ConfExp: Root-Cause Analysis of Service Misconfigurations in Enterprise Systems

Abstract

Misconfiguration is a known and increasingly serious problem in enterprise systems due to frequent code updates and retuning of the configuration parameters. Diagnosing complex, residual misconfiguration problems that lead to inaccessible services or failed transactions often starts with either a user complaint or observation by administrators, followed by a largely manual process of deciding what tests to run and how to proceed with further testing based on the test results. The goal of this paper is to automate this process and thereby make root-cause analysis of accessibility related misconfigurations much speedier and much more effective. We explore a domain-knowledge-driven methodology, called ConfExp using a network emulator that runs real enterprise networking protocols. Thus, by using commonly used tests, we show that the root-cause can be determined in all cases where discriminative tests exist. The methodology also highlights areas where more discriminative tests are needed to pinpoint the precise configuration variables at fault.