#3079 Inconsistencies between k8s metrics server `kubectl top nodes` and the `HostMetrics Dashboard (K8s)` metrics Issue created by pranay01 Hello everyone I followed the https://signoz.io/docs/tutorial/kubernetes-infra-metrics/#steps-to-export-k8s-metrics-to-signoz|steps to add k8s-metrics to signoz but I see inconsistencies between k8s metrics server

kubectl top nodes

and the

HostMetrics Dashboard (K8s)

metrics For example for one specific node •

kubectl top nodes

reports 13% •

CPU Usage

in

HostMetrics Dashboard (K8s)

reports 20% While looking at the promQL query, I found that CPU Usage is calculated using this formula

1 - (avg by(k8s_node_name) (irate(system_cpu_time{k8s_cluster_name="{{.k8s_cluster_name}}",k8s_node_name="{{.k8s_node_name}}",state="idle"}[5m])))

I also found that it’s CloudWatch CPU utilization metric and definition is similar to the one provided by k8s metrics server Why does Signoz dashboard use a different definition? how can I implement the metrics server / cloud watch definition in Signoz? Slack Message SigNoz/signoz