A two-layer reconfiguration scheme is elaborated to allow fault-tolerant control in networked systems. The presented two-layer structure separates the task of stabilisation (Layer 1) from the task of reference tracking (Layer 2). Stabilisation is achieved by using a modified virtual actuator or virtual sensor to handle actuator or sensor failures, respectively (Layer 1). It is shown that the states of the modified virtual actuator or virtual sensor can be used to guarantee reference tracking and disturbance rejection (Layer 2). The resulting fault-tolerant control scheme is applicable without a coordinating unit and without global model knowledge. Therefore, the subsystems use a communication network to exchange information in order to make fault tolerance possible. The overall reconfiguration scheme is translated to algorithms that can be implemented in each subsystem. Finally, the results are demonstrated in simulation studies.