MicroRacer: Detecting Concurrency Bugs for Cloud Service Systems

Modern cloud applications delivering global services are often built on distributed systems with a microservice architecture. In such systems, end-to-end user requests traverse multiple different services and machines, exhibiting intricate interactions. Consequently, cloud service systems are vulnerable to concurrency bugs, which pose significant challenges to their reliability. Existing methods for concurrency bug detection often fall short due to their intrusive nature and inability to handle the architectural complexities of microservices. To address these limitations, we propose MicroRacer, a non-intrusive and automated framework for detecting concurrency bugs in such environments. By dynamically instrumenting widely-used libraries at runtime, MicroRacer collects detailed trace data without modifying the application code. Such data are utilized to analyze the happened-before relationship and resource access patterns of common operations within service systems. Based on this information, MicroRacer identifies suspicious concurrent operations and employs a three-stage validation process to test and confirm concurrency bugs. Experiments on open-source microservice benchmarks with replicated industrial bugs demonstrate MicroRacer's effectiveness and efficiency in accurately detecting and pinpointing concurrency issues.