One advantage of the network coding based P2P live streaming system is that multiple source peers can serve the requests efficiently without explicit cooperation control. However, the lack of cooperation also brings redundant data due to the linear dependence, even when the Galois field of the coding coefficients is large enough. In this paper, we first analyze the causes of the redundant data. Then, we propose selecting active source peers from the source set given by the tracker to handle this problem. Active source peer selection problem can be regarded as a variant of 0-1 knapsack problem, which is NP hard, such that we design an approximation algorithm to compute the solution. Through formal proofs and simulations, we verify the validity of the algorithm, which can further improve the QoS of the P2P live streaming.