Automated Deployment Guide
Note: Helm installation option is still in preview.
This guide describes how to deploy Kubeflow on Amazon EKS using Cognito as your identity provider. Kubeflow uses Istio to manage internal traffic. In this guide, we will:
- create an Ingress to manage external traffic to the Kubernetes services
- create an Application Load Balancer (ALB) to provide public DNS
- enable TLS authentication for the Load Balancer
- create a custom domain to host Kubeflow (because the certificates needed for TLS are not supported for ALB’s public DNS names)
Prerequisites
This guide assumes you have Python 3.8 installed and that you have completed the prerequisites.
1.0 Create the required resources
The following steps automate:
- the creation of a custom domain and certificates as in section 1.0 of the Manual Deployment Guide to host Kubeflow and TLS certificates for the domain.
- the creation of a Cognito Userpool for user authentication.
- the configuration of an ingress and load balancer controller manifests.
Each section is detailed in Cognito Manual Deployment Guide.
-
Install the dependencies for the scripts.
pip install -r tests/e2e/requirements.txt
-
Update the following values in
tests/e2e/utils/cognito_bootstrap/config.yaml
:-
The registered root domain informations in
route53.rootDomain
.-
In the yaml example below,
route53.rootDomain.name
is set toexample.com
. -
If your domain is managed in route53, enter the Hosted zone ID found under
Hosted zone details
inroute53.rootDomain.hostedZoneId
. Skip this step if your domain is managed by other domain provider.
-
-
The name of the subdomain hosting Kubeflow (e.g.
platform.example.com
) inroute53.subDomain.name
. Read Create domain and certificates to understand why we use a subdomain. -
The cluster name and region where kubeflow will be deployed in
cluster.name
andcluster.region
(e.g.us-west-2
) respectively. -
The name of Cognito userpool in
cognitoUserpool.name
(e.g. kubeflow-users). -
The load balancer scheme (e.g.
internet-facing
orinternal
). Default is set tointernet-facing
. Useinternal
as the load balancer scheme if you want the load balancer to be accessible only within your VPC. See Load balancer scheme in the AWS documentation for more details.
The config file will look something like:
cognitoUserpool: name: kubeflow-users cluster: name: kube-eks-cluster region: us-west-2 kubeflow: alb: scheme: internet-facing route53: rootDomain: hostedZoneId: XXXX name: example.com subDomain: name: platform.example.com
-
-
Run the following script to create the resources.
cd tests/e2e PYTHONPATH=.. python utils/cognito_bootstrap/cognito_pre_deployment.py cd -
The script will update the config file with the resource names, ids, and ARNs it created. It will look something like:
cognitoUserpool: ARN: arn:aws:cognito-idp:us-west-2:123456789012:userpool/us-west-2_yasI9dbxF appClientId: 5jmk7ljl2a74jk3n0a0fvj3l31 domainAliasTarget: xxxxxxxxxx.cloudfront.net domain: auth.platform.example.com name: kubeflow-users kubeflow: alb: scheme: internet-facing serviceAccount: name: alb-ingress-controller namespace: kubeflow policyArn: arn:aws:iam::123456789012:policy/alb_ingress_controller_kube-eks-clusterxxx cluster: name: kube-eks-cluster region: us-west-2 route53: rootDomain: certARN: arn:aws:acm:us-east-1:123456789012:certificate/9d8c4bbc-3b02-4a48-8c7d-d91441c6e5af hostedZoneId: XXXXX name: example.com subDomain: us-west-2-certARN: arn:aws:acm:us-west-2:123456789012:certificate/d1d7b641c238-4bc7-f525-b7bf-373cc726 hostedZoneId: XXXXX name: platform.example.com us-east-1-certARN: arn:aws:acm:us-east-1:123456789012:certificate/373cc726-f525-4bc7-b7bf-d1d7b641c238
(Optional) Configure Culling for Notebooks
Enable culling for notebooks by following the instructions in configure culling for notebooks guide.
