High availability (HA) computing has recently gained much attention, especially in enterprise and mission critical systems. The HA is now a necessity that is no longer regarded as a luxury feature. Thus, we, conjunctively with the open source community, are in process of enhancing the HA feature to Open Source Cluster Application Resources (OSCAR), a widely adopted Linux PC cluster system. Server redundancy will be our initial key aspect of the next generation HA OSCAR cluster system. In this paper, we introduce a HA server for OSCAR cluster system. Its architecture and mechanism is discussed, and then we model and predict the dependability of the system by a Petri net-based model, Stochastic Reword Net (SRN). The reliability and instantaneous availability of the system are presented as a result.