LogDC: Problem Diagnosis for Declartively-Deployed Cloud Applications with Log
- 1 November 2017
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 282-287
Abstract
Recently, as the evolution of application's development and management paradigms, the deployment declaration becomes a standard interface connecting application developers and Cloud platforms. Kuberenetes is such a system for automating deployment, scaling, and management of micro-service based applications. However, managing and operating such a cloud benefit with additional complexities from the declarative deployment. This paper proposes a log model based problem diagnosis tool for declaratively-deployed cloud applications with the full lifecycle Kubernetes logs. With the runtime logs and deployment declarations, we can pinpoint the root causes in terms of abnormal declarative items and log entries. The advantage of this approach is that we provide a precise log model of a normal deployment to help diagnose problems. The experimental results show that our approach can find out the anomalies of some real-world Kubernetes problems, some of which have been confirmed as bugs. Within the given fault types, our approach can pinpoint the root causes at 91% in Precision and at 92% in Recall.Keywords
This publication has 10 references indexed in Scilit:
- LogSed: Anomaly Diagnosis through Mining Time-Weighted Control Flow Graph in Logs2017
- LogMine2016
- Log clustering based problem identification for online service systems2016
- LOGAN: Problem Diagnosis in the Cloud Using Log-Based Reference Models2016
- Applied Logistic Regression2013
- Declarative automated cloud resource orchestration2011
- ELT: Efficient Log-based Troubleshooting System for Cloud Computing Infrastructures2011
- Online System Problem Detection by Mining Patterns of Console Logs2009
- Declarative versus Imperative Process Modeling Languages: The Issue of Understandability2009
- Naive Bayes for Text Classification with Unbalanced Classes2006