Delta Debugging Microservice Systems with Parallel Optimization

Abstract

Microservice systems are complicated due to their runtime environments and service communications. Debugging a failure involves the deployment and manipulation of microservice systems on a containerized environment and faces unique challenges due to the high complexity and dynamism of microservices. To address these challenges, we propose a debugging approach for microservice systems based on the delta debugging algorithm, which is to minimalize failure-inducing deltas of circumstances (e.g., deployment, environmental configurations). Our approach includes novel techniques for defining, deploying/manipulating, and executing deltas during delta debugging. In particular, to construct a (failing) circumstance space for delta debugging to minimalize, our approach defines a set of circumstance dimensions that can affect the execution of microservice systems. To automate the testing of deltas, our approach includes the design of an infrastructure layer for automating deployment and manipulation of microservice systems. To optimize the delta debugging process, our approach includes the design of parallel execution for delta testing tasks. Our evaluation shows that our approach is scalable and efficient with the provided infrastructure resources and the designed parallel execution for optimization. Our experimental study on a medium-size microservice benchmark system shows that our approach can effectively identify failure-inducing deltas that help diagnose the root causes.