Traffic has become a universal metropolitan problem. This paper aims at easing the traffic jam situation through routing multiple cars cooperatively. We propose a novel distributed multi-vehicle routing algorithm with an objective of minimizing the road network breakdown probability. The algorithm is distributed, and hence highly scalable, making it applicable for large scale metropolitan road networks. Our algorithm always guarantees a much faster convergence rate than traditional distributed optimization techniques such as dual decomposition. Additionally, the algorithm always guarantees a feasible solution during the optimization process. This feature allows for real time decision making when applied to scenarios with time limits. We show the effectiveness of the algorithm by applying it to an arbitrarily large road network in simulation environment.