WARNING: Once upgraded, the EKS Cluster version cannot be downgraded in-place. If you wish to downgrade, the cluster must be destroyed and recreated with a previous version.
Terraform can upgrade your EKS cluster in-place to a newer version. This will also create new worker node infrastructure to support the new EKS cluster version and destroy the old worker node infrastructure.
Upgrading your EKS cluster to a newer EKS version update the value of cluster_version in your cluster definition examples/quickstart/workspaces/terraform.tfvars to the EKS version desired. A list of EKS version can be found here.
Also ensure that the AMI specified by ami_image_id is compatible with the desired EKS cluster version. Alternatively ami_image_id can be unset, in which case terraform will automatically use a compatible AMI.
Once the cluster definition has been updated run the following to apply the upgrade to your cluster:
NOTE: This may cause an outage during the upgrade when worker nodes are replaced
make apply workspace=<cluster name> path=<path>We've set up automated processes to do blue / green patching of your cluster. An AMI which the worker nodes will be upgraded to is required. In order to run the node patching you can run this using the following command:
make patch workspace=<cluster name> path=<path> ami=<ami>We've set up an automated process to deploy rolling updates to nodes/instances. The rolling patch deploy also ensures that the applications are healthy during patching. In order to run the rolling update, run the following command:
make roll-instances [wait_limit=<wait time for deployments - in seconds>] [max_nodes=<number of nodes to patch>]The command optionally excepts two parameters -
- wait_limit: Time to wait for deployments to be healthy. Default is set to 15 mins (900 seconds).
- max_nodes: Maximum number of nodes to patch in single go. Default is set to 50 nodes.