✅ Get started with minikube & kubefow.

❌ We will not be discussing the basics tenets or architecture of Kubernetes.

ℹ️ Focus on the following topics

The Elevator Pitch

minikube is excellent for running Kubernetes locally. It's an open-source tool that lets you run a single-node Kubernetes cluster on your machine, making it ideal for local development and testing. It comes with “batteries included”, common add-ons that span different contexts, such as Networking (ingress & DNS), Storage (provisioning, CSI, etc.), NVIDIA GPU support, and numerous others.

And, kubeflow is an open-source machine-learning platform built on Kubernetes. It provides a comprehensive set of components, tools, and services necessary for developing, orchestrating, deploying, and running scalable and portable ML workloads in the form of pipelines. Specifically

• Central dashboard — kubeflow’s UI.
• Kubeflow Notebooks — a project for promoting automated machine learning (AutoML). It supports hyperparameter tuning, early stopping, etc. Since it is framework agnostic, it virtually supports all major ML frameworks and libraries.
• Kubeflow Pipelines (KFP)
• Katib —
• and Training Operator — an umbrella project for scalable distributed training of ML models using frameworks such as PyTorch and Tensorflow (many more supported).

We are interested in the "Kubeflow Pipelines" (KFP) component, which provides a UI for managing and tracking experiments, a scheduler, an SDK (in Python), etc. Additionally, one can easily integrate CI/CD pipelines with KFP pipelines for automating testing of the components and deploying versioned pipelines.

Prepare & install minikube

Most of the following shell commands will work both on Linux (primarily) and perhaps with some adjustments on macOS.

As per the official documentation, you can install minikube with:

mkdir -p /opt/minikube
cd /opt/minikube
curl -LO <https://storage.googleapis.com/minikube/releases/latest/minikube-linux-amd64> -o minikube
chmod +x minikube