Hi,
I have installed PMM on our internal MongoDB 3.4.10 ubuntu cluster and noticed that the client consumes a lot of resources. This is affecting MongoDB operation execution times in a way that we basically can not use it with our current virtual hardware specs. Having a closer look at one of the the nodes (in this case one of the MongoDB config servers) I realized that this is caused by the percona-qan-agent which seems to use a lot of the cpu time. Stopping the mongodb:queries service with ‘sudo pmm-admin stop mongodb:queries’ brings down the cpu usage to a ‘normal’ level. The remaining two services ‘mongodb:metrics’ and ‘linux:metrics’ behave as expected. I have checked the logs in /var/log but could not find anything suspicious. Whether profiling in MongoDB is enabled or disabled does not really seem to change this. Can anybody give some advice?
Thanks, Bernd