The massive deployment of information infrastructure including 5G base station and edge computing server has increased the electricity purchasing cost for network operators. Thus it is necessary to promote the interaction of energy between the edge network and power grid to achieve energy saving and cost reduction. Current studies mainly focus on the edge network participating in a day-ahead economic dispatch strategy without considering that the double randomness of renewable energy and network traffic may cause the mismatch of energy supply and demand in the network. To cope with the network load management problem in a strongly random environment, a real-time energy management strategy faced to a virtualized network is proposed. First, to minimize the sum of time-average energy cost, a joint resource allocation, energy storage and energy sharing model is proposed. Second, to solve the proposed multi-slot problem with future network information being unknown, this paper proposes an online resource allocation algorithm based on a stochastic dual-subgradient method. Also, a full distributed energy-computing resource sharing strategy is investigated considering the protection of privacy of operators. Finally, the simulations show that proposed online algorithm effectively reduces the total energy purchasing cost of a 5G edge network without a-priori knowledge.