kubernetes-gpu

Install NVIDIA GPU Support on CoreOS based Kubernetes Cluster

Prerequisits

• CoreOS based Kubernetes cluster with GPU nodes (e.g. AWS P2 instances)

Installation

First install the nvidia driver via this daemonset

kubectl apply -f https://raw.githubusercontent.com/afritzler/kubernetes-gpu/master/k8s-nvidia-driver.yaml

Wait until the init container finishes on each node and install the device plugin

kubectl apply -f https://raw.githubusercontent.com/afritzler/kubernetes-gpu/master/k8s-nvidia-deviceplugin.yaml

Run

To run an example training on a GPU node, start first a base image with Tensorflow with GPU support & Keras

kubectl apply -f https://raw.githubusercontent.com/afritzler/deeplearning-workbench/master/manifests/dl-workbench.yaml

Now exec into the container and start an example Keras traing

kubectl exec -it deeplearning-workbench-8676458f5d-p4d2v -- /bin/bash
cd /keras/example
python imdb_cnn.py

Open Issues

• Label GPU nodes and add NodeSelector to daemonset

Acknowledgments & References

• Build and install NVIDIA driver on CoreOS: https://github.com/squat/modulus
• Nvidia Device Plugin: https://github.com/kubernetes/kubernetes/blob/master/cluster/addons/device-plugins/nvidia-gpu/daemonset.yaml