Hello,
In the “Top tables by rows read” dashboard, reads from performance_schema.events_statements_history appears to constantly increase over time until PMM is restarted (see attached screenshot, although this applies to all of our monitored instances). I’m guessing this is an error with the way the metric is collected, rather than an accurate statistic. can you please confirm?
Edit: I should add that we use performance_schema as a source for query analytics, in case that makes a difference
Cheers
Hi MickMc
That’s correct, it is due to fetching results from
PERFORMANCE_SCHEMA.events_statements_history
that is done by qan-agent (Query Analytics). This shouldn’t be an error.
[url]qan-agent/worker.go at 744504971772163e451fd0f4289e160b72d58c17 · percona/qan-agent · GitHub
You can evaluate the impact by stopping mysql:queries on one node and observe the line should flatten.
pmm-admin stop mysql:queries