Elasticsearch is one of solutions to monitor and analyze logs. Even with ALMA∗, observation logs are stored and anyone can look into it according to their purpose. For example, Hastings, which is a tool discovers the root cause of the defect, is utilized for ACA Correlator subsystem†. It queries logs to an ALMA Elasticsearch storing operational logs, analyzes specific messages which infer troubles, then outputs a result. Before the ALMA Elasticsearch was deployed, logs should have been collected manually in advance. Now the ALMA Elasticsearch has become available and we’ve known: 1) Elasticsearch can directly configure and access features by using REST API, 2) Logs taken even years ago can also be retrieved easily, 3) Elasticsearch’s major update didn’t cause much loss of time to change Hastings, 4) Python has several methods to manage Elasticsearch so that we can choose a favorite one. Therefore, we thought to apply Elasticsearch to the Subaru telescope‡. Size of Subaru logs are quite large but they are not stored in any database yet and just archived. We created a cluster system with Elasticsearch for the evaluation purpose and found ways to store data in a short time. We estimated the total ingestion time for 20 years of telescope status data to be at most 5 months. Our goal is to find a feasible cause of any defects in near real time, to predict any errors that may occur in near future, and to analyze communication between the telescope and observational equipment to optimize observations.
|