shiftstack-ci

This repository contains tools to help build and simplify a testing environment for the development of OpenShift 4.x installation against an OpenStack cloud.

It is complementary to ocp-doit that focus on deploying OCP using a standalone TripleO.

Prerequisites

Access to an OpenStack cloud fitting the OpenShift on OpenStack requirements.

The RHCOS image can be downloaded from the release browser.

It may be possible that the flavors available from your OpenStack cloud provider set a disk size smaller than what is required for the rhcos image. If that's the case, you can shrink the image with qemu-img, for instance:

qemu-img resize rhcos-maipo-400.7.20190312.0-openstack.qcow2 --shrink 10G

Additionally the cloud must have enough capacity for:

7 m1.medium nodes
1 floating IP

Finally, you'll need a go dev environment for building the installer. This guide has gifs of cats, and is highly recommended if you are struggling with this.

Set up

Before running any of the scripts in this repository, there is minimal setup required.

You will need to clone the openshift installer. Rather than clone using git, it is much more convenient to use go get, since it will put it where it belongs for you.

go get github.com/openshift/installer

Make sure to export the OS_CLOUD environment variable to the name of your OpenStack cloud provider from your $HOME/.config/openstack/clouds.yaml file.

Lastly, you need some binaries to work with OpenShift and OpenStack. This is easy, just run:

sudo dnf install jq python2-openstackclient origin-clients

Cluster Configuration

Make a copy of the cluster_config.sh.example file:

cp cluster_config.sh.example cluster_config.sh

Adjust the settings to match your environment. This will set up how and where your cluster gets built, so it is important to fill it out carefully. Here is a rundown of the important fields you will likely have to modify:

OS_CLOUD           The cloud in your openstack cluster that resources will be consumed from.
CLUSTER_NAME       What your ocp cluster will be nicknamed. This naming scheme is propogated to all resources in the cluster.

Finally you need to obtain a pull secret from here, and replace PULL_SECRET vaiable with the new value. If you don't do this you can still install the cluster, but you won't be able to create new applications in OpenShift because you won't have access to private images.

Once this has been set up, you can proceed with building the installer.

Building the Installer

Just run the convenience script: build_ocp.sh! You will have to do this before your first run, and every time you make a change to the installer.

Deploying OpenShift

This part is also pretty self-explanatory, the run_ocp.sh script will create a cluster, and the destroy_cluster.sh script will destroy it. Moving forwards, however, we will be looking to move away from building the cluster this way, and towards using the CI operator as our primary means of testing.