How to make Host's CPU Utilization Metric?

[techministrator]

I believe this should have been answered here and there but searching and using AI tools searching result no such thread for me to rely on. Right now with host_metrics https://vector.dev/docs/reference/configuration/sources/host_metrics/ I'm not sure how to make such metric? Is it the host_cpu_seconds_total metric that I should explore?

[satellite-no]

Not certain what exactly our asking here. With that source I'm sending the data to Prometheus and visualizing in grafana with a query like below. Hope this helps.

sum by(instance) (irate(node_cpu_seconds_total{instance="10.174.66.33:9598",job="vector-nodes", mode="iowait"}[1m0s])) / on(instance) group_left sum by (instance)((irate(node_cpu_seconds_total{instance="<IP>:9598",job="vector-nodes"}[1m0s])))

[techministrator]

Your example does help me figuring out the right metric to use. Though the metric name is not consistent here and there, I'm not sure why. Your example is node_cpu_seconds_total. In the doc, it's host_cpu_seconds_total (ref). For my test with the Amazon EC2 instance, it's just vector.cpu_seconds_total (I defined custom namespace via remap transform) without host_ or node_ prefix after the namespace.

For my "CPU Utilization" metric that I try to have is more about this (Datadog queries)

100 - (sum:vector.cpu_seconds_total{mode:idle,host:my_vector}.as_count().rollup(avg, 60) / sum:vector.cpu_seconds_total{host:my_vector}.as_count().rollup(avg, 60) * 100)

The resulting time series points are usage percentage % of CPU.

Thank you.