Prototyping Python Microservices in Production
summary
Prototypes use Thrift RPC + IDL for cross-language endpoints, keep the IDL with the service and generate runtimes, and measure everything with Prometheus metrics (latency, error counts, payload sizes). Avoid sharing DB tables across services; store the version of generated data so rollbacks and repairs are tractable. Use client-side discovery/load‑balancing and shadowing (duplicate requests to test instances) to roll out changes slowly; expect operational complexity from distributed state and plan for it.
AI-written summary · about
chapters (9)
- 0:00Start
- 7:06Why microservices? . Composability and comprehensibility • Reliability and mobility • Concurrent development
- 8:59Why NOT microservices?
- 13:06Textbook microservice scenarios
- 14:06Real life microservice insertion
- 18:02Apache Thrift • Cross platform/language • Simple RPC interface
- 19:25Thrift: IDL Concepts
- 23:16Deployment: rolling changes
- 23:43Deployment: shadowing
related talks
30mHow To Build a Python Microservice Without Losing a Job
33mLessons learned building Python microservices
44mPython in the MicroServices world
34mPython Micro-service Architecture
30mOn scaling Machine Learning microservices
39mObservable microservices
01h 01mMicroservices Gone Wrong
21mPractical Deployment Strategies for Production Python Apps