JupyterHub Demo Files

README.md

These are the configuration files used to setup the video demo shown in the presentation.

They use eksctl to setup a Kubernetes cluster with AWS EKS, and then Helm v2 to setup cluster components.

Note the cluster is hard-coded to eu-west-1a region & availability zone for simplicity, but this is not be suitable for a production setup.

autoscaler.yaml

# Install autoscaler using Helm https://github.com/helm/charts/tree/master/stable/cluster-autoscaler
# Replace <cluster-name> in this file with your eksctl cluster name
# helm upgrade --install cluster-autoscaler stable/cluster-autoscaler --namespace kube-system --values autoscaler.yaml
image:
  tag: v1.16.1
autoDiscovery:
  clusterName: <cluster-name>
awsRegion: eu-west-1
rbac:
  create: true
nodeSelector:
  role: main

jupyterhub.yaml

# Install JupyterHub with Helm https://zero-to-jupyterhub.readthedocs.io/en/latest/setup-jupyterhub.html
# helm upgrade --install jupyter jupyterhub/jupyterhub --namespace jupyter --version=0.8.2 --values jupyterhub.yaml 

# Setup Google auth as described in docs https://zero-to-jupyterhub.readthedocs.io/en/latest/authentication.html#google-oauth
auth:
  type: google
  google:
    clientId: ""
    clientSecret: ""
    callbackUrl: ""
    hostedDomain: ""
    loginService: ""

proxy:
  # Generate random secret token `openssl rand -hex 32`
  secretToken: ""
  https:
    enabled: true
    type: offload
  service:
    annotations:
      # Certificate ARN, setup certificate in AWS ACM
      service.beta.kubernetes.io/aws-load-balancer-ssl-cert: ""
      # The protocol to use on the backend, we use TCP since we're using websockets
      service.beta.kubernetes.io/aws-load-balancer-backend-protocol: "tcp"
      # Which ports should use SSL
      service.beta.kubernetes.io/aws-load-balancer-ssl-ports: "https"
      service.beta.kubernetes.io/aws-load-balancer-connection-idle-timeout: "3600"

singleuser:
  image:
    name: jupyter/datascience-notebook
    tag: 1386e2046833
  profileList:
    - display_name: "Large environment"
      description: "R language in a large Spot Instance environment"
      default: true
      kubespawner_override:
        cpu_guarantee: 1
        mem_guarantee: "6G"
        node_selector: # Links up with labels in nodes.yaml
          size: large
          pricing: spot
    - display_name: "Extra Large environment"
      description: "Python in a extra large Spot Instance environment"
      kubespawner_override:
        cpu_guarantee: 3
        mem_guarantee: "14G"
        node_selector:
          size: xlarge
          pricing: spot
    - display_name: "Large On Demand environment"
      description: "Python in a large On Demand environment"
      kubespawner_override:
        cpu_guarantee: 1
        mem_guarantee: "6G"
        node_selector:
          size: large
          pricing: on-demand

nodes.yaml

# Use eksctl to create an EKS cluster
# docs here https://eksctl.io/usage/creating-and-managing-clusters/
# replace <cluster-name> in this file with the name of your cluster
apiVersion: eksctl.io/v1alpha5
kind: ClusterConfig

metadata:
  name: <cluster-name>
  region: eu-west-1

nodeGroups:
- name: main2
  labels: 
    role: main
  instanceType: t3.medium
  desiredCapacity: 2
  privateNetworking: true
  availabilityZones:
  - eu-west-1a
  iam:
    withAddonPolicies:
      autoScaler: true
- name: ml-large-od-v3
  privateNetworking: true
  availabilityZones:
  - eu-west-1a
  minSize: 0
  maxSize: 2
  labels:
    size: large
    pricing: on-demand
  instanceType: "m5.large"
  tags:
    "k8s.io/cluster-autoscaler/enabled": "true"
    "k8s.io/cluster-autoscaler/<cluster-name>": "owned"
    "k8s.io/cluster-autoscaler/node-template/label/size": "large"
    "k8s.io/cluster-autoscaler/node-template/label/pricing": "on-demand"
- name: ml-large-v3
  privateNetworking: true
  availabilityZones:
  - eu-west-1a
  minSize: 0
  maxSize: 2
  labels:
    size: large
    pricing: spot
  instancesDistribution: # Including this struct turns the pool into Spot Instances
    instanceTypes: ["m5.large", "m4.large"]
    onDemandBaseCapacity: 0
    onDemandPercentageAboveBaseCapacity: 0
    spotInstancePools: 1
  tags:
    "k8s.io/cluster-autoscaler/enabled": "true"
    "k8s.io/cluster-autoscaler/<cluster-name>": "owned"
    "k8s.io/cluster-autoscaler/node-template/label/size": "large"
    "k8s.io/cluster-autoscaler/node-template/label/pricing": "spot"
- name: ml-xlarge-v4
  availabilityZones:
  - eu-west-1a
  privateNetworking: true
  minSize: 0
  maxSize: 2
  labels:
    size: xlarge
    pricing: spot
  instancesDistribution:
    instanceTypes: ["m5.xlarge", "m4.xlarge"]
    onDemandBaseCapacity: 0
    onDemandPercentageAboveBaseCapacity: 0
    spotInstancePools: 1
  tags:
    "k8s.io/cluster-autoscaler/enabled": "true"
    "k8s.io/cluster-autoscaler/<cluster-name>": "owned"
    "k8s.io/cluster-autoscaler/node-template/label/size": "xlarge"
    "k8s.io/cluster-autoscaler/node-template/label/pricing": "spot"