Standardization of Helm Chart more explicit in add-ons

awslabs · Jan 31, 2024 · 0cd51b8 · 0cd51b8
1 parent aad1668
commit 0cd51b8
Show file tree

Hide file tree

Showing 5 changed files with 32 additions and 13 deletions.
diff --git a/ai-ml/trainium-inferentia/addons.tf b/ai-ml/trainium-inferentia/addons.tf
@@ -250,6 +250,11 @@ module "eks_blueprints_addons" {
       clusterName: ${module.eks.cluster_name}
       ec2NodeClass:
         karpenterRole: ${split("/", module.eks_blueprints_addons.karpenter.node_iam_role_arn)[1]}
+        subnetSelectorTerms:
+          id: ${module.vpc.private_subnets[3]}
+        securityGroupSelectorTerms:
+          tags:
+            Name: ${module.eks.cluster_name}-node
       nodePool:
         labels:
           - provisioner: default
@@ -280,6 +285,11 @@ module "eks_blueprints_addons" {
       clusterName: ${module.eks.cluster_name}
       ec2NodeClass:
         karpenterRole: ${split("/", module.eks_blueprints_addons.karpenter.node_iam_role_arn)[1]}
+        subnetSelectorTerms:
+          id: ${module.vpc.private_subnets[3]}
+        securityGroupSelectorTerms:
+          tags:
+            Name: ${module.eks.cluster_name}-node
       nodePool:
         labels:
           - provisioner: inferentia-inf2
@@ -305,9 +315,6 @@ module "eks_blueprints_addons" {
           - key: "kubernetes.io/arch"
             operator: In
             values: ["amd64"]
-          - key: "topology.kubernetes.io/zone"
-            operator: In
-            values: ["${local.region}d"]
           - key: "karpenter.sh/capacity-type"
             operator: In
             values: ["spot", "on-demand"]
@@ -393,7 +400,7 @@ resource "random_password" "grafana" {
 
 #tfsec:ignore:aws-ssm-secret-use-customer-key
 resource "aws_secretsmanager_secret" "grafana" {
-  name                    = "${local.name}-oss-grafana"
+  name                    = "${local.name}-oss-grafana-remove"
   recovery_window_in_days = 0 # Set to zero for this example to force delete during Terraform destroy
 }
 

diff --git a/ai-ml/trainium-inferentia/examples/llama2/install-pre-requsites-for-ec2.sh b/ai-ml/trainium-inferentia/examples/llama2/install-pre-requsites-for-ec2.sh
@@ -81,5 +81,3 @@ install_helm
 install_boto3
 
 echo "Installation of prerequisites complete."
-
-
diff --git a/ai-ml/trainium-inferentia/helm-values/karpenter-resources/templates/node-class.yaml b/ai-ml/trainium-inferentia/helm-values/karpenter-resources/templates/node-class.yaml
@@ -11,14 +11,28 @@ spec:
     {{- toYaml .Values.ec2NodeClass.amiSelectorTerms | nindent 4 }}
   {{- end }}
   subnetSelectorTerms:
+    {{- if .Values.ec2NodeClass.subnetSelectorTerms.tags }}
     - tags:
-      {{- range $key, $value := .Values.ec2NodeClass.subnetSelectorTags }}
+      {{- range $key, $value := .Values.ec2NodeClass.subnetSelectorTerms.tags }}
         {{ $key }}: {{ $value | quote }}
       {{- end }}
-      karpenter.sh/discovery: {{ .Values.clusterName }}
+    {{- end }}
+    {{- if .Values.ec2NodeClass.subnetSelectorTerms.id }}
+    - id: {{ .Values.ec2NodeClass.subnetSelectorTerms.id }}
+    {{- end }}
   securityGroupSelectorTerms:
+    {{- if .Values.ec2NodeClass.securityGroupSelectorTerms.name }}
+    - name: {{ .Values.ec2NodeClass.securityGroupSelectorTerms.name }}
+    {{- end }}
+    {{- if .Values.ec2NodeClass.securityGroupSelectorTerms.id }}
+    - id: {{ .Values.ec2NodeClass.securityGroupSelectorTerms.id }}
+    {{- end }}
+    {{- if .Values.ec2NodeClass.securityGroupSelectorTerms.tags }}
     - tags:
-        Name: {{ .Values.clusterName }}-node
+      {{- range $key, $value := .Values.ec2NodeClass.securityGroupSelectorTerms.tags }}
+        {{ $key }}: {{ $value | quote }}
+      {{- end }}
+    {{- end }}
   role: {{ .Values.ec2NodeClass.karpenterRole }}
   tags:
     Name: karpenter-{{ .Values.name }}

diff --git a/ai-ml/trainium-inferentia/helm-values/karpenter-resources/values.yaml b/ai-ml/trainium-inferentia/helm-values/karpenter-resources/values.yaml
@@ -1,14 +1,14 @@
 # Shared values
 name: default
-clusterName:
+clusterName: test-cluster
 
 # EC2NodeClass specific values
 ec2NodeClass:
   enabled: true
   amiFamily: AL2
   amiSelectorTerms:
-  subnetSelectorTags:
-    kubernetes.io/role/internal-elb: "1"  # Added here to make sure that will be scheduled in private subnets
+  subnetSelectorTerms: # tag or id see documentation, https://karpenter.sh/docs/concepts/nodeclasses/
+  securityGroupSelectorTerms: # tag, name, id see documentation, https://karpenter.sh/docs/concepts/nodeclasses/
   karpenterRole:
   metadataOptions:
     httpEndpoint: enabled

diff --git a/ai-ml/trainium-inferentia/main.tf b/ai-ml/trainium-inferentia/main.tf
@@ -50,4 +50,4 @@ locals {
     Blueprint  = local.name
     GithubRepo = "github.com/awslabs/data-on-eks"
   }
-}
+}
Original file line number	Diff line number	Diff line change
Expand Up		@@ -81,5 +81,3 @@ install_helm
		install_boto3

		echo "Installation of prerequisites complete."
-Original file line number
+Diff line change
@@ Expand Up / @@ -50,4 +50,4 @@ locals { @@
         Blueprint  = local.name
         GithubRepo = "github.com/awslabs/data-on-eks"
       }
-    }
+    }