Scaling allows the EKS cluster to dynamically adjust to varying workloads, ensuring efficient resource utilization and cost management.
Kubernetes scaling types:
- manual
- automatic
Manual scaling can be performed with kubectl scale command which sets a new size for a deployment, replica set, replication controller, or stateful set.
Auto-scaling options provided in Amazon EKS:
- native to Kubernetes
- Horizontal Pod Autoscaler
- Cluster Autoscaler
- through AWS-specific features
- EKS Auto Mode
- Auto Scaling Groups
- Fargate
- 3rd party AWS-specific solutions
- Karpenter
- Vertical Pod Autoscaler (VPA) — adjusts CPU/memory requests for Pods (not installed by default in EKS).
- Kubernetes Event-Driven Autoscaler (KEDA) — adds event-driven scaling (e.g., SQS, Kafka, CloudWatch events). It’s widely used with EKS but not installed by default.
Kubernetes Cluster Autoscaler
The Kubernetes Cluster Autoscaler is designed to automatically adjust the number of nodes in your cluster based on the resource requests of the workloads running in the cluster.
A Kubernetes project, supported on EKS, that adjusts the number of nodes (EC2 instances) in your cluster based on pending Pods that cannot be scheduled due to resource constraints.
It works with node groups backed by AWS Auto Scaling Groups.
Key Features:
- Node Scaling: It adds or removes nodes based on the pending pods that cannot be scheduled due to insufficient resources.
- Pod Scheduling: Ensures that all pending pods are scheduled by scaling the cluster up.
How to check if it's enabled?
Installation and Setup:
To use the Cluster Autoscaler in the EKS cluster we need to deploy it using a Helm chart or a pre-configured YAML manifest.
kubectl apply -f https://raw.githubusercontent.com/kubernetes/autoscaler/master/cluster-autoscaler/cloudprovider/aws/examples/cluster-autoscaler-autodiscover.yaml
Configuration:
- Ensure the --nodes flag in the deployment specifies the min and max nodes for your node group.
- Annotate your node groups with the k8s.io/cluster-autoscaler tags to enable autoscaler to manage them.
EKS Auto Mode
AWS’s managed autoscaling approach for the cluster compute layer; it uses concepts similar to Karpenter under the hood.
AWS Autoscaling Groups
AWS Auto Scaling Groups (ASGs) can also be used to scale the worker nodes in your EKS cluster.
Key Features:
- EC2 Instance Scaling: Automatically adjusts the number of EC2 instances in the group.
Installation and Setup:
When you create EKS managed node groups, they are automatically managed by ASGs. (This is the default auto scaling provider in EKS cluster as it does not require installing any additional tools, it's provided out of the box when we create a node group.)
eksctl create nodegroup --cluster your-cluster-name --name your-nodegroup-name --nodes-min 1 --nodes-max 10
AWS Fargate
AWS Fargate allows you to run Kubernetes pods without managing the underlying nodes. It provides serverless compute for containers, eliminating the need to provision and scale EC2 instances.
Key Features:
- Serverless: No need to manage EC2 instances.
- Automatic Scaling: Automatically scales pods based on the specified compute resources.
Installation and Setup:
Create a Fargate profile for your EKS cluster, specifying which pods should run on Fargate.
eksctl create fargateprofile --cluster your-cluster-name --name your-fargate-profile --namespace your-namespace
Karpenter
Karpenter itself is an open-source cluster autoscaler (node scaler) that can be deployed to your EKS cluster and rapidly launches right-sized EC2 compute when needed. AWS supports it, but you must install/configure it, similar to Cluster Autoscaler.
No comments:
Post a Comment