Resilient Microservices Without the Chaos

Explore resilient microservices architecture and fault tolerance testing in this conference talk from Strange Loop 2022. Dive into Service-Level Fault Injection Testing and learn about Filibuster, a tool designed to verify microservice application behavior under failure conditions. Discover how to leverage existing functional tests to identify bugs before deployment, combining static analysis, test synthesis, and principled fault injection. Gain insights into the challenges of transferring academic code to practical implementations, based on a large-scale microservice deployment powering a popular app. Examine real-world incidents, including bad deployments and third-party dependency failures, to understand the importance of resilience testing. Learn about automatic instrumentation, JUnit integration, and standalone Filibuster implementation. Enhance your understanding of microservice fault tolerance and improve the reliability of your distributed systems.

Intro
Microservices Architecture
Microservice Architecture: Netflix
Microservices: Socio-Technical Problem
Challenge #1: Code Quality
Challenge #2: Distributed Systems
Microservice Fault Tolerance
Fault Tolerance: Example
What should happen?
What actually happens?
How Can We Do Better?
Motivating Example
Incident #1: Bad Deployment
Incident #2: 3rd Party Dependency Failure
Verifying The (Missing) Test Oracle
Service-level Fault Injection Testing
Test Adaptation: Failure-Driven Resilience
Fault Permutations: What Do We Test?
Dynamic Reduction Technique
Evaluation
Integrating the Approach
Automatic Instrumentation
JUnit Integration: Example
JUnit Integration: Faults
JUnit Integration: Conditional Assertions
Standalone Filibuster
Conclusion