{"id":22305,"date":"2025-03-18T18:31:39","date_gmt":"2025-03-18T13:01:39","guid":{"rendered":"https:\/\/opstree.com\/blog\/?p=22305"},"modified":"2025-07-15T21:40:50","modified_gmt":"2025-07-15T16:10:50","slug":"eks-cluster-autoscaler","status":"publish","type":"post","link":"https:\/\/opstree.com\/blog\/2025\/03\/18\/eks-cluster-autoscaler\/","title":{"rendered":"EKS Cluster AutoScaler"},"content":{"rendered":"

The Amazon EKS AutoScaler automatically adjusts the number of worker nodes in an <\/span><\/span><\/span>Amazon EKS cluster based on resource utilization and scaling demands.<\/span><\/span><\/span><\/p>\n

Before we started using EKS Cluster Autoscaler<\/span><\/span>, managing node scaling in our Kubernetes <\/span><\/span>cluster was a constant struggle. Here\u2019s what we were dealing with:<\/span><\/span><\/span>

\nHigh Costs: We had to keep extra EC2 nodes running all the time to avoid pod scheduling <\/span><\/span>issues<\/span><\/span> \u2014<\/span><\/span> even when they weren\u2019t being used.<\/span><\/span><\/span>

\nPod Failures: If we tried to reduce node count to save money, pods would often stay Pending<\/span><\/span>\u00a0due to lack of resources.<\/span><\/span><\/span>

\nManual Work: Scaling nodes up or down was a manual task<\/span><\/span>\u2014<\/span><\/span>time<\/span><\/span>–<\/span><\/span>consuming and error<\/span><\/span>–<\/span><\/span>prone.<\/span><\/span><\/span>

\nSlow Deployments: During deployments or traffic spikes, our workloads would get stuck <\/span><\/span>because there weren\u2019t enough resources.<\/span><\/span><\/span>

\nUnpredictable Demand: Our workloads change frequently, but our cluster capacity doesn\u2019t<\/span><\/span>\u2014<\/span><\/span> leading to overuse or underuse of nodes.<\/span><\/span><\/span>

\nThese challenges pushed us to explore Cluster Autoscaler, and it turned out to be a game<\/span><\/span>–<\/span><\/span>changer<\/span><\/span><\/span><\/p>\n

<\/p>\n

Overview<\/span><\/span><\/h2>\n

Setting up the Cluster Autoscaler on Amazon EKS using Helm can be tricky. Most guides <\/span><\/span>cover the basic steps, but there\u2019s an important tip that\u2019s often overlooked. In this post, <\/span><\/span>we\u2019ll share this helpful trick and walk you through the entire process step by<\/span><\/span> step.<\/span><\/span><\/p>\n

Prerequisites<\/h3>\n
\n

Before we dive into the deployment process, ensure you have the following:<\/span><\/span><\/span><\/p>\n

    \n
  1. An active Amazon EKS cluster<\/span><\/span><\/span><\/li>\n
  2. Helm installed on your local machine<\/span><\/span><\/span>
    <\/span><\/li>\n
  3. AWS CLI<\/span><\/span> configured with appropriate permissions<\/span><\/span><\/li>\n
  4. Kubectl<\/span><\/span> is\u00a0installed and configured to interact with your EKS cluster.<\/span><\/span><\/li>\n<\/ol>\n

    \"\"<\/p>\n<\/div>\n

    Procedure<\/h3>\n

    1. Add the Cluster Autoscaler Helm repository<\/strong><\/h4>\n
      \n
    • First, add the official Cluster Autoscaler<\/span><\/span> Helm repository:<\/span><\/span><\/li>\n<\/ul>\n
      \n
      helm repo<\/span> add<\/span> autoscaler <\/span><\/span>https:\/\/kubernetes.github.io\/autoscaler<\/span><\/span>
      <\/span><\/em>helm repo update<\/span><\/span><\/strong>
      <\/span><\/div>\n<\/div>\n
      <\/div>\n
      <\/div>\n
      <\/div>\n
      <\/div>\n
      \"\"<\/div>\n
      <\/div>\n

      2. Create IAM policy for Cluster Autoscaler:<\/strong><\/h4>\n

      The Cluster Autoscaler needs specific permissions to interact with AWS services. Create <\/span><\/span>an IAM policy with the required permissions:<\/span><\/span><\/p>\n