2.0 Install Kubeflow
Install Kubeflow using the following command:
make deploy-kubeflow INSTALLATION_OPTION=kustomize DEPLOYMENT_OPTION=cognito
make deploy-kubeflow INSTALLATION_OPTION=helm DEPLOYMENT_OPTION=cognito
3.0 Update the domain with ALB address
-
Check if ALB is provisioned. It takes about 3-5 minutes.
kubectl get ingress -n istio-system Warning: extensions/v1beta1 Ingress is deprecated in v1.14+, unavailable in v1.22+; use networking.k8s.io/v1 Ingress NAME CLASS HOSTS ADDRESS PORTS AGE istio-ingress <none> * k8s-istiosys-istioing-xxxxxx-110050202.us-west-2.elb.amazonaws.com 80 15d
If
ADDRESS
is empty after a few minutes, check the logs of thealb-ingress-controller
by following this troubleshooting guide. -
Substitute the ALB address under
kubeflow.alb.dns
intests/e2e/utils/cognito_bootstrap/config.yaml
. The kubeflow section of the config file will look like:kubeflow: alb: dns: ebde55ee-istiosystem-istio-2af2-1100502020.us-west-2.elb.amazonaws.com serviceAccount: name: alb-ingress-controller policyArn: arn:aws:iam::123456789012:policy/alb_ingress_controller_kube-eks-clusterxxx
-
Run the following script to update the subdomain with the ALB address:
cd tests/e2e PYTHONPATH=.. python utils/cognito_bootstrap/cognito_post_deployment.py cd -
4.0 Connect to Kubeflow central dashboard
Follow the instructions in the Connecting to Kubeflow central dashboard
section of the Manual Deployment Guide to:
- Create a user in Cognito user pool.
- Create a profile for the user from the user pool.
- Connect to the central dashboard.
5.0 Uninstall Kubeflow
Note: Delete all the resources you might have created in your profile namespaces before running these steps.
-
Run the following commands to delete the profiles:
kubectl delete profiles --all
-
Delete the kubeflow deployment:
make delete-kubeflow INSTALLATION_OPTION=kustomize DEPLOYMENT_OPTION=cognito
make delete-kubeflow INSTALLATION_OPTION=helm DEPLOYMENT_OPTION=cognito
-
To delete the rest of the resources (subdomain, certificates etc.), run the following commands from the root of your repository:
-
Ensure you have the configuration file
tests/e2e/utils/cognito_bootstrap/config.yaml
updated by thecognito_post_deployment.py
script. If you did not use the script, update the name, ARN, or ID of the resources that you created in a separate yaml file intests/e2e/utils/cognito_bootstrap/config.yaml
by referring to the following sample:# Sample config file cognitoUserpool: ARN: arn:aws:cognito-idp:us-west-2:123456789012:userpool/us-west-2_yasI9dbxF appClientId: 5jmk7ljl2a74jk3n0a0fvj3l31 domainAliasTarget: xxxxxxxxxx.cloudfront.net domain: auth.platform.example.com name: kubeflow-users kubeflow: alb: scheme: internet-facing serviceAccount: name: alb-ingress-controller namespace: kubeflow policyArn: arn:aws:iam::123456789012:policy/alb_ingress_controller_kube-eks-clusterxxx cluster: name: kube-eks-cluster region: us-west-2 route53: rootDomain: certARN: arn:aws:acm:us-east-1:123456789012:certificate/9d8c4bbc-3b02-4a48-8c7d-d91441c6e5af hostedZoneId: XXXXX name: example.com subDomain: us-west-2-certARN: arn:aws:acm:us-west-2:123456789012:certificate/d1d7b641c238-4bc7-f525-b7bf-373cc726 hostedZoneId: XXXXX name: platform.example.com us-east-1-certARN: arn:aws:acm:us-east-1:123456789012:certificate/373cc726-f525-4bc7-b7bf-d1d7b641c238
-
Run the following command to install the script dependencies and delete the resources:
cd tests/e2e pip install -r requirements.txt PYTHONPATH=.. python utils/cognito_bootstrap/cognito_resources_cleanup.py cd -
You can rerun the script in case some resources fail to delete.
-