WIP: Raise number of query groups collected in K000 from 100 to 500
Right now, K00* reports are ordered by total_time desc
. It helps to perform top-down analysis and macro-optimization oriented to reducing resource consumption. But sometimes, different kinds of analyses are needed:
- by frequency (order by
calls
): when we want to reduce the frequency of queries having high QpS (such as 1000 per second or more); - by average time (
mean_time
): when we want to improve the experience of particular users; - by data volumes involved (
shared_blks_read/hit
): when we want to reduce actual or possible disk IO.
While these approaches should be covered in new reports in group K in the future, I'm raising the number of queries collected as a fast solution to have more visibility. Previously, we collected only 100 query groups to JSON report. Increasing it to 500 will help to see a wide picture.
Edited by Nikolay Samokhvalov