In each iteration of microservice system, regression testing should be executed. A large number of repeat testing will cause waste of resources. Therefore, it is necessary to minimize the test suite to reduce costs and to improve testing efficiency. Current test suite minimization technologies mainly rely on system specification and architecture description as input, which is limited to the practicability of microservice system with the characteristics of service autonomy and uncertain call relationship. Moreover, current test suite minimization technologies rarely take the usage scenarios into consideration, and the test suite is difficult to reflect user's concerns. This study proposes a test suite minimization technology based on API gateway access logs mining. This technology mines frequent paths from API gateway access logs which reflects the dynamic operation of microservice system. The relationship between frequent paths and test cases is established to construct search graph. Then, origin test suite is minimized with heuristic search of the graph. This paper explains the whole process of the technology. The experiments based on an integrated OA microservice system show that the scale of the test suite is reduced by more than 40%, and its defect detection ability is reduced by no more than 10%.