terraform-kustomization-os4ml

This Terraform module installs Open Space for Machine Learning, or Os4ML for short, on a k3d Kubernetes cluster using Argo CD.

Step 1: Create the k3d cluster

Install k3d by following the instructions on the k3d website, and create a local cluster using the provided configuration.

k3d cluster create --config ./k3d-default.yaml

Step 2: Install Argo CD using Terraform

Install Terraform and clone this repository.
Go to the examples/kubernetes folder and execute the following commands:

terraform init
terraform apply --auto-approve

Argo CD will now begin installing Os4ML. Relax and wait until the Os4ML namespace shows up, and the frontend service is running. If there are any issues, such as the argocd-server service not being created, try running the last command one more time.

Step 3: Connect to the Cluster

There are multiple ways to connect to your app once it's running in your cluster.
The two methods mentioned here are the ones we mainly use, so feel free to connect to your services however you feel comfortable.

Step 3.1 Port-Forward

Argocd

You can track the progress of your deployment easily and in a sophisticated way, using the ArgoCD Frontend.
To make it available on any free port on your local device, use the following command (in this example, it's port 8000):

kubectl port-forward -n argocd services/argocd-server 8000:80

Then access it at http://localhost:8000.
The initial username and password are admin, as specified in /manifests/argocd/base/argocd-secret.yaml.

Kubeflow-pipelines

When the Kubeflow-pipeline servers are up and running, forward the required port to an available port on your local device (in this example, it's port 8001):

kubectl port-forward -n kubeflow services/kserve-models-web-app 8001:80

Then access it at the forwarded port http://localhost:8001.

Os4ML

Lastly, if the application is readily deployed, you can access the Os4ML frontend in a similar way. Here is the command to forward the frontend to your local port 8002 as an example:

kubectl port-forward -n os4ml svc/frontend 8002:80

Then access it at http://localhost:8002.

Step 3.2 Telepresence

Alternatively, instead of forwarding each port manually, we use Telepresence. Here's how you do it with this tool.

First, follow the recommended steps on the Telepresence website to install it locally.
Then run:

telepresence helm install

After this, you are ready to connect to your cluster:

telepresence connect

Now you can access the application from your browser using the cluster-internal DNS names:

os4ml: http://frontend.os4ml.svc.cluster.local
argocd: http://argocd-server.argocd.svc.cluster.local
kubeflow-pipelines: http://ml-pipeline-ui.kubeflow.svc.cluster.local

Step 4: Start using Os4ML

For more information and examples on how to use the platform, please refer to the Documentation.