Configuration Error Localization Based on Delta Debugging

User misconfiguration has become one of the leading causes of complex software failures in the cloud and data center. Once a vulnerability caused by misconfiguration occurs, effective software error localization technology is needed to help developers quickly find relevant code fragments to complete the repair. At present, software error localization technology mainly faces two challenges when it is applied to complex software in clouds and data centers. On the one hand, there are many configuration parameters input into the software, which makes it difficult to quickly locate the error configuration parameters; on the other hand, the complex composition of software function modules makes it difficult to locate the relevant error codes. To solve the above challenges, this paper proposes a configuration error localization technique based on delta debugging, which is based on the running results of unit tests. It includes two parts: error-configuration location technique based on dichotomous reset algorithm and error-code location technique based on execution comparison strategy. We tested our techniques on a complex cloud system software. The experimental results show that when the target configuration set size was 6 and the initial configuration set size was 20, the dichotomous reset-based algorithm improved the selection efficiency by 5.8x compared to the permutation-based algorithm with an accuracy rate of 98.6%. At the same time, the execution comparison-based strategy can be verified on the HADOOP-18362 in the real world, and the error code that leads to its configuration vulnerability can be located.