Log message with JSON item count for root cause analysis in microservices

Abstract

System administrator takes time to find application error in one microservice caused by HTTP response from another microservice. Although Istio default log message includes the data length of HTTP response body, it doesn’t indicate the structure of HTTP response body such as key-value pair counts. Therefore, system administrator takes time to identify root cause of error by Istio default log message and needs to have skills in fault diagnosis. This paper proposes an algorithm that calculates an indicator for application error identification. The algorithm counts items such as map or list in JSON format from HTTP response body as the indicator. The indicator represents the body structure of HTTP response and is recorded in log messages as an additional field. The indicator enables system administrator to identify application error by log messages. The experiment measures log message length increase by the indicator and response time increase by the indicator calculation. The average response time of the proposed log format increases by 7% compared with Istio default log format. The average log message length increase in the proposed method is 216 bytes compared to Istio default log format.