Cluster Autoscaler

This is a cloud-agnostic controller, for more info check the Kubernetes tech-docs directly

• What this controller does is to adjust the resource AWS::AutoScaling::AutoScalingGroup of the NodeGroups based on the total usage of the cluster
• The ASG already have min and max nodes defined at the NodeGroup creation. Autoscaler will obey this boundaries. Example: if the max number of nodes is 4 but there are still pending pods, autoscaler won't launch new nodes.

Permissions

• The controller needs access to the AutoScalingGroup AWS Resource in order scale out/in the EC2 Instances
• Pass the flag --asg-access when creating the NodeGroup via eksctl so that the workloads running in the worker nodes have access to ASG