A safe and efficient route from the start point to destination is a precondition for drone to realize its various applications. During the flight, the drone must be able to respond to various dynamic changes quickly and replan its route. For this purpose, the online route planning problem of multi-rotor drone is studied in the discrete urban environments in this paper. The model of discrete urban environments is established, based on which the flight rules of multi-rotor drone are defined. The route planning model is developed, and the constraints on path points and the index of evaluating the route planning are designed. According to the different influences of dynamic changes on multi-rotor drone's flight, the dynamic changes are classified into three groups, i.e., fixed no-fly zone, cooperative drone and non-cooperative drone. Three strategies (replanning 3D route, changing flight velocity and roll optimization of 3D route) are proposed to cope with different types of dynamic changes. The proposed strategies are also valid when addressing the combined types of dynamic changes. The rapid-exploring random tree (RRT) algorithm is used to generate the route, and some modifications are carried out to make it suitable for the discrete urban environments. The validity of the proposed three online route planning strategies are verified in simulations, and it is also effective when the multi-rotor drone is facing more than one type of dynamic change. Simulation results demonstrate that the route generated online can avoid the bad effects brought by the dynamic changes.