Cost optimization in Kubernetes - Commands & Configuration

Introduction

Running applications on Kubernetes can use more resources than needed, which costs money. Cost optimization helps you use just the right amount of resources so you save money without hurting your apps.

When your Kubernetes cluster bills are higher than expected and you want to reduce costs.

When you want to avoid paying for unused CPU or memory in your app containers.

When you want to automatically adjust resources based on app demand to save money.

When you want to find and remove idle or unnecessary workloads in your cluster.

When you want to set limits so no app uses more resources than it should.

Commands

Check current CPU and memory usage of all nodes to see if any are underused.

Terminal

kubectl top nodes

Expected Output

NAME           CPU(cores)   MEMORY(bytes)

worker-node-1  250m         512Mi

worker-node-2  100m         256Mi

worker-node-3  50m          128Mi

See resource usage of all pods to find which ones use too much or too little CPU and memory.

Terminal

kubectl top pods --all-namespaces

Expected Output

NAMESPACE     NAME                      CPU(cores)   MEMORY(bytes)

default       my-app-1                  100m        200Mi

default       my-app-2                  50m         150Mi

kube-system   coredns-558bd4d5db-7x9zq  10m         50Mi

→

--all-namespaces - Shows pods from all namespaces, not just the current one

Save the pod configuration to a file so you can edit resource requests and limits.

Terminal

kubectl get pod my-app-1 -o yaml > my-app-1.yaml

Expected Output

No output (command runs silently)

→

-o yaml - Outputs the pod details in YAML format

Apply the updated pod configuration with optimized resource requests and limits to save costs.

Terminal

kubectl apply -f my-app-1.yaml

Expected Output

pod/my-app-1 configured

Check if Horizontal Pod Autoscaler is set up to automatically adjust pod count based on load.

Terminal

kubectl get hpa

Expected Output

NAME       REFERENCE           TARGETS         MINPODS   MAXPODS   REPLICAS   AGE

my-app-hpa Deployment/my-app   50%/80%        1         5         2          10m

Key Concept

If you remember nothing else from this pattern, remember: monitor your resource usage and set proper requests and limits to avoid paying for unused capacity.

Common Mistakes

Not setting resource requests and limits in pod specs.

Without limits, pods can use more resources than expected, causing higher costs and unstable clusters.

Always define CPU and memory requests and limits in your pod or deployment YAML files.

Ignoring idle or low-usage nodes and pods.

Idle resources waste money because you pay for capacity that is not used.

Regularly check resource usage with kubectl top and remove or scale down unused workloads.

Not using autoscaling features like Horizontal Pod Autoscaler.

Without autoscaling, you either over-provision (waste money) or under-provision (poor performance).

Set up autoscaling to adjust pod counts based on real demand automatically.

Summary

Use 'kubectl top' commands to monitor node and pod resource usage.

Edit pod resource requests and limits in YAML files to control how much CPU and memory each pod can use.

Apply changes with 'kubectl apply' to update running workloads.

Use Horizontal Pod Autoscaler to automatically scale pods based on load and save costs.