Deploy Kubeflow

In this section, you will learn how to deploy Kubeflow on your Kubernetes cluster. Kubeflow is an open-source machine learning (ML) toolkit for Kubernetes. It is designed to simplify the process of training, serving, and deploying ML models at scale.

Having a local instance of Kubeflow is useful for development and testing purposes. It allows you to experiment with different ML models and workflows without incurring the costs of cloud resources. It is an invaluable step for learning MLOps practices and understanding how Kubernetes works by examing a real-world use case.

What you’ll need

To complete this guide, you will need the following:

A running Kubernetes cluster. You can use the Kubernetes cluster you created in the previous section.
Longhorn installed on your Kubernetes cluster. You can install it by following the instructions in the previous section.
Kustomize installed on your local machine. You can install it by following the instructions in the official documentation.

Procedure

Follow the steps below to deploy Kubeflow on your Kubernetes cluster:

Clone the Kubeflow manifests repository:

user:~$ git clone https://github.com/kubeflow/manifests.git

Change to the manifests directory:
```
user:~$ cd manifests
```
Checkout a specific version of the Kubeflow manifests repository:
```
user:~/manifests$ git checkout tags/v1.8.1
```
Note

This documentation uses version 1.8.1 of the Kubeflow manifests. You can replace v1.8.0 with the version you want to use.
Deploy Kubeflow using the Kustomize:
```
user:~/manifests$ while ! kustomize build example | kubectl apply -f -; do echo "Retrying to apply resources"; sleep 10; done
```
Note

The while loop retries the kubectl apply command until it succeeds. The whole process can take up to 10 minutes.

Verify

Verify that Kubeflow is running.

Check the status of the pods in the kubeflow namespace:
```
user:~$ kubectl get pods -n kubeflow
```
Note

The output should show a list of pods running in the kubeflow namespace.

Port-forward the Istio Ingress Gateway service to access the Kubeflow dashboard:

user:~$ kubectl port-forward svc/istio-ingressgateway -n istio-system 8080:80

Open a web browser and navigate to http://localhost:8080. You should see the Kubeflow dashboard.

Note

The default username is user@example.com and the default password is 12341234.