In the realm of Kubernetes (K8s) resource scheduling, studies frequently focus on individual resource efficiency or performance metrics. Alternatively, they may merely refine K8s' default scheduling policy and often lead to a local optimal solution dilemma. Particularly when dealing with a large number of Pods and high problem complexity, their effects are nonideal in load balancing and resource utilization. To clear such dilemma, this paper initially examines four resources: CPU, memory utilization, disk I/O, and network bandwidth. Subsequently, baed on K8s' default scheduling strategy and the Elitist Non-dominated Sorting Genetic Algorithm II, we devolop a Kubernetes Non-dominated Sorting Genetic Algorithm II (KU-NSGA-II) algorithm with dual objectives of resource utilization and load balancing. Comparison experiments demonstrate that the proposed algorithm outperforms the K8s default scheduling policy in terms of load balancing and resource utilization. In addition, the feasility of the proposed algorithm is verified by a real case.