Halliax

GPUSpotNodeGroup:
  Type: AWS::AutoScaling::AutoScalingGroup
  Properties:
    AutoScalingGroupName: !Sub "${ClusterName}-${NodeGroupName}"
    DesiredCapacity: !Ref NodeAutoScalingGroupDesiredSize # 0
    MinSize: !Ref NodeAutoScalingGroupMinSize # 0
    MaxSize: !Ref NodeAutoScalingGroupMaxSize # 10, arbitrarily
    MixedInstancesPolicy:
      InstancesDistribution:
        OnDemandBaseCapacity: !Ref OnDemandBaseCapacity # 0

Halliax

Parameters:
  ...
  BootstrapArgumentsForSpotFleet:
    Description: Arguments to pass to the bootstrap script. See files/bootstrap.sh in https://github.com/awslabs/amazon-eks-ami
    Type: String
    Default: "--kubelet-extra-args '--node-labels=lifecycle=Ec2Spot,nvidia.com/gpu=true,k8s.amazonaws.com/accelerator=nvidia-tesla
                                    --register-with-taints=spotInstance=true:PreferNoSchedule,nvidia.com/gpu=true:NoSchedule'"
  ...
...
Resources:

Halliax

apiVersion: apps/v1
kind: Deployment
metadata:
  name: cuda-vector-add
  labels:
    app: cuda-vector-add
spec:
  replicas: 3
  selector:
    matchLabels:

Halliax

apiVersion: argoproj.io/v1alpha1
kind: Workflow
metadata:
  generateName: cuda-vector-add-
spec:
  entrypoint: main
  templates:
  - name: main
    # requires this pod to be run on an nvidia.com/gpu labeled node
    nodeSelector: