Best Practices deploying OpenShift 3.11 on IBM Cloud Pak System

IBM worked with a number of clients earlier this year to deploy Red Hat OpenShift Container Platform 3.11 on their IBM Cloud Pak System platform. We followed the process described on the IBM Developer tutorial “Accelerate your Red Hat OpenShift Container Platform deployment with IBM Cloud Pak System”.

In this blog post I would like to share some of the best practices we came across. I am planning to get the IBM Developer tutorial updated shortly, but in the mean time consider the points below.

1. Start by upgrading to IBM Cloud Pak System 2.3.1.1 or higher first
   If your IBM Cloud Pak System is on 2.3.0.1 and you are unable to upgrade immediately, please raise a PMR with IBM Support. You are likely to be able to proceed by downloading and importing the OpenShift Container Platform 1.0.2.1 pattern type and the IBM OS Image for Red Hat Linux Systems v 3.0.14.0 – RHEL v7.7 virtual image from IBM Fix Central.
2. Ensure that the OpenShift Container Platform 1.0.2.1 pattern type has been imported and enabled on your IBM Cloud Pak System
3. Ensure that you have the “IBM OS Image for Red Hat Linux Systems v 3.0.14.0 – RHEL v7.7” virtual image has been imported on your IBM Cloud Pak System.
   This is very important as the OpenShift Container Platform 1.0.2.1 pattern type will install OpenShift 3.11.146 (see point 6), which has minimum kernel version requirements met by the aforementioned virtual image.
4. When using the Virtual System Pattern “” as private docker registry, note that this uses a self-signed TLS certificate. When populating this with Docker images, make sure that you configure Docker on your client to use treat the registry as insecure as documented here.
   Note that you need to configure the FQDN of your private Docker registry in the docker /etc/docker/daemon.json and restart Docker.
5. When populating your private Docker registry with the Docker images for Red Hat OpenShift Container Platform 3.11, make sure to explicitly pull version 3.11.146. This version is what is expected when deploying using the OpenShift Container Platform 1.0.2.1 pattern type on IBM Cloud Pak System. So avoid simply pulling the latest version!
6. IBM Cloud Pak System 2.3.1.1 and the OpenShift Container Platform 1.0.2.1 pattern type introduced a new shared service called Cloud Pak System Software Registry Service. Refer to the IBM Cloud Pak System 2.3.1 Knowledge Center – Cloud Pak System Software Registry Service, you need to deploy this prior to deploying the Red Hat OpenShift Container Platform pattern.
7. Before you deploy the OpenShift Container Platform Virtual System Pattern, make sure that you do not use an IP group used for Data that has been associated with any part of the 172.30.0.0/16 IPv4 network. This network is reserved for use as the overlay network by OpenShift, as documented here and here.
8. The name of the domain that you specify for your OpenShift cluster must meet the following criteria – for example you cannot use capital letters or underscores. Refer to Configuring Cluster Variables of the OpenShift Container Platform 3.11 documentation.
9. Do not override the default value of 8443 for the OpenShift Master Console Port, currently that is not supported.
10. Taking all the above in mind, you should be able to deploy OpenShift 3.11 on IBM Cloud Pak System.

Note that the OpenShift Container Platform 1.0.2.1 pattern type has some limitations. Refer to the IBM Cloud Pak System 2.3.1 Knowledge Center – OpenShift Container Platform pattern, where the following is noted:

OpenShift Container Platform pattern is only supported on Single System and cloud groups that have a “Deployment type” of “Single Cloud”.

This implies the following:

• Multi-system deployments are not supported – you cannot deploy the VMs of a single pattern across multiple CPS racks.
• Deployments on multi cloud cloud groups are not supported – raise a PMR with IBM Support and ask for the OpenShift Container Platform 1.0.2.3 pattern type (which is also available from IBM FixCentral here).

A big thank you to Pallavi Singh, Mallanagouda Patil and Anshu Garg from IBM for their contributions here!