How the cAdvisor gets the container_cpu_load_average_10s? and what is the unit of this metric?

Question

I'm trying to check which container makes the cpu load higher in my cluster.
So that I deployed the cAdvisor daemonset on my k8s cluster with the args below.

args:
- -enable_load_reader=true
- -disable_metrics=sched,network,tcp,advtcp,udp

also .spec.hostNetwork: true

I want to know how the cAdvisor calculates or reports the value of container_cpu_load_average_10s. (how the cAdvisor gets that value)

And if the value of container_cpu_load_average_10s is 100, then that container loads 100 cores? or 100 milicores? or sth else? (the unit of this metric)

And when I checked the metrics (container_cpu_load_average_10s), the value of cadvisor pod's metrics is over 3000.
The others are below 300, but the cadvisor pod's metrics are among 4000~7000.

Please let me know why the values of cadvisor pods' container_cpu_load_average_10s are so high.
Thanks a lot

Answer 1

container_cpu_load_average_10s=loadAvg * 1000 . convert to 'milliLoad' to avoid floats and preserve precision.