With the rapid development of online music streaming service, it allows users to easily access and listen to music via the Internet. Music recommendation can assist the users easily finding their favorite music. For improving the accuracy of music recommendation, a music recommender system integrating the session information of clicking interaction and repeat consumption behavior of the platform users can provide a favorite recommendation closing to the music preference of users for online music service. Therefore, a session-based music recommendation with the feature of repeat consumption behavior is proposed in this paper. It also adopts parallel recurrent neural network (p-RNN) model to develop the recommender system and evaluate the effectiveness of recommendation in order to provide the better system experience of using online music service for users.