LOG ANOMALY DETECTION ON KUBERNETES-BASED MICROSERVICE WITH AUTOLOG

• Main Author: Fahreza, Afif
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/75284
• Institution: Institut Teknologi Bandung

The research on log anomaly detection in Kubernetes-based microservices with AutoLog presents a study on anomaly detection in microservice logs using the AutoLog method implemented in a Kubernetes environment. This study aims to evaluate the performance of AutoLog in detecting anomalies in system logs generated by microservices running in a Kubernetes cluster and identify the conditions required for retraining the log anomaly detection process using the AutoLog method. The methodology employed in this research is MLOps, which aligns with the development of systems supported by machine learning. Grafana Loki is utilized as the supporting tool for logging in Kubernetes. The effectiveness of AutoLog is assessed through fault injection using Chaos Mesh and comparing the AutoLog prediction results with the anomalies induced by fault injection. The research findings demonstrate that AutoLog is effective in detecting anomalies in system logs generated by microservices running on Kubernetes, achieving an F1 score of 0.943. Additionally, this study identifies the necessary conditions for retraining the log anomaly detection process to ensure optimal performance.