docs: terraforming an EKS cluster with autoscaling and EFS.

docs/setup-cluster/k8s/setup-eks-cluster.rst

@@ -10,6 +10,10 @@
    information that is obsolete. This documentation is preserved because it may contain useful
    insights relevant to legacy systems.
 
+   See `Github repo <https://github.com/determined-ai/determined/tree/main/examples/deploy/eks>` for
+   an up-to-date example for terraform code deploying Determined on EKS with autoscaling and EFS
+   support.
+
 Determined can be installed on a cluster that is hosted on a managed Kubernetes service such as
 `Amazon EKS <https://aws.amazon.com/eks/>`_. This document describes how to set up an EKS cluster
 with GPU-enabled nodes. The recommended setup includes deploying a cluster with a single non-GPU

examples/deploy/eks/.gitignore

@@ -0,0 +1,3 @@
+.terraform
+terraform.tfstate*
+.terraform.tfstate*

examples/deploy/eks/README.md

@@ -0,0 +1,43 @@
+# Terraformed EKS cluster for Determined
+
+This is an example terraform code to configure an EKS cluster to run Determined on.
+
+Supported features:
+- autoscaling via Karpenter,
+- postgresql volume on EBS,
+- shared fs on EFS.
+
+Based on [original Karpenter example](https://github.com/terraform-aws-modules/terraform-aws-eks/tree/master/examples/karpenter)
+
+## Prerequisites
+
+- terraform
+- helm
+- aws CLI
+
+## Installation
+
+First, edit the `locals` section in `main.tf` to set your cluster name and AWS region.
+
+```bash
+$ terraform init
+$ terraform apply -auto-approve
+$ aws eks --region us-west-2 update-kubeconfig --name <CLUSTER NAME>
+$ helm install determined determined-ai/determined --values values.yaml
+```
+
+## Teardown
+
+Warning: shut down all the jobs in determined first.
+
+```bash
+$ helm uninstall determined
+$ terraform destroy -auto-approve
+```
+
+## Future work
+
+In the future, we may want to:
+- Make the code configurable: currently, custom configurations will require changing the terraform code directly.
+- Rework this code as `det deploy eks` utility.
+- Switch from a postgres instance installed by helm and using an EBS volume to a terraform-provisioned RDS.