JupyterHub on Kubernetes

We will install and test JupyterHub in k8s as a service from VK Cloud Solutions

Tested with Kubernetes 1.20.4, Ubuntu 20.04, Helm chart for JupyterHub 1.1.3, JupyterHub 1.4.2

Running JupyterHub on Kubernetes useful links

Official doc

https://zero-to-jupyterhub.readthedocs.io/en/latest/index.html
https://zero-to-jupyterhub.readthedocs.io/en/latest/jupyterhub/installation.html

Prerequisites

(Optional) Create host VM

It would be easier to create a host VM in cloud
All work in the cloud can be done from that VM
You can install kubectl, Helm, Docker and all other things on this VM and don`t mess with your own local machine

How to create VM: https://mcs.mail.ru/help/ru_RU/create-vm/vm-quick-create
How to connect: https://mcs.mail.ru/help/ru_RU/vm-connect/vm-connect-nix
Steps:

1. Create VM
2. Connect to VM with SSH
3. Perform all steps described further in this instruction from this VM
4. Enjoy cloud:)

Create K8s cluster in mcs and download kubeconfig

Instruction: https://mcs.mail.ru/help/ru_RU/k8s-start/create-k8s
Kubernetes as a Service: https://mcs.mail.ru/app/services/containers/add/

Install kubectl

https://mcs.mail.ru/help/ru_RU/k8s-start/connect-k8s
https://kubernetes.io/ru/docs/tasks/tools/install-kubectl/

curl -LO https://storage.googleapis.com/kubernetes-release/release/`curl -s https://storage.googleapis.com/kubernetes-release/release/stable.txt`/bin/linux/amd64/kubectl
chmod +x ./kubectl
sudo mv ./kubectl /usr/local/bin/kubectl

Set path to kubeconfig for kubectl

export KUBECONFIG=/replace_with_path/to_your_kubeconfig.yaml

also it will be easier to work with kubectl while enabling autocomplete and using alias

alias k=kubectl
source <(kubectl completion bash)
complete -F __start_kubectl k

Install helm

https://helm.sh/docs/intro/install/

curl https://raw.githubusercontent.com/helm/helm/HEAD/scripts/get-helm-3 | bash

(Optional) Install Docker if you want to build your own images for JupyterHub and log into a Docker registry

https://docs.docker.com/engine/install/ubuntu/
https://docs.docker.com/engine/reference/commandline/login/
https://ropenscilabs.github.io/r-docker-tutorial/04-Dockerhub.html

JupyterHub installation and testing part

Install JupyterHub

helm repo add jupyterhub https://jupyterhub.github.io/helm-chart/
helm repo update

Also we need to mark one of storage classes as default for successful installation

kubectl get storageclass
kubectl patch storageclass csi-ceph-ssd-dp1-retain  -p '{"metadata": {"annotations":{"storageclass.kubernetes.io/is-default-class":"true"}}}'

Create config for basic installation

warning: this config for demo use only! NOT A PRODUCTION SOLUTION

nano config_basic.yaml
#paste this to config_basic.yaml

singleuser:
  defaultUrl: "/lab"
  storage:
    dynamic:
      storageClass: csi-ceph-ssd-dp1-retain 
hub:
  config:
    Authenticator:
      admin_users:
        - admin
      allowed_users:
        - your_another_non_admin_user
#DummyAuthenticator not for production
    DummyAuthenticator:
      password: insertyourpasswordhereMVeP2VXfr
    JupyterHub:
      authenticator_class: dummy

Apply basic config and install JupyterHub

helm upgrade --cleanup-on-fail \
  --install defaultinstall jupyterhub/jupyterhub \
  --namespace jupyterhub \
  --create-namespace \
  --version=1.1.3 \
  --values config_basic.yaml \
  --timeout 20m0s

To access JupyterHub we need to find external ip

kubectl get services -n jupyterhub

Look for LoadBalancer Service type. Then look for external ip.
You can access JupyterHub by entering this external ip to browser.

For debug and troubleshouting

kubectl get pods -n jupyterhub
kubectl get events -n jupyterhub
kubectl describe pod <pod-name> -n jupyterhub
kubectl logs <POD_NAME> -n jupyterhub

Create config for advanced installation

Let`s add some security measures

nano config_lb.yaml
#paste this to config_lb.yaml

singleuser:
  defaultUrl: "/lab"
  storage:
    dynamic:
#you could use different storage classes
#get storage classes with: kubectl get storageclasses.storage.k8s.io 
      storageClass: csi-ceph-ssd-dp1-retain 
hub:
  config:
    Authenticator:
      admin_users:
        - admin
      allowed_users:
        - your_another_non_admin_user
#DummyAuthenticator not for production
    DummyAuthenticator:
      password: insertyourpasswordhereMVeP2VXfr
    JupyterHub:
      authenticator_class: dummy
proxy:
  service:
#if you set loadBalancerSourceRanges, you can access JupyterHub only from ip address from this setting.
#you can set a bunch of IP adresses
#https://zero-to-jupyterhub.readthedocs.io/en/latest/administrator/security.html#restricting-load-balancer-access
    loadBalancerSourceRanges:
      - PLACE_YOUR_IP_HERE
      - PLACE_ANOTHER_YOUR_IP_HERE_OR_REMOVE_THIS_LINE
#EXAMPLE
#      - 91.74.148.161/32

