This paper investigates an unreliable $M/G(P_1,P_2)/1$ retrial queueing system with a woking vacation. An arriving customer successfully starts the first phase service with the probability $\alpha$ or the server fails with the probability $\overline{\alpha }$. Once failure happens, the serving customer is taken to the orbit. The failed server is taken for repair with some delay. Once the repair is comleted, the server is ready to provide service once again. In this background, we implemented the working vacation scenario. During working vacation, the service will be provided at a slower rate, rather than entirely stopping the service. The supplementary variable method was adopted to find the orbit and system lengths. Additionally, some unique results and numerical evaluations have been presented.