The support of reliability as adopted in conventional replication or transaction processing techniques is not enough due to their distinct objectives: Replication guarantees the liveness of computational operations by using forward error recovery, while transaction processing guarantees the safety of application data by using backward error recovery. Combining the two mechanisms for stronger reliability is a challenging task. Current solutions, however, are typically on the assumption of simple transaction pattern where only a server transaction exists at the middle-tier application server, and seldom think about some complex patterns, such as client transaction or nested transaction. To address this problem, four typical transaction patterns in J2EE application are recognized first. Then a Web application server replication scheme based state synchronization point concept, RSCTP (replication scheme for complex transaction pattern), is presented to uniformly provide exactly-once semantic reliability support for these complex transaction patterns. In this scheme, EJB components are replicated to endow business logics with high availability. In addition, by replicating transaction coordinator, the blocking problem of 2PC protocol during distributed transactions processing is eliminated. Different transaction scenarios are also discussed to illustrate the effectivity of this scheme. This scheme has been implemented, and it has been integrated into J2EE compatible application server, OnceAS, and the performance evaluation shows that its overhead is acceptable.