This repository contains the playbooks and other required Ansible content to automate OpenShift 4.x UPI-based installation on VMware vSphere 6.x platfrom.

Official UPI on VMware documentation can be found on https://docs.openshift.com/container-platform/latest/installing/installing_vsphere/installing-vsphere.html

To deploy OpenShift 4.x, one needs to configure DHCP, create required DNS records, and provision the required load balancers, an HTTP server and virtual machines before starting the actual installation. For details please refer to official documentation.

The automation playbooks first creates a network utility vm which is configured with the all required roles (i.e. DHCP, DNS, HTTP and LB) automatically. For DHCP and DNS roles, dnsmasq software is used. DHCP is only used during the first boot after which all OpenShift nodes are assigned the predefined IP addresses. DHCP server is also configured to serve only to OpenShift nodes using their mac addresses, so it's safe to have this DHCP server in the network. For load balancing role, HAProxy is used. If the base domain is example.com and cluster name is openshift, then the endpoints for api.openshift.example.com, api-int.openshift.example.com and *.apps.openshift.example.com are automatically configured on this HAProxy to load balance api requests to bootstrap and master nodes, ingress traffic to all worker nodes.

Initially, all the OpenShift nodes will use this network utility server as their DNS server and all the API and Ingress traffic will flow through it. After configuring the required DNS records on the enterprise DNS servers and provisioning the Load Balancers, one can easily eliminate and power off this network utility server by pointing all the nodes to enterprise DNS servers by updating the /etc/resolv.conf on OpenShift nodes.

Nearly all of the manual steps in the UPI method is automated by the playbooks in this repository. The only pre-requisites to start the automation is downloading the required Red Hat CoreOS OVA file and import it into VMware vSphere platform, determine the IP addresses that will be assigned statically to the OpenShift nodes and provide internet connectivity (either directly or through a web proxy) to the OpenShift nodes. There is no need to create any DNS records or configure any enterprise load balancers to initiate the installation.

For the smallest OpenShift cluster, 3 master, and 2 worker nodes are required. In addition to these, a bootstrap node is used to initiate deployment and a network utility server is used to provide basic networking services as described above. Therefore, 7 IP addresses must be allocated at minimum before starting the installation.he

To start the installation, one must set the VMware vCenter credentials as environment variables and then simply run the following command on an Ansible control node:

export VMWARE_PASSWORD='vCenter Password'
export VMWARE_USER='vCenter Username'
export VMWARE_HOST='vCenter IP/FQDN'

time ansible-playbook site.yaml -e @cluster_vars.yaml

Some defaults are defined in the default_vars.yaml file which is included by all the playbooks/plays. However, some of these defaults must be updated for individual clusters, like openshift cluster name or datacenter name, and it's better to consolidate all the updates in a new file and specify this file as Ansible extra variables to override the defaults (i.e. cluster_vars.yaml as shown in the example above).

Here you can find the list of variables used and their roles explained.