Query Monitor

TE v3.7 provides the capability of checking the query initiated recently in the TE system, which could be used by the root admin and system admin having permission to access this screen. Monitor query consists of two parts: live query and query log.

Live Query

Select the scope of monitoring (including the time range, the project, and the cluster used) to check the information related to queries within the monitoring scope, including the user, query types and display names of queries, the time taken for queries, detailed SQL, etc. Click a single record to check detailed information. When queries are blocked occurs, it is reasonable to cancel the query based on various information. Just click "cancel query" in the operation column.

To improve the sensitivity toward abnormal queries, you can set alerts by taking "time taken for real-time query", "time taken for timed queries", and "queue length" as the monitoring metrics. When such metrics exceed the preset value for N consecutive sampling cycle, a pop-up window would be displayed to inform the system admin with relevant permissions of such a situation. If you do not want to receive multiple similar monitor warnings within a short time, you can select "no more prompts within 1h" on the alert.

Query Log

TE would track query behaviors and submit the data to your system to help you analyze and check previous queries. You can check the data generated 90 days ago by enquiring about the logs, identify queries that have taken a relatively long time or with relatively high-performance consumption through combinatorial screening or column sorting, and check the detailed information. If a large number of resources have been consumed due to frequent queries, it is recommended that query behaviors be optimized according to the usage scenes.

If the number of results returned during the query period exceeds 1000 items, 1000 items of results would be displayed. To check all the results returned, you can click "download" to check locally. It should be noted that when the number of log queries is excessively large, the performance of the system would be undermined. Therefore, it is recommended that troubleshooting be conducted when the system is not busy. You can use "real-time monitoring" to identify existing problems when the system is busy.