I installed kind to play around with Kubernetes.
If I use top and sort by CPU usage (key C), then I see that kube-apiserver is constantly consuming 5 to 10% CPU.
Why?
I don't have installed something up to now:
guettli@p15:~$ kubectl get pods --all-namespaces
NAMESPACE NAME READY STATUS RESTARTS AGE
kube-system coredns-558bd4d5db-ntg7c 1/1 Running 0 40h
kube-system coredns-558bd4d5db-sx8w9 1/1 Running 0 40h
kube-system etcd-kind-control-plane 1/1 Running 0 40h
kube-system kindnet-9zkkg 1/1 Running 0 40h
kube-system kube-apiserver-kind-control-plane 1/1 Running 0 40h
kube-system kube-controller-manager-kind-control-plane 1/1 Running 0 40h
kube-system kube-proxy-dthwl 1/1 Running 0 40h
kube-system kube-scheduler-kind-control-plane 1/1 Running 0 40h
local-path-storage local-path-provisioner-547f784dff-xntql 1/1 Running 0 40h
guettli@p15:~$ kubectl get services --all-namespaces
NAMESPACE NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE
default kubernetes ClusterIP 10.96.0.1 <none> 443/TCP 40h
kube-system kube-dns ClusterIP 10.96.0.10 <none> 53/UDP,53/TCP,9153/TCP 40h
guettli@p15:~$ kubectl get nodes
NAME STATUS ROLES AGE VERSION
kind-control-plane Ready control-plane,master 40h v1.21.1
guettli@p15:~$ kubectl get nodes --all-namespaces
NAME STATUS ROLES AGE VERSION
kind-control-plane Ready control-plane,master 40h v1.21.1
I am curious. Where does the CPU usage come from? How can I investigate this?
Even in an empty cluster with just one master node, there are at least 5 components that reach out to the API server on a regular basis:
This is because API Server acts as the only entry point for all components in Kubernetes to know what the cluster state should be and take action if needed.
If you are interested in the details, you could enable audit logs in the API server and get a very verbose file with all the requests being made. How to do so is not the goal of this answer, but you can start from the apiserver documentation.