Skip to main content

Create spot capacity

Lets deploy a managed node group that creates Spot instances, followed by modifying the existing catalog component of our application to run on the newly created Spot instances.

We can get started by listing all of the nodes in our existing EKS cluster. The kubectl get nodes command can be used to list the nodes in your Kubernetes cluster, but to get additional detail about the capacity type, we'll use the -L eks.amazonaws.com/capacityType parameter.

The following command shows that our nodes are currently on-demand instances.

~$kubectl get nodes -L eks.amazonaws.com/capacityType
NAME                                          STATUS   ROLES    AGE    VERSION                CAPACITYTYPE
ip-10-42-103-103.us-east-2.compute.internal   Ready    <none>   133m   v1.29-eks-48e63af    ON_DEMAND
ip-10-42-142-197.us-east-2.compute.internal   Ready    <none>   133m   v1.29-eks-48e63af    ON_DEMAND
ip-10-42-161-44.us-east-2.compute.internal    Ready    <none>   133m   v1.29-eks-48e63af    ON_DEMAND
tip
If you want to retrieve nodes based on a specific capacity type, such as on-demand instances, you can utilize "label selectors". In this particular scenario, you can achieve this by setting the label selector to capacityType=ON_DEMAND.
~$kubectl get nodes -l eks.amazonaws.com/capacityType=ON_DEMAND
NAME                                         STATUS   ROLES    AGE     VERSION
ip-10-42-10-119.us-east-2.compute.internal   Ready    <none>   3d10h   v1.29-eks-48e63af
ip-10-42-10-200.us-east-2.compute.internal   Ready    <none>   3d10h   v1.29-eks-48e63af
ip-10-42-11-94.us-east-2.compute.internal    Ready    <none>   3d10h   v1.29-eks-48e63af
ip-10-42-12-235.us-east-2.compute.internal   Ready    <none>   4h34m   v1.29-eks-48e63af
In the below diagram, there are two separate "node groups" representing the managed node groups within the cluster. The first Node Group box represents the node group containing On-Demand instances while the second represents the node group containing Spot instances. Both are associated with the specified EKS cluster.
spot arch
Let's create a node group with Spot instances. The following command creates a new node group managed-spot.
~$aws eks create-nodegroup \
  --cluster-name $EKS_CLUSTER_NAME \
  --nodegroup-name managed-spot \
  --node-role $SPOT_NODE_ROLE \
  --subnets $PRIMARY_SUBNET_1 $PRIMARY_SUBNET_2 $PRIMARY_SUBNET_3 \
  --instance-types c5.large c5d.large c5a.large c5ad.large c6a.large \
  --capacity-type SPOT \
  --scaling-config minSize=2,maxSize=3,desiredSize=2 \
  --disk-size 20
The --capacity-type SPOT argument indicates that all capacity in this managed node group should be Spot.
tip
The aws eks wait nodegroup-active command can be used to wait until a specific EKS node group is active and ready for use. This command is part of the AWS CLI and can be used to ensure that the specified node group has been successfully created and all the associated instances are running and ready.
~$aws eks wait nodegroup-active \
  --cluster-name $EKS_CLUSTER_NAME \
  --nodegroup-name managed-spot
Once our new managed node group is Active, run the following command.
~$kubectl get nodes -L eks.amazonaws.com/capacityType,eks.amazonaws.com/nodegroup
NAME                                          STATUS   ROLES    AGE     VERSION                CAPACITYTYPE   NODEGROUP
ip-10-42-103-103.us-east-2.compute.internal   Ready    <none>   3h38m   v1.29-eks-48e63af    ON_DEMAND      default
ip-10-42-142-197.us-east-2.compute.internal   Ready    <none>   3h38m   v1.29-eks-48e63af    ON_DEMAND      default
ip-10-42-161-44.us-east-2.compute.internal    Ready    <none>   3h38m   v1.29-eks-48e63af    ON_DEMAND      default
ip-10-42-178-46.us-east-2.compute.internal    Ready    <none>   103s    v1.29-eks-48e63af    SPOT           managed-spot
ip-10-42-97-19.us-east-2.compute.internal     Ready    <none>   104s    v1.29-eks-48e63af    SPOT           managed-spot
The output shows that two additional nodes got provisioned under the node group managed-spot with capacity type as SPOT.