HRCA: A Heterogeneous Graph-based Adaptive Root Cause Analysis Framework
Enyu Yu, Hui Dong, Yuxiang Ren, Minzhi Yan, Xuecang Zhang, Yi Ping Yang, Le Yue, Zhengbin Huang · 2023
The paper introduces HRCA, a Heterogeneous graph-based Root Cause Analysis framework for large-scale cloud platforms. As cloud platforms expand rapidly, ensuring stability and reliability becomes increasingly important. However, the dynamic and complex call relationships among services, along with the massive infrastructure components, pose challenges to Root Cause Analysis (RCA). HRCA addresses these challenges by providing an adaptive root cause analysis plan that integrates data from multiple sources and employs unified heterogeneous graph modeling. The framework leverages a subgraph extracting module to improve efficiency and accuracy, as well as a supervised random walk algorithm for diagnosing root causes. Comparative evaluations with MicroRCA, AutoMAP, MicroDiag and Groot demonstrate that HRCA outperforms these state-of-the-art methods in terms of accuracy and generalization ability. Currently, HRCA is actively deployed on Huawei Cloud Stack platform for root cause analysis in production environments.