      \n
      aws iam<\/span> create<\/span>–<\/span>policy<\/span> –<\/span> policy<\/span>–<\/span>name<\/span> AmazonEKSClusterAutoscalerPolicy<\/span> – <\/span>policy<\/span>–<\/span>document <\/span><\/span>file:\/\/cluster<\/span>–<\/span>autoscaler<\/span>–<\/span>policy.json<\/span><\/span><\/em><\/strong><\/div>\n<\/div>\n
      \n
      \n
      <\/div>\n
      {
      \n“Version”<\/span><\/span>:<\/span><\/span> “2012<\/span><\/span>–<\/span><\/span>10<\/span><\/span>–<\/span><\/span>17″<\/span><\/span>,<\/span><\/span><\/div>\n<\/div>\n
      \n
      “Statement”<\/span><\/span>:<\/span><\/span> [<\/span><\/span><\/span><\/span>
      {<\/span><\/span><\/span><\/span>
      “Action”<\/span><\/span>:<\/span><\/span> [<\/span><\/span><\/span><\/span>
      “autoscaling:DescribeAutoScalingGroups”<\/span><\/span>,<\/span><\/span><\/span><\/span>
      “autoscaling:DescribeAutoScalingInstances”<\/span><\/span>,<\/span><\/span><\/span><\/span>
      “autoscaling:DescribeLaunchConfigurations”<\/span><\/span>,<\/span><\/span><\/span><\/span>
      “autoscaling:DescribeTags”<\/span><\/span>,<\/span><\/span><\/span><\/span>
      “autoscaling:SetDesiredCapacity”<\/span><\/span>,<\/span><\/span><\/span><\/span>
      “autoscaling:TerminateInstanceInAutoScalingGroup”<\/span><\/span>,<\/span><\/span><\/span><\/span>
      “ec2:DescribeLaunchTemplateVersions”<\/span><\/span><\/span><\/span>
      ],<\/span><\/span><\/span><\/span>
      “Resource”<\/span><\/span>:<\/span><\/span> “*”<\/span><\/span>,<\/span><\/span><\/span><\/span>
      “Effect”<\/span><\/span>:<\/span><\/span> “Allow”<\/span><\/span><\/span><\/div>\n
      }<\/div>\n
      ]<\/div>\n
      }<\/div>\n<\/div>\n<\/div>\n
      <\/div>\n

      3. Create<\/span> IAM<\/span>–<\/span>OIDC Provider:<\/span><\/strong><\/h4>\n
      <\/div>\n
      First, check if an OIDC provider is already associated:<\/span><\/span>
      aws eks<\/span> describe<\/span>–<\/span>cluster<\/span> —<\/span>name<\/span> <your<\/span>–<\/span>cluster<\/span>–<\/span>name><\/span> —<\/span>query<\/span><\/strong><\/em>
      “cluster.identity.oidc.issuer”<\/span> —<\/span>output<\/span> text<\/span><\/strong><\/em><\/div>\n
      <\/div>\n
      If this<\/span><\/span> command returns an<\/span><\/span> OIDC<\/span><\/span> URL,<\/span><\/span> you\u2019re good<\/span><\/span> \u2014<\/span><\/span> you already<\/span><\/span> have an<\/span><\/span>
      OIDC provider.<\/span><\/span><\/div>\n
      <\/div>\n
      If it’s<\/span><\/span> empty or<\/span><\/span> errors, then you need to create it.<\/span><\/span><\/div>\n
      <\/div>\n
      \n
      \n
      eksctl<\/span> utils<\/span> associate<\/span>–<\/span>iam<\/span>–<\/span>oidc<\/span>–<\/span>provider<\/span> \/<\/span>
      —<\/span>region=ap<\/span>–<\/span>south<\/span>–<\/span>1 \/<\/span>
      —<\/span>cluster=testing<\/span> \/<\/span>
      —<\/span>approve<\/span><\/span><\/strong><\/div>\n<\/div>\n<\/div>\n
      <\/div>\n
      <\/div>\n
      <\/div>\n
      <\/div>\n
      <\/div>\n
      \n
      \n
      <\/div>\n<\/div>\n<\/div>\n
      \"\"<\/div>\n
      <\/div>\n

