Currently query history stored in the memory of TajoMaster or QueryMaster. Query detail information is disappeared when the cluster is restarted. And because of memory limitations, only a few hours or a few days histories are kept.
I suggest the following features for the query history.
- Worker saves a query task history to the HDFS after finishing a query.
- QueryMaster saves a query history to the HDFS after finishing a query.
- If query is finished, Information Web server in TajoMaster shows query history in the web page, If query is running TajoMaster forward to QueryMaster.
- History data will be saved to HDFS with JSON format.
- JSON format has history file version for backward compatibility.