This paper investigates the reliable sampled-data control problem for a class of singular networked cascade control systems with randomly occurring actuator failures, time-varying delay and disturbances. In particular, the randomly occurring actuator failures describes the phenomenon of the actuator failure appearing in a random way and is modelled by a Bernoulli distributed white noise sequences with a known conditional probability. The main purpose of this work is to design a reliable sampled-data controller such that an optimised upper bound of the predefined performance index is well guaranteed for the systems under consideration even in the presence of both the randomly occurring actuator failures and disturbances. The resulting controllers are reliable in the sense that they will provide guaranteed stochastically admissible and performance index when all actuators are operational as well as when some actuators experiences failure. Finally, two numerical examples are given to exhibit the effectiveness of the proposed control scheme.