Characterized by decentralization, irreversibility, and traceability, blockchain systems have garnered significant attention recently. However, performance limitations have emerged as one of the primary obstacles in blockchain applications. For blockchain systems with multiple layers and highly complicated interactions among nodes, detecting the nodes and stages where bottlenecks occur has proven challenging. Metrics like TPS and latency fail to offer sufficient detailed information. To solve this problem, we propose 3 metrics and corresponding methods based on the spatio-temporal state transition of transactions. The spatio-temporal state transition cost serves as an indicator reflecting the overall operational status of blockchain systems. A method to detect abnormalities in blockchain using spatio-temporal state transition cost is also proposed. The spatial state transition costs and temporal state transition costs offer additional detailed insights into the blockchain system. We introduce methods based on these indicators for locating abnormal nodes and abnormal stages. We implement the framework on ChainMaker with a log-based method and deploy it on 16 machines using docker and Kubernetes. We verify the stability and sensitivity of metrics and the precision of anomaly detection method. The experimental results show that our metrics can reliably reflect the operational status of blockchain systems and accurately pinpoint nodes and stages with abnormalities.