Prerequisites

You can install Kube AI Hub on virtual machines and bare metal with Kubernetes also provisioned. In addition, Kube AI Hub can also be deployed on cloud-hosted and on-premises Kubernetes clusters as long as your Kubernetes cluster meets the prerequisites below.

Kubernetes Version

• Minimum version: v1.28.x
• Recommended version: v1.34.x
• Supported versions: v1.28.x, v1.29.x, v1.30.x, v1.31.x, v1.32.x, v1.33.x, v1.34.x

To check your Kubernetes version, run:

kubectl version

The output should show:

Client Version: v1.34.x
Server Version: v1.34.x

Available Resources

• CPU: At least 2 cores
• Memory: At least 4 GB
• Disk: At least 40 GB

Default Storage Class

Your Kubernetes cluster must have a default Storage Class configured. To check, run:

kubectl get sc

The output should show a Storage Class with (default) annotation. If not, set one:

kubectl patch storageclass <your-storage-class> -p '{"metadata": {"annotations":{"storageclass.kubernetes.io/is-default-class":"true"}}}'

Network Requirements

• Ensure that the cluster nodes can communicate with each other.
• The cluster must have a CNI plugin installed (Calico, Flannel, etc.).
• If you plan to use the Kube AI Hub console, ensure that port 30880 is accessible.

Container Runtime

The cluster must use containerd as the container runtime. Docker is no longer supported.

To check your container runtime, run:

kubectl get nodes -o wide

The CONTAINER-RUNTIME column should show containerd://....

Architecture Support

Both x86_64 and ARM64 (aarch64) architectures are supported.

Helm Version

If you plan to install Kube AI Hub using Helm, ensure Helm 3.x is installed:

helm version