With wafer residency time constraints, to obtain a feasible schedule for cluster tools is to balance the wafer sojourn time to some extent among the processing steps. However, the wafer sojourn time fluctuation caused by activity time variation can result in the violation of residency time constraint at some step such that a feasible schedule under the assumption of deterministic activity times may become infeasible. Thus, it is a great challenge to operate a cluster tool subject to wafer residency time constraints and activity time variation. This paper addresses the real-time operation issues of single-arm cluster tools. The system is modeled by a Petri net. With this model, a real-time control policy is proposed to offset the activity variation as much as possible by dynamically adjusting the robot waiting times. In this way, an off-line schedule under deterministic activity times is made to be adaptable to certain activity time disturbance such that it becomes feasible. An illustrative example is given to show the applications of the work.