support ExtraNodeLabels and auto create node label NVIDIAGPU=1 for GP…

…U instances(#2 #3)
aws-samples · Jul 7, 2019 · e4726c2 · e4726c2
1 parent 0d90748
commit e4726c2
Show file tree

Hide file tree

Showing 3 changed files with 32 additions and 6 deletions.
diff --git a/Makefile b/Makefile
@@ -100,6 +100,7 @@ create-eks-cluster:
 	ParameterKey=KeyName,ParameterValue="$(SSH_KEY_NAME)" \
 	ParameterKey=LambdaRoleArn,ParameterValue="$(EKS_ADMIN_ROLE)" \
 	ParameterKey=OnDemandBaseCapacity,ParameterValue="$(OnDemandBaseCapacity)" \
+	ParameterKey=NodeImageId,ParameterValue="$(NodeImageId)" \
 	ParameterKey=NodeAutoScalingGroupMinSize,ParameterValue="$(NodeAutoScalingGroupMinSize)" \
 	ParameterKey=NodeAutoScalingGroupDesiredSize,ParameterValue="$(NodeAutoScalingGroupDesiredSize)" \
 	ParameterKey=NodeAutoScalingGroupMaxSize,ParameterValue="$(NodeAutoScalingGroupMaxSize)" \
@@ -127,6 +128,7 @@ update-eks-cluster:
 	ParameterKey=KeyName,ParameterValue="$(SSH_KEY_NAME)" \
 	ParameterKey=LambdaRoleArn,ParameterValue="$(EKS_ADMIN_ROLE)" \
 	ParameterKey=OnDemandBaseCapacity,ParameterValue="$(OnDemandBaseCapacity)" \
+	ParameterKey=NodeImageId,ParameterValue="$(NodeImageId)" \
 	ParameterKey=NodeAutoScalingGroupMinSize,ParameterValue="$(NodeAutoScalingGroupMinSize)" \
 	ParameterKey=NodeAutoScalingGroupDesiredSize,ParameterValue="$(NodeAutoScalingGroupDesiredSize)" \
 	ParameterKey=NodeAutoScalingGroupMaxSize,ParameterValue="$(NodeAutoScalingGroupMaxSize)" \

diff --git a/cloudformation/eks.yaml b/cloudformation/eks.yaml
@@ -51,6 +51,11 @@ Parameters:
   #   #Type: List<String>
   #   Type: String
 
+  NodeImageId:
+    Type: String
+    Default: ""
+    Description: Specify AMI id for the node instances. Leave it blank if you preferred the latest EKS-optimized AMI
+
   NodeVolumeSize:
     Type: Number
     Description: Node volume size
@@ -194,11 +199,13 @@ Resources:
         NodeAutoScalingGroupDesiredSize: !Ref NodeAutoScalingGroupDesiredSize
         NodeAutoScalingGroupMaxSize: !Ref NodeAutoScalingGroupMaxSize
         NodeVolumeSize: !Ref NodeVolumeSize
+        NodeImageId: !Ref NodeImageId
         OnDemandBaseCapacity: !Ref OnDemandBaseCapacity
         LatestAMIId: !GetAtt AMI.Outputs.AmiId
         Subnets: !Ref SubnetIds
         ExistingNodeSecurityGroups: !Ref ExistingNodeSecurityGroups
         ExtraNodeSecurityGroups: !Ref ExtraNodeSecurityGroups
+
 
 
         # RootStackName: !Sub "${AWS::StackName}"

diff --git a/cloudformation/nodegroup.yaml b/cloudformation/nodegroup.yaml
@@ -230,15 +230,20 @@ Parameters:
   ClusterName:
     Description: The cluster name provided when the cluster was created.  If it is incorrect, nodes will not be able to join the cluster.
     Type: String
+
+  ExtraNodeLabels:
+    Description: Extra Node Labels(seperated by comma)
+    Type: String
+    Default: "foo=bar"
 
   BootstrapArgumentsForOnDemand:
     Description: Arguments to pass to the bootstrap script. See files/bootstrap.sh in https://github.com/awslabs/amazon-eks-ami
-    Default: "--kubelet-extra-args --node-labels=lifecycle=OnDemand"
+    Default: "--kubelet-extra-args --node-labels=lifecycle=OnDemand,$ExtraNodeLabels"
     Type: String
 
   BootstrapArgumentsForSpotFleet:
     Description: Arguments to pass to the bootstrap script. See files/bootstrap.sh in https://github.com/awslabs/amazon-eks-ami
-    Default: "--kubelet-extra-args '--node-labels=lifecycle=Ec2Spot --register-with-taints=spotInstance=true:PreferNoSchedule'"
+    Default: "--kubelet-extra-args '--node-labels=lifecycle=Ec2Spot,$ExtraNodeLabels --register-with-taints=spotInstance=true:PreferNoSchedule'"
     Type: String
 
   ClusterControlPlaneSecurityGroup:
@@ -249,6 +254,11 @@ Parameters:
     Description: Unique identifier for the Node Group.
     Type: String
 
+  PauseTime:
+    Description: Pause Time
+    Type: String
+    Default: PT1M
+
   # RootStackName:
   #   Type: String
 
@@ -435,7 +445,7 @@ Resources:
       AutoScalingRollingUpdate:
         MinInstancesInService: !Ref NodeAutoScalingGroupDesiredSize
         MaxBatchSize: '1'
-        PauseTime: 'PT5M'
+        PauseTime: !Ref PauseTime
 
   LCH:
     Type: AWS::AutoScaling::LifecycleHook
@@ -473,14 +483,21 @@ Resources:
             set -o xtrace
             yum install -y ec2-instance-connect
             iid=$(curl -s http://169.254.169.254/latest/meta-data/instance-id)
+            it=$(curl -s http://169.254.169.254/latest/meta-data/instance-type)
             export AWS_DEFAULT_REGION=${AWS::Region}
             ilc=`aws ec2 describe-instances --instance-ids  $iid  --query 'Reservations[0].Instances[0].InstanceLifecycle' --output text`
+            if [[ $it == p* ]] || [[ $it == g* ]]; then
+              # GPU instance type
+              ExtraNodeLabels="${ExtraNodeLabels},NVIDIAGPU=1"
+            else
+              ExtraNodeLabels="${ExtraNodeLabels}"
+            fi
             if [ "$ilc" == "spot" ]; then
-              /etc/eks/bootstrap.sh ${ClusterName} ${BootstrapArgumentsForSpotFleet}
+              bash -c "/etc/eks/bootstrap.sh ${ClusterName} ${BootstrapArgumentsForSpotFleet}"
             else
-              /etc/eks/bootstrap.sh ${ClusterName} ${BootstrapArgumentsForOnDemand}
+              bash -c "/etc/eks/bootstrap.sh ${ClusterName} ${BootstrapArgumentsForOnDemand}"
             fi
-            # /etc/eks/bootstrap.sh ${ClusterName} $BootstrapArgumentsForOnDemand
+            /etc/eks/bootstrap.sh ${ClusterName} $BootstrapArgumentsForOnDemand
             /opt/aws/bin/cfn-signal --exit-code $? \
                      --stack  ${AWS::StackName} \
                      --resource NodeGroup  \