In the daily lives, people always buy complements together instead of substitutes. These item relationships in the user-item interaction sequences can have impact on the target item in the recommender systems. It is essential to recognize that strength of these item-item influence varies over time. However, most previous related studies don't explicitly model both the relations between the items and time factor together. In this work, we propose a Relation and Dynamic aware Graph Convolutional Network (GCN) for sequential Recommendation (ReDRec), which explicitly model item relations and time information from data. Our model uses a GCN to model item relations and utilizes different time kernel functions for each item relation to better simulate the time decay of various relationship. Our model acquires better results in extensive experiments on public datasets, which proves that our methods have a competitive performance against previous baseline models.