Operation 'request–response': generating clients for interservice interaction
summary
Magnit engineering teams describe their approach to automatically generating client code for inter-service communication across HTTP, gRPC and Kafka. They explain the initial situation, why the existing solution no longer worked for dozens of microservices, and how they moved to dynamically generating communication pipelines.
AI-written summary · about
chapters (17)
- 0:00Introduction
- 2:11Magnit team and services
- 4:59Microservice integration problems
- 6:03Inter-service interaction methods
- 8:13Service spec requirements
- 10:00Custom client generator
- 12:46Client issues and fixes
- 15:13Centralized spec storage
- 18:28Centralized approach benefits
- 20:39Centralized approach problems
- 22:25Alternative client generation
- 23:55Summary and recommendations
- 25:20Q&A
- 30:47Motivation and alternatives
- 35:37Client versioning and updates
- 38:07Generated code storage and testing
- 40:57Conclusion
related talks
18mService Interoperability With gRPC
55mForget about HTTP
59mEffective Microservice Communication and Conversation Patterns
18mScaling AI Agents With gRPC: Intuit's Enterprise MCP Service Marketplace
44mFrom Synchronous to Asynchronous Distributed Microservices
21mReal World Migration from HTTP to gRPC in Ruby
52mMovin beyond request-reply: how smart APIs are different
29mThe Story of Why We Migrate to gRPC and How We Go About It