Modelling Microservices at Spotify
summary
System‑Z models Spotify microservices by combining runtime traces from Apollo/nameless and mostly‑static YAML metadata to produce dependency graphs, ownership, PagerDuty/Slack contacts and deployment views. It harvests outgoing‑call data, compares declared vs runtime dependencies, generates architecture diagrams and drives Helios-based rollouts. The system tracks about 1000 components across roughly 8000 machines, enables safe migrations via proxying and traffic ramp‑ups, and shows ~200 weekly users and clearer discovery of infra tooling—effectively centralizing systems metadata.
AI-written summary · about
chapters (15)
- 0:00Introduction
- 2:18Scaling
- 4:52Teams
- 8:57How we work
- 10:52Naming
- 12:12Terminology
- 22:00Data Model
- 35:30Warnings
- 36:30Impact
- 38:30Survey
- 40:30Dirty metadata
- 41:30Dependencies
- 42:30Key Features
- 44:30Size
- 45:30Conclusion
related talks
49mMicroservices at Spotify
29mTune Your Microservices by Learning from Traces
24mMicroservices in Practice
27mBalancing autonomy, technical alignment, & speed through strategic collaboration
31mHundreds of Microservices without breaking your APIs
36mAdopting gRPC at Spotify
40mScaling code health with always on rewrites and AI at Spotify
15mPlatforming the Platform: Running Our IDP With Our IDP