Observe Metrics Server Data
Explore how to observe and analyze resource usage metrics from Kubernetes pods using the Metrics Server and kubectl commands. Understand how to retrieve CPU and memory consumption data at pod and container levels across namespaces. This lesson prepares you to leverage these metrics for advanced deployment autoscaling strategies.
Resource usage of the nodes is useful but is not what we’re looking for. In this chapter, we’re focused on auto-scaling Pods. But, before we get there, we should observe how much memory each of our Pods is using. We’ll start with those running in the kube-system Namespace.
Memory consumption of pods running in kube-system #
Execute the following from your command line to see the memory consumption of all the pods running in the Kube-system.
kubectl -n kube-system top pod
The output (on Docker For Desktop) is as follows.
NAME CPU(cores) MEMORY(bytes)
etcd-docker-for-desktop 16m 74Mi
kube-apiserver-docker-for-desktop 33m 427Mi
kube-controller-manager-docker-for-desktop 44m 63Mi
kube-dns-86f4d74b45-c47nh 1m 39Mi
kube-proxy-r56kd 2m 22Mi
kube-scheduler-docker-for-desktop 13m 23Mi
We can see resource usage (CPU and memory) for each of the Pods currently running in kube-system. If we do not find better tools, we could use that information to adjust the requests of those Pods to be more accurate. However, there are better ways to get that info, so we’ll skip adjustments for now. Instead, let’s try to get current resource usage of all the Pods, no matter the Namespace.