In this paper we address the problem of uniprocessor probabilistic scheduling of real-time systems with variable execution times. For these systems the tasks have an associated probability of missing the deadline, i.e., some jobs may miss their deadlines without affecting the schedulability of the system. Therefore dropping these jobs does not affect the schedulability of the system and it could increase the probability of other jobs to meet their deadline. The problem of deciding what jobs to drop is not trivial and we discuss a possible solution.