Apply new config and upgrade JupyterHub

helm upgrade --cleanup-on-fail \
  defaultinstall jupyterhub/jupyterhub \
  --namespace jupyterhub \
  --version=1.1.3 \
  --values config_lb.yaml \
  --timeout 20m0s

You can check versions of Helm chart and JupyterHub here:
https://jupyterhub.github.io/helm-chart/

Also we can enable https and integrate JupyterHub with Github for authentication
Read more here:
https://zero-to-jupyterhub.readthedocs.io/en/latest/administrator/security.html#https
https://zero-to-jupyterhub.readthedocs.io/en/latest/administrator/authentication.html#github
https://docs.github.com/en/organizations/collaborating-with-groups-in-organizations/creating-a-new-organization-from-scratch

You can find working examples of config.yaml in config directory in this repo.
Read config_https_github.yaml

INTEL oneAPI demo

You could read more about oneAPI:
https://software.intel.com/content/www/us/en/develop/tools/oneapi/ai-analytics-toolkit.html
https://medium.com/intel-analytics-software/save-time-and-money-with-intel-extension-for-scikit-learn-33627425ae4

You need to install Docker if you want to build your own image or you could use our image: mcscloud/jupyter-ds-intel-mcs:v2

Login to Docker Hub

sudo docker login --username=YOUR_DOCKERHUB_USER_NAME
#sudo docker login --username=mcscloud

additional instruction about Docker Hub
https://jsta.github.io/r-docker-tutorial/04-Dockerhub.html

Create Dockerfile

#make separate dir
mkdir ~/intel_based_docker_image && cd ~/intel_based_docker_image

#then create Dockerfile
nano Dockerfile

#paste this to Dockerfile
FROM jupyter/datascience-notebook:hub-1.4.2
RUN pip install --no-cache-dir nbgitpuller
#Доп информация про nbgitpuller
#https://github.com/jupyterhub/nbgitpuller 
#Install git extension
RUN pip install --no-cache-dir jupyterlab-git
#Install Intel part
RUN conda install -c conda-forge scikit-learn-intelex

Build and push image

Let`s build custom image with Intel ML package

export YOUR_DOCKER_REPO=
#example export YOUR_DOCKER_REPO=mcscloud

sudo docker build -t jupyter-ds-intel-mcs .

sudo docker images
#find image id and copy it

sudo docker tag YOUR_IMAGE_ID  $YOUR_DOCKER_REPO/jupyter-ds-intel-mcs:v2
sudo docker push $YOUR_DOCKER_REPO/jupyter-ds-intel-mcs:v2

Create config for JupyterHub with Intel image

If you want to test Intel libraries for ML you need more resources
As you can see we add cpu and memory requirements to config under singleuser part

nano config_intel.yaml
#paste this to config_intel.yaml

singleuser:
  defaultUrl: "/lab"
  storage:
    dynamic:
#you could use different storage classes
#get storage classes with: kubectl get storageclasses.storage.k8s.io 
      storageClass: csi-ceph-ssd-dp1-retain
  cpu:
    limit: 3
    guarantee: 2
  memory:
    limit: 3G
    guarantee: 512M
  # Defines the default image
  image:
    name: jupyter/minimal-notebook
    tag: hub-1.4.2
  profileList:
    - display_name: "Minimal environment"
      description: "To avoid too much bells and whistles: Python."
      default: true
    - display_name: "Tensorflow"
      description: "If you want the additional bells and whistles: Python, R, and Julia."
      kubespawner_override:
        image: jupyter/tensorflow-notebook:hub-1.4.2
    - display_name: "Spark environment"
      description: "The Jupyter Stacks spark image!"
      kubespawner_override:
        image: jupyter/all-spark-notebook:hub-1.4.2
    - display_name: "JupyterLab with Intel libraries"
      description: "Use some Intel optimizations"
      kubespawner_override:
        image: PLACE_YOUR_DOCKER_REPO_OR_USE_mcscloud/jupyter-ds-intel-mcs:v2
#       image: mcscloud/jupyter-ds-intel-mcs:v2
hub:
  config:
    Authenticator:
      admin_users:
        - admin
      allowed_users:
        - your_another_non_admin_user
#DummyAuthenticator not for production
    DummyAuthenticator:
      password: insertyourpasswordhereMVeP2VXfr
    JupyterHub:
      authenticator_class: dummy
proxy:
  service:
#if you set loadBalancerSourceRanges, you can access JupyterHub only from ip address from this setting.
#you can set a bunch of IP adresses
#https://zero-to-jupyterhub.readthedocs.io/en/latest/administrator/security.html#restricting-load-balancer-access
    loadBalancerSourceRanges:
      - PLACE_YOUR_IP_HERE
      - PLACE_ANOTHER_YOUR_IP_HERE_OR_REMOVE_THIS_LINE
#EXAMPLE
#      - 91.74.148.161/32

Apply new config and upgrade JupyterHub

helm upgrade --cleanup-on-fail \
  defaultinstall jupyterhub/jupyterhub \
  --namespace jupyterhub \
  --version=1.1.3 \
  --values config_intel.yaml \
  --timeout 20m0s

For testing Intel library clone repo https://github.com/intel/scikit-learn-intelex with Git extension already installed in Jupyter
You can find Git extension on the left vertical bar
After cloning repo you can find test Notebooks in scikit-learn-intelex/examples/notebooks/

To use JupyterHub, enter the external IP for the proxy-public service in to a browser.

kubectl get service -n jupyterhub

Help

If you have any questions you can ask me here:
Telegram @volinski
Email [email protected]