This paper deals with the leader-follower consensus problem for double-integrator multi-agent systems using sampled position data information. An observer-based output feedback controller is designed to study this problem while taking into account intermittent sensor failures. The non-uniformity and randomness of the sampling times due to intermittent information lead to a µ−varying linear system on a discrete stochastic time domain for the closed-loop system dynamics (here µ is the graininess function). Some necessary and sufficient conditions for the observer and controller gains are derived, using positive perturbation and Lyapunov operators on the space of symmetric matrices, to guarantee mean-square exponential stability for the observation and tracking errors. Some simulation results illustrate the effectiveness of the proposed observer-based output feedback controller.