Kubeflow

This guide describes how to install Charmed Kubeflow (CKF) on NVIDIA DGX hardware. DGX systems are purpose-built hardware for enterprise AI use cases, featuring NVIDIA Tensor Core GPUs.

Requirements

• NVIDIA DGX-enabled hardware setup, including no NVIDIA drivers preinstalled, BIOS settings and bootloader.
• kubectl.

Install MicroK8s

Install MicroK8s and enable required add-ons as follows:

sudo snap install microk8s --classic --channel 1.22
 
sudo microk8s enable dns:10.229.32.21 storage ingress registry rbac helm3 metallb:10.64.140.43-10.64.140.49,192.168.0.105-192.168.0.111
 
sudo usermod -a -G microk8s ubuntu
sudo chown -f -R ubuntu ~/.kube
newgrp microk8s

Edit /var/snap/microk8s/current/args/containerd-template.toml by adding:

[plugins."io.containerd.grpc.v1.cri".registry.configs]

[plugins."io.containerd.grpc.v1.cri".registry.configs."registry-1.docker.io".auth]
username = "afrikha"
password = "<>"

Finally , restart MicroK8s:

microk8s.stop
microk8s.start

Enable GPU add-on

Install the required GPU operator as follows:

sudo microk8s.enable gpu
mkdir .kube
microk8s config > ~/.kube/config

Check the GPU count for MicroK8s:

kubectl get nodes --show-labels | grep gpu.count

Configure MIG

Configure MIG devices running the following command:

kubectl label nodes blanka nvidia.com/mig.config=all-1g.5gb --overwrite

Check again the GPU count to confirm it has increased:

kubectl get nodes --show-labels | grep gpu.count

Deploy CKF

Follow the instructions in General installation for this section.

Explore some examples

CKF can be run on different types of DGX hardware:

• See kubeflow-single-node-dgx for single-node examples.
• See kubeflow-multi-node-dgx for multi-node examples.

---