The increasing complexity of distributed enterprise systems has made the task of managing these systems difficult and time consuming. The only way to simplify the management process is to automate much of the work so that minimum human effort needs to be invested. This has lead to research in autonomic systems that are selfhealing, self-configuring, self-protecting and self-securing. We believe that the starting point of any autonomic system is to understand the dependencies between various components of the system and use it to perform higher order management tasks. As a proof of concept, we are trying to build self-healing capabilities into distributed enterprise applications by modeling the failure dependencies within the system. In a complex distributed environment, failures tend to propagate from one part of the system to the other. Hence, the failure symptoms may be observed at a point far removed from the actual cause of these failures. Therefore, localizing observed failures to its root cause is an important prerequisite to initiating micro-recovery procedures on a failed system. In this paper, we suggest a methodology to obtain a failure model of the application and use it to perform root-cause analysis based on observed system failures.