This paper proposes a slice admission control scheme to maximize profit and resource utilization. It considers an online scenario where the infrastructure provider (InP) does not have full knowledge of future requests and requests can renege after their waiting time is exceeded. Moreover, it employs a dynamic priority and a capacity sharing mechanism to buy back idle resources from already accepted slices. The performance evaluation demonstrates that the proposed algorithm yields about 26%, 7%, and 7.3% higher profit, acceptance rate and resource utilization gains over the state of the art schemes, respectively.