looking for options to reduce frequency of metrics collection in PMM for mongodb


We are running PMM 1.17.0 built from AWS AMI, so running on an AWS EC2. Using PMM just for mongodb clusters. Installed PMM server and clients 02/2019, so everything is recent and at default values.

Since mongodb clusters tend to have a lot of nodes to monitor, I want to try and keep the CPU load on the PMM server to reasonable levels and hence keep the instance type cost down.

To do this I thought that reducing the frequency with which PMM collects metrics would help. Would welcome thoughts on that and other options.

In the past I was using PMM for mysql, with a docker install. That gave me the option of METRICS_RESOLUTION, but the AMI install of PMM doesn’t have that option.

I’ve looked at the prometheus datasource on the pmm web interface, and that has a scrape_interval setting. Is this all I need to set? I am kind of doubtful because it mentions in the help popup that this is driven or set from the yml config. Or do I need to alter settings within the /etc/prometheus.yml file and bounce prometheus.

I did search thru the forum and help and there was mention of advice to not edit the prometheus.yml because of instability.


Hi mthomas

You’re correct that there isn’t a supported way to modify METRIC_RESOLUTION when deploying the AWS Marketplace image for PMM 1.

You can attempt to edit the yml file and restart Prometheus, however the file will be updated automatically during normal server operation which might revert your changes unexpectedly.

In PMM 2 we will offer a way via GUI and API to persist setting changes, including METRIC_RESOLUTION. We’re looking at first public builds to be out by end the end of March.

Thanks Michael. We’ll upgrade when pmm 2 is out.