The current interbank transaction system is huge and complicated. How to quickly locate and analyze the cause of transaction failure from the massive amount of logs generated by the transaction system is very important. This paper proposes an analysis method based on the transaction log slice, which cuts the whole log from the start to the end of a transaction into slice. Then it leverages Hadoop framework and executes big data analysis with predefined scripts, which can locate the error information and find the solution more rapidly. A fault analysis system is implemented with this method. The experiments show that the method proposed in this paper can obviously improve the efficiency of transaction log query and fault analysis, which will reduce the operational cost of interbank transaction system.