Intern, Canspirit.ai (Mentor — Mr. Arunkumar Nair Canspirit AI Arun Nair) Student, School of ECE, MIT-WPU (Mentor — Dr.Vinaya Gohokar)

This is a tutorial on deploying Kubeflow on a local Kubernetes cluster from scratch. While trying to deploy it myself using the documentation on the official websites, I came across a lot of errors, mostly compatibility issues so I am documenting the process here, for reference.

We will create an Ubuntu VM first. I have created an instance on GCP.

VM Requirements:

1. VM OS — Ubuntu 20.04 LTS
2. 16 GB RAM
3. 4 vCPUs
4. The official site recommends machine with 250 GB storage, I have made a 150 GB one. Ensure that the root is allocated at least 50GB.

Prerequisites :

1. Docker — version 1.19
2. kubectl — version 1.15
3. minikube — version 1.15

I am using NoMachine as a GUI for VM instance. This is required since we need to access Kubeflow which cannot be accessed from host machine’s browser. Please feel free to use other methods, eg.PuTTY, to access the Kubeflow Dashboard.

• https://download.nomachine.com/download/6.9/Linux/nomachine_6.9.2_1_amd64.deb
• How to setup Linux GUI on google cloud with nomachine remote desktop | step by step guide — YouTube

#kubernetes #docker #kubeflow