Overview
This is a Vagrant based project that demonstrates an advanced Openshift Origin Latest or Container Platform 3.3 install i.e. one using an Ansible playbook.
The documentation for the installation process can be found at
https://docs.openshift.org/latest/welcome/index.html
or
https://docs.openshift.com/container-platform/3.3/install_config/install/planning.html
Pre-requisites
- (If intending to install Openshift Container Platform then) a Red Hat Account is required so that the VM can be registered via subscription manager.
- Vagrant
- VirtualBox or Libvirt (--provider=libvirt)
Install the following vagrant plugins:
- landrush (1.1.2)
- vagrant-hostmanager (1.8.5)
- (If intending to install Openshift Container Platform then) vagrant-registration (found within the Red Hat CDK 2.2)
The Openshift Container Platform install requires importing a RHEL 7.2 box, the easiest way to do this is use the packet tool from hashicorp. The steps are described at
https://stomp.colorado.edu/blog/blog/2015/12/24/on-building-red-hat-enterprise-linux-vagrant-boxes/
The iso image that the vagrant image is created from should be the 'RHEL 7.2 Binary DVD' image on the Red Hat downloads site. The box name I have used in the Vagrantfile is 'rhel/7.2'
When installing Openshift Container Platform the Vagrantfile assumes a Red Hat Employee subscription 'Employee SKU'. If you aren't a Red Hat Employee then simply override with your pool id of the subscription that gives you access to the Openshift Container Platform rpms (this could be a 30 day trial subscription).
export RHSM_POOL=<pool id>
The OS for the origin install defaults to centos but can be overridden by the following environment variable
export ORIGIN_OS=<desired OS>
Installation
git clone https://github.com/openshift/openshift-ansible-contrib.git
cd openshift-ansible-contrib/vagrant
then for an Origin install
vagrant up
or for an Openshift Container Platform install
export DEPLOYMENT_TYPE=openshift-enterprise
export SUB_USERNAME=<Red Hat Username>
export SUB_PASSWORD=<Red Hat Password>
vagrant up (you will be prompted for your Red Hat account details and the sudo account password on the host during this process)
Two ansible playbooks will start on admin1 after it has booted. The first playbook bootstraps the pre-requisites for the Openshift install. The second playbook is the actual Openshift install. The inventory for the Openshift install is declared inline in the Vagrantfile.
The install comprises one master and two nodes. The NFS share gets created on admin1.
The inventory makes use of the 'openshift_ip' property to force the use of the eth1 network interface which is using the 192.168.50.x ip addresses of the vagrant private network.
Once complete AND after confirming that the docker-registry pod is up and running then
Logon to https://master1.example.com:8443 as admin/admin123, create a project test then
ssh to master1:
ssh master1
oc login -u=system:admin
oc annotate namespace test openshift.io/node-selector='region=primary' --overwrite
On the host machine (the following assumes RHEL/Centos, other OS may differ) first verify the contents of /etc/dnsmasq.d/vagrant-landrush gives
server=/example.com/127.0.0.1#10053
then update the dns entries thus
vagrant landrush set apps.example.com 192.168.50.20
In the web console create a PHP app and wait for it to complete the deployment. Navigate to the overview page for the test app and click on the link for the service i.e.
cakephp-example-test.apps.example.com
What has just been demonstrated? The new app is deployed into a project with a node selector which requires the region label to be 'primary', this means the app gets deployed to either node1 or node2. The landrush DNS wild card entry for apps.example.com points to master1 which is where the router is running, therefore being able to render the home page of the app means that the SDN of Openshift is working properly with Vagrant.
Notes
The landrush plugin creates a small DNS server to that the guest VMs can resolve each others hostnames and also the host can resolve the guest VMs hostnames. The landrush DNS server is listens on 127.0.0.1 on port 10053. It uses a dnsmasq process to redirect dns traffic to landrush. If this isn't working verify that:
cat /etc/dnsmasq.d/vagrant-landrush
gives
server=/example.com/127.0.0.1#10053
and that /etc/resolv.conf has an entry
# Added by landrush, a vagrant plugin
nameserver 127.0.0.1