rust
34mSearch Engine in production with Rust
26mDon't Let Your Kubernetes Cluster Go Wild: Ensuring Etcd Reliability
04mOrchestrating chaos: Applying database research in the wild
37mArchitecture Guidance