      4. Create IAM role for Cluster Autoscaler<\/strong><\/h4>\n
      <\/div>\n
      This step allows Cluster Autoscaler to access EC2 and Auto Scaling resources securely. <\/span><\/span>We create an IAM role and link it to a Kubernetes service account using IRSA.<\/span><\/span><\/div>\n
      <\/div>\n
      Next<\/span><\/span><\/span>, create an IAM role and attach the policy:<\/span><\/span><\/span><\/div>\n
      <\/div>\n
      \n
      \n
      eksctl<\/span> create iamserviceaccount<\/span> \\<\/span>
      –<\/span> cluster=<your<\/span>–<\/span>cluster<\/span>–<\/span>name><\/span> \\<\/span>
      –<\/span> namespace=kube<\/span>–<\/span>system<\/span> \\<\/span>cluster<\/span>
      –<\/span> name=cluster<\/span>–<\/span>autoscaler<\/span> \\<\/span>
      –<\/span> attach<\/span>–<\/span>policy<\/span>–<\/span>arn=arn:aws:iam::<your<\/span>–<\/span>account<\/span>–<\/span>
      id>:policy\/AmazonEKSClusterAutoscalerPolicy<\/span> \\<\/span>
      –<\/span> override<\/span>–<\/span>existing<\/span> –<\/span>serviceaccounts<\/span> \\<\/span>
      –<\/span> approve<\/span><\/span><\/strong><\/div>\n

      This command <\/span><\/span>does three<\/strong><\/span>\u00a0things<\/span>:<\/span><\/strong><\/span><\/div>\n

      \u2022<\/span><\/span><\/span>Creates\u00a0<\/span><\/span>a Kubernetes<\/strong><\/span>\u00a0service\u00a0<\/span><\/strong>account\u00a0<\/strong>named<\/span>\u00a0cluster<\/span><\/span>–<\/span><\/span>autoscaler in the kube<\/span><\/span>–<\/span><\/span>system namespace<\/span><\/span><\/span>
      \u2022<\/span><\/span><\/span>Creates\u00a0<\/span><\/span>an IAM <\/strong><\/span>role\u00a0<\/strong>and<\/span>\u00a0attaches\u00a0<\/span><\/span>the Cluster<\/strong><\/span>\u00a0Autoscaler policy<\/span><\/strong><\/span>
      \u2022<\/span><\/span><\/span>Links the IAM role to the service account\u00a0<\/span><\/span>using<\/span> OIDC identity provider<\/span><\/strong><\/span><\/div>\n<\/div>\n<\/div>\n
      <\/div>\n
      <\/div>\n
      <\/div>\n
      <\/div>\n
      <\/div>\n
      <\/div>\n
      <\/div>\n
      <\/div>\n
      \"\"<\/div>\n
      \n
      \n

      <\/h4>\n


      5. Deploy Cluster Autoscaler using Helm<\/span><\/strong><\/span><\/h4>\n

      Now, here\u2019s the trick that often goes unmentioned: When deploying the Cluster <\/span><\/span>Autoscaler using Helm, you need to set the `awsRegion` value explicitly. Th<\/span><\/span><\/span><\/span>is ensures that the autoscaler works correctly with your specific AWS region:<\/div>\n
      \n
      \n
      <\/div>\n
      helm install<\/span> cluster<\/span>–<\/span>autoscaler<\/span> autoscaler\/cluster<\/span>–<\/span>autoscaler<\/span> \\<\/span><\/span>
      —<\/span>namespace<\/span> kube<\/span>–<\/span>system<\/span> \\<\/span><\/span>
      —<\/span>set autoDiscovery.clusterName=testing<\/span> \\<\/span><\/span>
      —<\/span>set<\/span> awsRegion=ap<\/span>–<\/span>south<\/span>–<\/span>1<\/span> \\<\/span><\/span>
      —<\/span>set rbac.serviceAccount.create=false<\/span> \\<\/span><\/span>
      —<\/span>set rbac.serviceAccount.name=cluster<\/span>–<\/span>autoscaler<\/span><\/span><\/strong><\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n
      <\/div>\n
      <\/div>\n
      <\/div>\n
      <\/div>\n
      <\/div>\n
      <\/div>\n
      <\/div>\n
      \n
      \n
      \n
      \n
      \"\"<\/span><\/div>\n
      <\/div>\n


      6. Verify the deployment:<\/span><\/strong><\/span><\/h4>\n

      Check if the Cluster Autoscaler pod is running:<\/span><\/span><\/span><\/div>\n

      kubectl<\/span> get<\/span> pods<\/span> –<\/span>n kube<\/span>–<\/span>system<\/span> |<\/span> grep<\/span> cluster<\/span>–<\/span>autoscaler<\/span><\/strong><\/span><\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n
      <\/div>\n
      <\/div>\n
      <\/div>\n
      <\/div>\n
      <\/div>\n
      <\/div>\n
      \"\"<\/div>\n
      \n
      \n
      \n
      \n
      <\/div>\n


      7. After Verifying also check for Asg Policies for Scale Up<\/span> And<\/span> Scale<\/span>–<\/span>down<\/span><\/strong>
      The nodes using Autoscaler:<\/span><\/strong><\/span><\/h4>\n

