Diagnostic log collecting system for accelerator control at SPring-8

In this paper, we report on a diagnostic log collecting system for accelerator control system at SPring-8. The SPring-8 accelerator control system is based on distributed computers using approximately 360 hosts. Diagnostic logs (e.g. messages between control terminals and equipments, records of the data collection) from each host are collected and transferred into a single log server, and are investigated once a trouble occurs. However the amount of logs transferred to the log server reach the size of 30 GByte per day, which is equivalent of 30 billion rows. It is very time-consuming to utilize these logs because of its large volume and varieties in the data format. To improve the efficiency, a log analysis system using fluentd and kibana is constructed and for the moment applied to monitor computer resources and web logs. We found that the log analysis system is efficient to find the causes of the trouble immediately. (author)