The invention relates to a service selection method and device in a micro-service cluster, and belongs to the technical field of computer software. According to the method, at least three characteristics of recent n-time average response time and the concurrency of each service, a real-time load of a server, the number of running services of the server and a historical fault service number of theserver are comprehensively considered; and a recommendation value of each service is determined according to the weight of each feature, and the optimal service is selected according to the size sequence of the recommendation values. The selection method comprehensively considers the real-time operation information of the service, the real-time operation information of the server where the serviceis located and the historical operation information of the server, ensures that the selected service is the optimal service, and solves the problem that the selected service cannot achieve the optimal service calling effect due to the fact that few features are considered during current service selection.