This repo helps set up the prerequisites for this blog post.

Request Access

Open a ticket with AWS support and request access to the p3.2xlarge instance. This will take a couple days to go through so do this first!

Use CloudFormation to Set up Infrastructure

Once you have access to the instance, upload CloudFormationTemplateOpenShift.yaml to an S3 bucket. From there, copy the URL and create a new CloudFormation Stack

Clone openshift-ansible

$ git clone https://github.com/openshift/openshift-ansible.git
$ cd openshift-ansible
$ git checkout origin/release-3.10

This gives us the playbooks to install OpenShift and its prerequisites.

Set up your inventory file

Go to your EC2 Dashboard and grab the public DNS names for your worker node and your master node and paste them into your inventory file.

Run the playbooks

1. prepare.yaml This playbook does some initial set up that is not included in the openshift-ansible prerequisites playbook. It's included in this repo.

   $ ansible-playbook ~/Workspace/openshift-origin-ansible/prepare.yaml -i ~/Workspace/openshift-gpu-prep/inventory --key-file ~/.ssh/id_rsa
   

2. prerequisites.yml We can run this from the openshift-ansible repo. It will take a while

   $ ansible-playbook -i ~/Documents/openshift-origin-ansible/inventory --key-file ~/.ssh/id_rsa \
     ~/Documents/openshift-ansible/playbooks/prerequisites.yml
   

3. deploy_cluster.yml Installs OpenShift!

   $ ansible-playbook -i ~/Documents/openshift-origin-ansible/inventory --key-file ~/.ssh/id_rsa \
     ~/Documents/openshift-ansible/playbooks/deploy_cluster.yml -e openshift_disable_check=package_version,disk_availability,memory_availability
   

Done!

From here you can view your dashboard at https://[master node DNS name]:8443. You now have an OpenShift cluster installed with a master node and a worker node with a GPU. Next steps are to follow the blog post to configure your cluster to consume the GPU resource!