The invention discloses a node exception judgment method in a cloud computing environment. The method comprises a node performance data acquisition step; a node performance data conversion step of training a node anomaly detection model; judging a node state based on the performance data; and a node anomaly detection model updating step. According to the invention, by collecting and analyzing nodeperformance data and comparing states of adjacent nodes, abnormal nodes can be judged in time. The sliding time window mode is used, the anomaly detection model is continuously updated, unknown anomalies can be recognized, and node anomalies are judged in a self-adaptive mode in the cloud computing environment.