      If they are not there add these two policies for scale<\/span><\/span>–<\/span><\/span>up and scale<\/span><\/span>–<\/span><\/span>down<\/span><\/span><\/span><\/div>\n
      \n


      aws autoscaling put<\/span>–<\/span>scaling<\/span>–<\/span>policy<\/span> —<\/span>policy<\/span>–<\/span>name ScaleOutPolicy<\/span> —<\/span>auto<\/span>–<\/span><\/strong>scaling<\/span>–<\/span>group<\/span>–<\/span>name <<\/span>your<\/span>–<\/span> auto<\/span>–<\/span>scaling<\/span>–<\/span>group<\/span>–<\/span>name ><\/span> —<\/span>scaling<\/span>–<\/span><\/strong><\/span><\/p>\n

      <\/div>\n<\/div>\n<\/div>\n
      \n
      adjustment 1<\/span> —<\/span>adjustment<\/span>–<\/span>type ChangeInCapa<\/span>city<\/span><\/span><\/strong><\/div>\n

      aws autoscaling put<\/span>–<\/span>scaling<\/span>–<\/span>policy<\/span> —<\/span>policy<\/span>–<\/span>name ScaleInPolicy<\/span> —<\/span>auto<\/span>–<\/span><\/strong>
      scaling<\/span>–<\/span>group<\/span>–<\/span>name<\/span> <your<\/span>–<\/span> auto<\/span>–<\/span>scaling<\/span>–<\/span>group<\/span>–<\/span>name ><\/span> —<\/span>scaling<\/span>–<\/span><\/strong>
      adjustment<\/span><\/strong><\/span><\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n
      <\/div>\n
      <\/div>\n
      <\/div>\n
      <\/div>\n
      <\/div>\n
      <\/div>\n
      <\/div>\n
      \n
      \n
      \n
      \n
      \"\"<\/span><\/div>\n
      <\/div>\n

      We are done with eks Autoscaler we need to verify the Autoscaler<\/span> is<\/span><\/strong>
      working fine or not so we are making an ingress file for the nginx<\/span>–<\/span><\/strong>
      webserver<\/span> and testing it.<\/span><\/strong><\/span><\/div>\n

      These are the steps:<\/span>–<\/span><\/strong><\/span><\/div>\n

      1<\/span>. Check for active nodes:<\/span><\/strong><\/span><\/h5>\n

      kubectl<\/span> get<\/span> nodes<\/span><\/strong><\/span><\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n
      <\/div>\n
      <\/div>\n
      <\/div>\n
      <\/div>\n
      <\/div>\n
      <\/div>\n
      \n
      \n
      \n
      \n
      \"\"<\/div>\n

      2. Create a Test Ingress file for checking Autoscaler:<\/span><\/strong><\/span><\/h5>\n

      apiVersion<\/span>:<\/span> apps\/v1<\/span><\/strong><\/span>
      kind<\/span>:<\/span> Deployment<\/span><\/span>
      metadata<\/span>:<\/span><\/span>
      name<\/span>:<\/span> webserver<\/span><\/span>
      spec<\/span>:<\/span><\/span>
      replicas<\/span>:<\/span> 20<\/span><\/span>
      selector<\/span>:<\/span><\/span>
      matchLabels<\/span>:<\/span><\/span>
      app<\/span>:<\/span> nginx<\/span>–<\/span>webserver<\/span><\/span>
      template<\/span>:<\/span><\/span>
      metadata<\/span>:<\/span><\/span>
      labels<\/span>:<\/span><\/span>
      app<\/span>:<\/span> nginx<\/span>–<\/span>webserver<\/span><\/span>
      spec<\/span>:<\/span><\/span>
      containers<\/span>:<\/span><\/span><\/strong><\/div>\n<\/div>\n
      \n
      –<\/span> name<\/span>:<\/span> nginx<\/span>–<\/span>container<\/span><\/span>
      image<\/span>:<\/span> nginx:latest<\/span><\/span>
      resources<\/span>:<\/span><\/span>
      requests<\/span>:<\/span><\/span>
      memory<\/span>:<\/span> “1Gi”<\/span><\/span>
      cpu<\/span>:<\/span> 1<\/span><\/span>
      <\/span><\/strong>—<\/span><\/span><\/div>\n
      <\/div>\n

      3. Apply this Ingress file and check for the nodes to scale<\/span>–<\/span>in<\/span><\/strong><\/span><\/h5>\n

      Nodes became 2 to 10 after deploying ingress<\/span><\/strong><\/span><\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n
      <\/div>\n
      <\/div>\n
      <\/div>\n
      <\/div>\n
      <\/div>\n
      <\/div>\n
      <\/div>\n
      <\/div>\n
      \n
      \n
      \n
      \n
      \"\"<\/div>\n
      <\/div>\n

