Kubernetes Cluster Autoscaler for AWS EKS

Terraform module for deploy Cluster Autoscaler for AWS EKS to your kubernetes cluster.

Wokrflow

This module creates all necessary resources for deploy AWS Cluster Autoscale inside your kubernetes cluster. It helps you to manage your EKS cluster node size (as EC2 Auto Scaling Group) based on internal kubernetes metrics.

Usage

module "cluster_autoscaler" {
  source = "../"

  aws_region   = "us-west-1"
  cluster_name = "my-EKS-cluster-name"

  // Optional
  resources = {
    request_cpu    = "150m"
    request_memory = "70Mi"
  }
}

AWS Auto Scaling Group TAGs requirements

Please add next aws tags to your Auto Scaling Group

k8s.io/cluster-autoscaler/enabled = true
k8s.io/cluster-autoscaler/CLUSTER_NAME = true

You can add additional Cluster Autoscaler container arguments via additional_deployment_commands. See all possible args in official Cluster Autoscaler docs:

 additional_deployment_commands = [
    "--balance-similar-node-groups=true",
    "--skip-nodes-with-local-storage=false"
 ]

NOTE: If you want to use other metrics, please set 'service_monitor_enable' to true and set namespace where your monitoring system deployed:

 service_monitor_enable = true
 metrics_namespace      = "monitoring-prometheus"

Inputs

Name	Description	Type	Default	Example	Required
cloud_provider	Cloud Provier name (support only 'aws')	string	aws	n/a	no
aws_region	Region where EKS deployed	string	n/a	us-east-1	yes
aws_role_arn	AWS role for Cluster Autoscaler	string	null	arn:aws:iam::123456789:role/eks-oidc-cluster-autoscaler	no
app_name	Name of Cluster Autoscaler	string	cluster-autoscaler	n/a	no
namespace	Name of kubernetes namespace for cluster autoscaler	string	cluster-autoscaler	n/a	no
create_namespace	Create namespace by module ? true or false	bool	true	n/a	no
cluster_name	EKS cluster name	string	n/a	my-EKS-cluste-name	yes
node_selector	Choose node where you want to deploy Cluster Autoscaler	map(string)	null	{ spot_node_pool = true service_node_pool = false }	no
service_monitor_enable	Enable Service Monitor? If true set variable 'metrics_namespace'	bool	false	n/a	no
metrics_namespace	Namespace where metrics collector (such as Prometheus) is deployed	string	null	monitoring	no
prometheus_release_label	Prometheus release label ( typically name of helm chart )	string	kube-prometheus-stack	kube-prometheus-stack	no
rbac_psp_enable	Enable Pod Security Policies (PSP)	bool	false	n/a	no
docker_image	Image for cluster autoscaler	string	k8s.gcr.io/autoscaling/cluster-autoscaler:v1.21.1	n/a	no
additional_deployment_commands	Your additional args to Cluster Autoscaler deployment	list(string)	[]	[ "--balance-similar-node-groups=true", "--skip-nodes-with-local-storage=false" ]	no
service_ports	Ports for auth request from ingress to service	list(object({ name = string internal_port = string external_port = string }))	[ { name = "http" internal_port = "http" external_port = "http" } ]	n/a	no
resources	Your Cluster Autoscaler deployment limits	map(string)	{ request_cpu = "100m" request_memory = "300Mi" }	n/a	no

Outputs

Name	Description
name	Name of Cluster Autoscaler
namespace	Name of namespace where Cluster Autoscaler deployed
service_account_name	Name of Cluster Autoscaler service account

Software Requirements

Name	Description
Terraform	>= 0.14.9
Kubernetes provider	>= 2.6.1