{"id":892,"date":"2025-11-25T01:55:16","date_gmt":"2025-11-24T17:55:16","guid":{"rendered":"https:\/\/vinta.ws\/code\/?p=892"},"modified":"2026-02-18T01:20:34","modified_gmt":"2026-02-17T17:20:34","slug":"gke-autopilot-cluster-pay-for-pods-not-nodes","status":"publish","type":"post","link":"https:\/\/vinta.ws\/code\/gke-autopilot-cluster-pay-for-pods-not-nodes.html","title":{"rendered":"GKE Autopilot Cluster: Pay for Pods, Not Nodes"},"content":{"rendered":"

If you're already on Google Cloud, it's highly recommended to use GKE Autopilot Cluster: you only pay for resources requested by your pods (system pods and unused resources are free<\/strong> in Autopilot Cluster). No need to pay for surplus node pools anymore! Plus the entire cluster applies Google's best practices by default.<\/p>\n

ref:
\nhttps:\/\/cloud.google.com\/kubernetes-engine\/pricing#compute<\/a><\/p>\n

Create an Autopilot Cluster<\/h2>\n

DO NOT enable Private Nodes, otherwise you\u00a0MUST<\/strong>\u00a0pay for a Cloud NAT Gateway (~$32\/month) for them to access the internet (to pull images, etc.).<\/p>\n

# create\ngcloud container clusters create-auto my-auto-cluster \n--project YOUR_PROJECT_ID \n--region us-west1\n\n# connect\ngcloud container clusters get-credentials my-auto-cluster \n--project YOUR_PROJECT_ID \n--region us-west1<\/code><\/pre>\n
You can update some configurations later on Google Cloud Console.<\/p>\n
ref:
\nhttps:\/\/docs.cloud.google.com\/sdk\/gcloud\/reference\/container\/clusters\/create-auto<\/a><\/p>\n
Autopilot mode works in both Autopilot and Standard clusters. You don't necessarily need to create a new Autopilot cluster; you can simply deploy your pods in Autopilot mode as long as your Standard cluster meets the requirements:<\/p>\n
gcloud container clusters check-autopilot-compatibility my-cluster \n--project YOUR_PROJECT_ID \n--region us-west1<\/code><\/pre>\n
ref:
\nhttps:\/\/cloud.google.com\/kubernetes-engine\/docs\/concepts\/about-autopilot-mode-standard-clusters<\/a>
\nhttps:\/\/cloud.google.com\/kubernetes-engine\/docs\/how-to\/autopilot-classes-standard-clusters<\/a><\/p>\n
Deploy Workloads in Autopilot Mode<\/h2>\n
The only thing you need to do is add one magical config<\/strong>: nodeSelector: cloud.google.com\/compute-class: "autopilot"<\/code>. That's it. You don't need to create or manage any node pools beforehand, just write some YAMLs and kubectl apply<\/code>. All workloads with cloud.google.com\/compute-class: "autopilot"<\/code> will run in Autopilot mode.<\/p>\n
More importantly, you are only billed for the CPU\/memory resources your pods<\/strong> request, not for nodes that may have unused capacity or system pods (those running under the kube-system<\/code> namespace). Autopilot mode is both cost-efficient and developer-friendly.<\/p>\n
apiVersion: apps\/v1\nkind: Deployment\nmetadata:\n  name: nginx\nspec:\n  replicas: 3\n  selector:\n    matchLabels:\n      app: nginx\n  template:\n    metadata:\n      labels:\n        app: nginx\n    spec:\n      nodeSelector:\n        cloud.google.com\/compute-class: \"autopilot\"\n      containers:\n        - name: nginx\n          image: nginx:1.29.3\n          ports:\n            - name: http\n              containerPort: 80\n          resources:\n            requests:\n              cpu: 100m\n              memory: 128Mi\n            limits:\n              cpu: 500m\n              memory: 256Mi<\/code><\/pre>\n
If your workloads are fault-tolerant (stateless), you can use Spot instances to save a significant amount of money. Just change the nodeSelector<\/code> to autopilot-spot<\/code>:<\/p>\n
apiVersion: apps\/v1\nkind: Deployment\nmetadata:\n  name: nginx\nspec:\n  template:\n    spec:\n      nodeSelector:\n        cloud.google.com\/compute-class: \"autopilot-spot\"\n      terminationGracePeriodSeconds: 25 # spot instances have a 25s warning before preemption<\/code><\/pre>\n
ref:
\nhttps:\/\/docs.cloud.google.com\/kubernetes-engine\/docs\/how-to\/autopilot-classes-standard-clusters<\/a><\/p>\n
You will see something like this in your Autopilot cluster:<\/p>\n
kubectl get nodes\nNAME                             STATUS   ROLES    AGE     VERSION\ngk3-my-auto-cluster-nap-xxx      Ready    <none>   2d18h   v1.33.5-gke.1201000\ngk3-my-auto-cluster-nap-xxx      Ready    <none>   1d13h   v1.33.5-gke.1201000\ngk3-my-auto-cluster-pool-1-xxx   Ready    <none>   86m     v1.33.5-gke.1201000<\/code><\/pre>\n
The nap<\/code> nodes are auto-provisioned by Autopilot for your workloads, while pool-1<\/code> is a default node pool created during cluster creation. System pods may run on either, but in Autopilot cluster, you are never billed for the nodes themselves<\/strong> (neither nap<\/code> nor pool-1<\/code>), nor for the system pods. You only pay for the resources requested by your application pods.<\/p>\n
FYI, the minimum resources for Autopilot workloads are:<\/p>\n