      4. Now I Can scale out from 1o to 2 I put a replica number to 2 here<\/span> and after <\/span><\/strong>that<\/span> Check for nodes that are scaling down or not<\/span><\/strong><\/span><\/h5>\n

      now apply this and check for the nodes to scale out<\/span><\/strong><\/span><\/p>\n

      as we see 2 nodes are there only because the load is less and it is scale-out<\/span><\/strong>
      successfully<\/span><\/strong><\/span>
      <\/span><\/p>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n

      <\/div>\n
      <\/div>\n
      \"\"<\/div>\n
      \n
      \n
      \n
      \n
      <\/div>\n

      Conclusion:<\/span>–<\/span><\/span><\/strong><\/h2>\n


      This document guides you<\/span><\/span> through setting up EKS Cluster Autoscaler to automatically <\/span><\/span>scale your worker nodes based on pod requirements. By enabling OIDC, creating the <\/span><\/span>necessary IAM role, and deploying the autoscaler, you ensure your cluster scales <\/span><\/span>efficiently, reduces manual effort<\/span><\/span>, and optimizes resource utilization. <\/span><\/span><\/span><\/p>\n

      Let the Cluster Autoscaler handle node scaling<\/span><\/span> \u2014<\/span><\/span> so you can focus on running your <\/span><\/span>workloads, not managing infrastructure<\/span><\/span><\/span><\/p>\n<\/div>\n

      CONTACT US<\/span><\/a><\/p>\n<\/div>\n<\/div>\n<\/div>\n

      <\/div>\n","protected":false},"excerpt":{"rendered":"

      The Amazon EKS AutoScaler automatically adjusts the number of worker nodes in an Amazon EKS cluster based on resource utilization and scaling demands. Before we started using EKS Cluster Autoscaler, managing node scaling in our Kubernetes cluster was a constant struggle. Here\u2019s what we were dealing with: High Costs: We had to keep extra EC2 … Continue reading “EKS Cluster AutoScaler”<\/span><\/a><\/p>\n","protected":false},"author":244582695,"featured_media":22323,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_coblocks_attr":"","_coblocks_dimensions":"","_coblocks_responsive_height":"","_coblocks_accordion_ie_support":"","jetpack_post_was_ever_published":false,"_jetpack_newsletter_access":"","_jetpack_dont_email_post_to_subs":false,"_jetpack_newsletter_tier_id":0,"_jetpack_memberships_contains_paywalled_content":false,"_jetpack_memberships_contains_paid_content":false,"footnotes":"","jetpack_publicize_message":"","jetpack_publicize_feature_enabled":true,"jetpack_social_post_already_shared":true,"jetpack_social_options":{"image_generator_settings":{"template":"highway","enabled":false},"version":2}},"categories":[36349927,28070474],"tags":[768739503,768739506,768739504,768739505,392947966,768739507,768739508],"jetpack_publicize_connections":[],"jetpack_featured_media_url":"https:\/\/opstree.com\/blog\/wp-content\/uploads\/2025\/03\/EKS-Cluster-AutoScaler-.png","jetpack_likes_enabled":true,"jetpack_sharing_enabled":true,"jetpack_shortlink":"https:\/\/wp.me\/pfDBOm-5NL","jetpack-related-posts":[],"_links":{"self":[{"href":"https:\/\/opstree.com\/blog\/wp-json\/wp\/v2\/posts\/22305"}],"collection":[{"href":"https:\/\/opstree.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/opstree.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/opstree.com\/blog\/wp-json\/wp\/v2\/users\/244582695"}],"replies":[{"embeddable":true,"href":"https:\/\/opstree.com\/blog\/wp-json\/wp\/v2\/comments?post=22305"}],"version-history":[{"count":7,"href":"https:\/\/opstree.com\/blog\/wp-json\/wp\/v2\/posts\/22305\/revisions"}],"predecessor-version":[{"id":22330,"href":"https:\/\/opstree.com\/blog\/wp-json\/wp\/v2\/posts\/22305\/revisions\/22330"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/opstree.com\/blog\/wp-json\/wp\/v2\/media\/22323"}],"wp:attachment":[{"href":"https:\/\/opstree.com\/blog\/wp-json\/wp\/v2\/media?parent=22305"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/opstree.com\/blog\/wp-json\/wp\/v2\/categories?post=22305"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/opstree.com\/blog\/wp-json\/wp\/v2\/tags?post=22305"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}