eks-gpu-autoscaling/ref-eksctl/gpu-cluster-ng.yaml at master · DevSecOpsSamples/eks-gpu-autoscaling · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
#
# eksctl create nodegroup -f gpu-cluster-ng.yaml
#
apiVersion: eksctl.io/v1alpha5
kind: ClusterConfig

metadata:
  name: gpu-autoscaling-dev
  region: us-east-1

vpc:
  id: <vpc-id>
  subnets:
    public:
      us-east-1a: { id: <public-subnet1> }
      us-east-1b: { id: <public-subnet2> }
    private:
      us-east-1a: { id: <private-subnet1> }
      us-east-1b: { id: <private-subnet2> }

managedNodeGroups:
- name: cpu-ng
  instanceType: c5.xlarge
  spot: true
  minSize: 2
  maxSize: 10
  desiredCapacity: 2
  privateNetworking: true
  tags:
    nodegroup-role: worker
- name: gpu-ng
  instanceTypes: ['g4dn.xlarge', 'g4dn.2xlarge', 'g4dn.4xlarge']
  spot: true
  minSize: 2
  maxSize: 10
  desiredCapacity: 2
  privateNetworking: true
  labels: {accelerator: nvidia-gpu}
  tags:
    nodegroup-role: worker
# - name: gpu-ng-p2
#   # instanceTypes: ['p2.xlarge', 'g4dn.2xlarge', 'g4dn.4xlarge']
#   spot: true
#   instanceType: p2.xlarge
#   minSize: 2
#   maxSize: 10
#   desiredCapacity: 2
#   privateNetworking: true
#   labels: {accelerator: nvidia-gpu}
#   tags:
#     nodegroup-role: worker