Slow log as query source but still get expensive perfschema queries

Percona Monitoring and Management (PMM) PMM 2.x
1

Description:

I’ve added a mysql server with slowlog as query source:

pmm-admin add mysql --query-source=slowlog --size-slow-logs=1GiB --disable-tablestats=true --username=pmm --password=xxxxxx mysql-0 10.0.10.10:3306

Can still see these expensive perfschema queries running by the pmm mysql user

select event_name, count_star, sum_timer_wait from performance_schema.events_waits_summary_global_by_event_name

select event_name, count_read, sum_timer_read, sum_number_of_bytes_read, count_write, sum_timer_write, sum_number_of_bytes_write, count_misc, sum_timer_misc from performance_schema.file_summary_by_event_name

Did I miss something? I’ve tried removing the server and adding it again

Version:

ProjectName: pmm-admin
Version: 2.42.0
PMMVersion: 2.42.0
Timestamp: 2024-06-06 15:28:26 (UTC)
FullCommit: 74e57527735bd062c4bd37adbd89c31bb14ebc15

2

Hello @Davesson,

I can see the same behavior in my test.

The --query-source option applies to the QAN dashboard. It specifies where to get the query details for QAN. But for other dashboards, PMM uses system tables like performance_schema to get information on events, locking, replication etc. Because of this, we still see queries running on performance_schema.

3

Hello @Davesson
Regarding expensive performence_schema queries, if you see performance issues due to them, please report them to our bug database with full details.
https://perconadev.atlassian.net