CAFault - Enhance Fault Injection Technique in Practical Distributed Systems via Abundant Fault-Dependent Configurations

Learn about CAFault, a novel testing framework that enhances fault injection techniques in distributed systems by incorporating configuration-aware testing methodologies. Discover how this research addresses the limitations of traditional fault injection testing that relies on fixed default configurations, potentially missing critical vulnerabilities that emerge under different system configurations. Explore the framework's two key innovations: the Fault-Dependent Model (FDModel) that intelligently prunes the vast combinatorial search space between fault inputs and configuration inputs, and fault-handling guided fuzzing that effectively detects bugs hidden in deep execution paths. Examine the comprehensive evaluation results across four widely-used distributed systems including HDFS, ZooKeeper, MySQL-Cluster, and IPFS, where CAFault demonstrated significant improvements in fault tolerance logic coverage compared to state-of-the-art tools like CrashFuzz, Mallory, and Chronos, achieving 31.5%, 29.3%, and 81.5% more coverage respectively. Understand how this approach successfully identified 16 previously unknown serious bugs, highlighting the importance of configuration-aware fault injection in ensuring distributed system reliability and availability in real-world deployment scenarios.