This Vagrantfile create 4 machines, one with the roles "master" and "infra", other two with the "node" role and another one for storage/identity purposes:
| Machine | Address | Roles |
|---|---|---|
| okd.example.com | 172.27.11.10 | master, infra |
| node1.example.com | 172.27.11.20 | node |
| node2.example.com | 172.27.11.30 | node |
| extras.example.com | 172.27.11.40 | storage, ldap |
Everything is installed during the provisioning stage, this means that after the provisioning step, vagrant execute these two commands:
ansible-playbook /root/openshift-ansible/playbooks/prerequisites.yml
ansible-playbook /root/openshift-ansible/playbooks/deploy_cluster.yml
The ansible on master machine is preconfigured with the keys and the hosts to access the other machines without problems.
During provisioning step, the preconfigured inventory presented in files/hosts is copied to the master's /etc/ansible/hosts.
Some configurations on the inventory /etc/ansible/hosts was added to overcome problems with default NAT interface that vagrant creates:
- etcd_ip
- openshift_public_ip
- openshift_public_hostname
- openshift_logging_install_logging
- openshift_enable_olm
- openshift_enable_service_catalog
- ansible_service_broker_install
- template_service_broker_install
From a software point of view, you will need a least VirtualBox.
From a hardware point of view, each machine uses 2 cpu cores, unless the storage one. The master is configured to use 6GiB of RAM, the node ones 2GiB and the storage 256MiB, so you will need at least 11GiB of RAM, or even less if you lower the memory of each vm.
If you disable the openshift_metrics_install_metrics you can lower the memory from 6GiB to ~2GiB on master, and each node to 1GiB.
This takes a lot of time, just go to the cloned folder and type:
vagrant up
If you want to access the webconsole and/or see metrics you can add the hostnames and ip address on /etc/hosts:
echo '172.27.11.10 okd.example.com' | sudo tee -a /etc/hosts
On Windows you should modify
c:\Windows\System32\Drivers\etc\hosts
Remeber to access the address https://hawkular-metrics.172-27-11-10.nip.io and accept the self-signed certificate.
And then access the address https://okd.example.com:8443. The username and password are developer and 4linux.
If your machine have 8 GB or less of free memory still possible to provision OKD with one machine, disabling the metrics service. To do this clone the repository and do the following:
cd 4542/
mv Vagrantfile old-Vagrantfile
mv Allinone Vagrantfile
vagrant upIn this way only two virtual machines are created, the master with 4GB and all services on it and extras with NFS and LDAP.