This paper presents RANDREL, a simulator based trainer for reliability testing. Reliability testing is modeled as a random walk through the program execution state space into which an initial set of faults has been seeded. When a fault is encountered during the simulation, a failure is reported, the tester fIxes the fault and resumes reliability testing. After each failure, the reliability growth model is updated and used to determine whether the reliability goal has been achieved. RANDREL provides a low-cost environment for learning reliability testing and modeling, and for exploring factors that contribute to the effectiveness of reliability testing. This paper describes the capabilities of RANDREL and the methodology being used to validate the model.