EKS-CLI

EKS cluster bootstrap with batteries included

• Supports creation of multiple node groups of different types with communication enabled between them
• Taint and label your nodegroups
• Authorize IAM users for cluster access
• Manage IAM policies that will be attached to your nodes
• Easily configure docker repository secrets to allow pulling private images
• Manage Route53 DNS records to point at your Kubernetes services
• Export nodegroups to SpotInst Elastigroups
• Auto resolving AMIs by region & instance types (GPU enabled AMIs)
• Supports both kubernetes 1.12 and 1.13
• Configuration is saved on S3 for easy collaboration

$ gem install eks_cli
$ eks create --kubernetes-version 1.13 --cluster-name my-eks-cluster --s3-bucket my-eks-config-bucket
$ eks create-nodegroup --cluster-name my-eks-cluster --group-name nodes --ssh-key-name <my-ssh-key> --s3-bucket my-eks-config-bucket --yes
$ eks delete-cluster --cluster-name my-eks-cluster --s3-bucket my-eks-config-bucket

You can type eks in your shell to get the full synopsis of available commands

Commands:
  eks add-iam-user IAM_ARN                                     # adds an IAM user as an authorized member on the EKS cluster
  eks create                                                   # creates a new EKS cluster
  eks create-default-storage-class                             # creates default storage class on a new k8s cluster
  eks create-dns-autoscaler                                    # creates kube dns autoscaler
  eks create-nodegroup                                         # creates all nodegroups on environment
  eks delete-cluster                                           # deletes a cluster, including nodegroups
  eks delete-nodegroup                                         # deletes cloudformation stack for nodegroup
  eks enable-gpu                                               # installs nvidia plugin as a daemonset on the cluster
  eks export-nodegroup                                         # exports nodegroup auto scaling group to spotinst
  eks help [COMMAND]                                           # Describe available commands or one specific command
  eks scale-nodegroup                                          # scales a nodegroup
  eks set-docker-registry-credentials USERNAME PASSWORD EMAIL  # sets docker registry credentials
  eks set-iam-policies --policies=one two three                # sets IAM policies to be attached to created nodegroups
  eks set-inter-vpc-networking TO_VPC_ID TO_SG_ID              # creates a vpc peering connection, sets route tables and allows network access on SG
  eks show-config                                              # print cluster configuration
  eks update-auth                                              # update aws auth configmap to allow all nodegroups to connect to control plane
  eks update-cluster-cni                                       # updates cni with warm ip target
  eks update-dns HOSTNAME K8S_SERVICE_NAME                     # alters route53 CNAME records to point to k8s service ELBs
  eks version                                                  # prints eks_cli version
  eks wait-for-cluster                                         # waits until cluster responds to HTTP requests

Options:
  c, [--cluster-name=CLUSTER_NAME]  # eks cluster name (env: EKS_CLI_CLUSTER_NAME)
  s3, [--s3-bucket=S3_BUCKET]       # s3 bucket name to save configurtaion and state (env: EKS_CLI_S3_BUCKET)

1. Ruby
2. kubectl version >= 10 on your PATH
3. aws-iam-authenticator on your PATH
4. aws-cli version >= 1.16.18 on your PATH
5. S3 bucket with write/read permissions to store configuration

You are encouraged to export both EKS_CLI_CLUSTER_NAME and EKS_CLI_S3_BUCKET environment variables instead of using the corresponding flags on each command. It makes the command clearer and reduces the chance for typos. The following selected commands assumes you have exported both environment variables:

export EKS_CLI_S3_BUCKET=my-eks-config-bucket
export EKS_CLI_CLUSTER_NAME=my-eks-cluster

EKS_CLI_S3_BUCKET can be safely put in your ~/.bash_profile and EKS_CLI_CLUSTER_NAME may be exported on a cluster basis

Nodegroups are created separately from the cluster.

You can use eks create-nodegroup multiple times to create several nodegroups with different instance types and number of workers. Nodes in different nodegroups may communicate freely thanks to a shared Security Group.

Scale nodegroups up and down using

$ eks scale-nodegroup --group-name nodes --min 1 --max 10

$ eks add-iam-user arn:aws:iam::XXXXXXXX:user/XXXXXXXX --yes

Edits aws-auth configmap and updates it on EKS to allow an IAM user access the cluster via kubectl

$ eks set-iam-policies --policies=AmazonS3FullAccess AmazonDynamoDBFullAccess

Sets IAM policies to be attached to nodegroups once created. This settings does not work retro-actively - only affects future eks create-nodegroup commands.

$ eks update-dns my-cool-service.my-company.com cool-service --route53-hosted-zone-id=XXXXX --elb-hosted-zone-id=XXXXXX

Takes the ELB endpoint from cool-service and puts it as an alias record of my-cool-service.my-company.com on Route53

$ eks enable-gpu

Installs the nvidia device plugin required to have your GPUs exposed

Assumptions:

1. You have a nodegroup using EKS GPU AMI
2. This nodegroup uses a GPU instance (p2.x / p3.x etc)

$ eks set-docker-registry-credentials <dockerhub-user> <dockerhub-password> <dockerhub-email>

Adds your dockerhub credentials as a secret and attaches it to the default ServiceAccount's imagePullSecrets

$ eks create-default-storage-class

Creates a standard gp2 default storage class named gp2

$ eks create-dns-autoscaler

Creates coredns autoscaler with production defaults

$ eks set-inter-vpc-networking VPC_ID SG_ID

Assuming you have some shared resources on another VPC (an RDS instance for example), this command opens communication between your new EKS cluster and your old VPC:

1. Creating and accepting a VPC peering connection from your EKS cluster VPC to the old VPC
2. Setting route tables on both directions to allow communication
3. Adding an ingress rule to SG_ID to accept all communication from your new cluster nodes.

$ eks export-nodegroup --group-name=other-nodes

Exports the corresponding Auto Scaling Group to a Spotinst Elastigroup

Requires the following environment variables to be set:

• SPOTINST_ACCOUNT_ID
• SPOTINST_API_TOKEN

Is more than welcome! ;)