Zero Support in Production
summary
Claims a production fleet running Erlang+OTP and Elm requires virtually no out-of-hours support, handling half a million live events a year by treating client inputs as an immutable source of truth, creating autonomous jobs supervised to restart on crash, and rebuilding state from scratch. Tooling includes Docker local emulation of entire clusters, InfluxDB metrics, automated end-to-end chaos testing, and UI fix-yourself buttons that kill local state so controllers reallocate resources without human intervention.
AI-written summary · about
chapters (11)
- 0:00Introduction and background
- 3:45Building self-supporting software
- 10:31Absolute truth in system design
- 16:57One-way data flow patterns
- 20:04Erlang supervision and resilience
- 26:02Local emulation of production
- 29:52Testing and chaos engineering
- 33:07Monitoring and observability
- 38:03Fixing failures without support
- 41:17Why not use existing tools
- 42:42Q&A session
related talks
36mOTP Has Done It
49mInto Production
16mLiving without pre-production environments
56mGetting all the 99.99(9) you always wanted
39mInstrumenting mayhem: functional chaos engineering
45mZero production incidents. How to use an unpopular solution on a project
58mCultivating Production Excellence
57mChaos Engineering Primer