Recently, both the academic and industrial communities have devoted significant efforts to unveil the true entities behind pseudo-anonymous accounts on the blockchain and to recover stolen funds from vast transaction data. We model blockchain transaction records as transaction graphs and view blockchain transaction tracking as a graph search task. To achieve efficient and effective tracking of fund transfers in transaction graphs, we propose an scalable transaction tracking tool, TRacer. To the best of our knowledge, this is the first tool designed for multiple account-based blockchain platforms and capable of handling complex transaction behaviors in decentralized finance (DeFi). Specifically, we address the transaction tracking task using a subgraph search approach, which employs a novel ranking method to infer the correlation between accounts during the graph search process in multi-relational blockchain transaction graphs. Theoretical analysis and experimental results on datasets from multiple blockchain platforms demonstrate that TRacer can perform transaction tracking tasks more efficiently at lower costs and achieve better tracking results compared to existing methods or even expert manual audits.