IllumiDesk Helm Chart

Overview

Use this helm chart to install IllumiDesk into your Cluster. This chart depends on the jupyterhub.

This setup pulls images defined in the illumidesk/values.yaml file from DockerHub. To push new versions of these images or to change the image's tag(s) (useful for testing), then follow the instructions in the build images section.

TL;DR

  helm repo add illumidesk https://illumidesk.github.io/helm-chart/
  helm repo update
  helm upgrade --install $RELEASE illumidesk/illumidesk --version $SEMVER --namespace $NAMESPACE --values example-config/values.yaml --debug

Prerequsites

• helm >= v3
• Kubectl >= 1.17

Testing Locally

1. Install Kind
2. Create a directory called illumidesk-nb-exchange
3. Inside that directory create a sub directory called illumidesk-courses
4. Create a cluster with the following YAML config

     kind: Cluster
     apiVersion: kind.x-k8s.io/v1alpha4
     nodes:
     - role: control-plane
       extraMounts:
       - hostPath: ./illumidesk-nb-exchange
         containerPath: "/illumidesk-nb-exchange"
       - hostPath: ./illumidesk-nb-exchange/illumidesk-courses
         containerPath: "/illumidesk-courses"

5. Run the following command to create the cluster
   • kind create cluster --name {cluster name} --config test-cluster.yaml
6. Load images as needed into local Kind cluster
   • kind load docker-image illumidesk/grader-notebook:712 --name {cluster name}
7. Create namespace in kubernetes
   • kubectl create namespace {}
8. If testing locally, fetch pull request and cd into root directory, helm-chart
9. To Deploy:
   • helm upgrade --install {name} charts/illumidesk/ -n {namespace} -f bar-us-west-2-config_go_grader.yaml --debug --dry-run
   • NOTE: name and namespace should match
   • NOTE: remove --dry-run flag to run deploy illumidesk stack

10. Port forward with

    • kubectl port-forward svc/proxy-public -n {namespace} 8000:80

    Troubleshooting

    • Get logs
      • kubectl logs {pod} -n {namespace}
    • Exec into pod
      • kubectl exec -it {pod} -n {namespace} -- /bin/bash
    • If helm chart is hanging, it is likely beacuse you need to load the docker image locally
      • kubectl get pods -n {namespace}
      • NOTE: if the hook-image-puller is in init status, log the pod and see if there are images that the cluster cannot pull

Installing the chart

Create config.yaml file and update it with your setup.

NOTE: to get a token use openssl rand -hex 32:

• Add Illumidesk repository to HELM:

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

• Install a release of the illumidesk helm chart:

  RELEASE=illumidesk
  NAMESPACE=illumidesk
  helm upgrade \
    --install $RELEASE \
    illumidesk/illumidesk \
    --version 3.2.0
    --namespace $NAMESPACE \
    --values my-custom-config.yaml

• Install a release wtihout config.yaml file

kubernetes create namespace test
helm upgrade --install test --set proxy.secretToken=XXXXXXXXXX illumidesk/illumidesk --version 3.2.0 -n test

Steps to setup argo on a new cluster

helm install argo argo/argo-workflows -n argo --create-namespace

helm install argo-events argo/argo-events --set singleNamespace=false --set namespace=''  -n argo-events --create-namespace

helm install argo-events-nats nats/nats --set cluster.enabled=true -n argo-events

helm install argo-events-stan nats/stan --set stan.nats.url=nats://argo-events-nats.argo-events.svc.cluster.local:4222 --set stan.clusterID=argo-events-stan --set cluster.enabled=true -n argo-events

Uninstall the Chart

    helm uninstall $RELEASE -n $NAMESPACE

Configuration

Note: Please follow instructions to install the Cert Manager if you are using the ALB or Nginx Ingress Controller

Note: Please follow instructions to setup external dns, if you plan to use this resource

Note: Please follow reference guides in the values.yaml in order to properly configure the resource during a deployment

NOTE: The following envars must be set:

• Jupyterhub.hub.extraEnvar.POSTGRES_NBGRADER_PASSWORD
• Jupyterhub.hub.extraEnvar.POSTGRES_JUPYTERHUB_PASSWORD
• Jupyterhub.hub.extraEnvar.JUPYTERHUB_API_TOKEN
• Jupyterhub.hub.extraEnvar.JUPYTERHUB_CRYPT_KEY

NOTE: The following envars must be set depending on autentication type

• Jupyterhub.hub.extraEnvar.LTI_SHARED_SECRET
• Jupyterhub.hub.extraEnvar.OIDC_CLIENT_SECRET

Cluster Helm Chart

EFS CSI Driver

1. Navigate to the `policy folder' and create a policy for EFS
2. Create a policy for EFS CSI driver using the policy document

       aws iam create-policy \
     --policy-name AmazonEKS_EFS_CSI_Driver_Policy \
     --policy-document file://policy/iam-policy-efs-csi-driver.json

3. Get the region-code and oidc-id to pass into trust policy

     aws eks describe-cluster --name {cluster} --query "cluster.identity.oidc.issuer" --output text

4. Use the example policy/trust-efs-csi-driver-policy-example.json to create the trust policy for efs csi driver

     aws iam create-role \
   --role-name AmazonEKS_EFS_CSI_DriverRole \
   --assume-role-policy-document file://"policy/trust-efs-csi-driver-policy-example.json"

   5. Attach efs csi driver IAM policy to the role created in the previous step

   aws iam attach-role-policy \
   --policy-arn arn:aws:iam::{account_id}:policy/AmazonEKS_EFS_CSI_Driver_Policy \
   --role-name AmazonEKS_EFS_CSI_DriverRole

6. Deploy Cluster level resources

   helm upgrade --install {release} illumidesk/cluster --namespace kube-system -f {cluster-stage-custom-config}.yaml --debug --dry-run

Configuration

Validate the Helm Chart

• For nodeport you will need to use your one of your node ips and also the port you defined in your values file.
  • Open up your browser and use the NODE_IP:NODE_PORT
  • Use the following command to list out your nodes:

   kubectl get nodes -o wide

• For load balancer you will need to get the external IP for proxy-public
  • Use this command to view your services and then paste the loadbalancer dns that is is the external ip of proxy-public

  kubectl get svc -n $NAMESPACE

• For Application Load Balancer, you must have specified the host in your values file
  • Verify the dns has propgates your domain

    dig $HOST 

• Open up your browser and paste the value for your host

Cleanup

  helm delete <release name> --purge