A Root Cause Localization Method Based on Event Call Chains for Microservices
Yu Chen, Zhiming Xiao, Fei Teng · 2024
The localization of root causes in microservice systems is critical for quickly restoring services. However, existing root cause localization methods that rely on single monitoring data have stringent requirements for target data, while those based on multiple monitoring sources struggle with modeling the relationships of different monitoring data. This article proposes a method for locating root causes in microservices based on event call chains. First, the method constructs and extracts trace and log data from the system during the normal phase into events, unifying the representation of monitoring data through these events. Next, leveraging the contextual relationships between events trained using LSTM, the method builds a set of all possible event call chains during the diagnosis phase, effectively addressing the challenges posed by the microservice chain coordination mechanism. Finally, the most likely root cause set is generated based on the event call chains. Experiments demonstrate that the proposed method offers higher localization accuracy and faster positioning times compared to existing approaches.