The system consists of collection of services, running on several of nodes. The workload generated by services may cause nodes overload, which negatively impacts the response time of services. The reconfiguration of the system incrementally transforms one system state to the other. The goal of the reconfiguration is to transform the system state into one, which all services meet their response time. Finding the appropriate reconfiguration that satisfies all timing requirements is the dynamic, preemptive scheduling problem. We consider this problem as the planning problem, where plan consists of consequent steps of the reconfigurations actions. Classical planning algorithms cannot be applied to this problem because of the incomplete information. We used the concept of the conformant planning (Smith and Weld 98), which refers to planning with incomplete information that achieves the goal from any initial state compatible with the available information. The planner consists of the plan generator and the plan selector. The generator creates a collection of plans. The selector selects that one, which in the most probable degree transforms system to the state that satisfies all timing requirements. The selected plan is applied to the system. The simulation results proved the concept of our approach.