new: Deploy and monitor ML models with GPUs on Amazon EKS

cluster/eksctl/cluster.yaml

@@ -21,7 +21,7 @@ vpc:
     publicAccess: true
 addons:
   - name: vpc-cni
-    version: 1.16.0
+    version: 1.18.3
     configurationValues: '{"env":{"ENABLE_PREFIX_DELEGATION":"true", "ENABLE_POD_ENI":"true", "POD_SECURITY_GROUP_ENFORCING_MODE":"standard"},"enableNetworkPolicy": "true", "nodeAgent": {"enablePolicyEventLogs": "true"}}'
     resolveConflicts: overwrite
 managedNodeGroups:

lab/iam/policies/iam.yaml

@@ -83,6 +83,7 @@ Statement:
           - eks.amazonaws.com
           - eks-nodegroup.amazonaws.com
           - eks-fargate.amazonaws.com
+          - scraper.aps.amazonaws.com
           - guardduty.amazonaws.com
           - spot.amazonaws.com
           - fis.amazonaws.com

lab/iam/policies/labs1.yaml

@@ -12,14 +12,21 @@ Statement:
   - Effect: Allow
     Action:
       - aps:DeleteWorkspace
+      - aps:DeleteScraper
       - aps:Describe*
       - aps:List*
       - aps:QueryMetrics
+      - aps:CreateScraper
+      - aps:TagResource
     Resource: ["*"]
     Condition:
       StringLike:
         aws:ResourceTag/env:
           - ${Env}*
+  - Effect: Allow
+    Action:
+      - aps:DescribeScraper
+    Resource: ["*"]
   - Effect: Allow
     Action:
       - dynamodb:ListTables

manifests/.workshop/terraform/base.tf

@@ -15,8 +15,8 @@ terraform {
       version = "2.15.0"
     }
     kubectl = {
-      source  = "gavinbunney/kubectl"
-      version = "1.14.0"
+      source  = "alekc/kubectl"
+      version = "2.0.4"
     }
     local = {
       version = "2.5.1"

manifests/.workshop/terraform/outputs.tf

@@ -5,4 +5,5 @@ output "environment" {
 export ${k}='${v}'
 %{endfor}
 EOF
-}
+sensitive = true
+}

manifests/modules/aiml/deploy-monitor-genai-model/.workshop/cleanup.sh

@@ -0,0 +1,14 @@
+#!/bin/bash
+
+kubectl delete ns dogbooth --ignore-not-found
+
+uninstall-helm-chart jupyterhub jupyterhub
+uninstall-helm-chart nginx-ingress default
+uninstall-helm-chart kuberay-operator kuberay
+
+kubectl delete ns jupyterhub --ignore-not-found
+
+# Uninstall gpu-operator
+uninstall-helm-chart gpu-operator gpu-operator
+
+kubectl delete ns gpu-operator --ignore-not-found