The early task scheduling algorithm generally completes the static scheduling action from the perspective of the inherent resource usage. This way also lacks the feedback regulation of the running environment, and it is easy to fall into local optimum and cannot achieve the best convergence. To solve this problem, a rule-based collaborative scheduling algorithm for container cloud (CSC2T) is proposed. Firstly, we construct the scheduling rules according to the relationship between nodes and tasks, and describe the problem of finding the best node for scheduling. Then, the nodes are divided into task-related nodes and taskirrelated nodes according to their states. For task-related nodes, the correlation coefficient of twin services and dependent services are described by AHP model. For the task-irrelated nodes, pearson similarity model is used to describe the correlation coefficient of the nodes with twin instances. Combined with two scoring methods, the forward scheduling calculation based on multi-objective is realized. Finally, four kinds of indicators are selected as the feedback evaluation criteria to update the parameters. The experimental part verifies the feasibility of the algorithm from different angles such as resource imbalance degree, resource utilization rate, replica scheduling imbalance degree and dependency scheduling imbalance degree, and improves the overall efficiency of task scheduling in container cloud.