kube-fedora
kube-fedora is a project for my home lab that uses Vagrant to spin up a cluster of libvirt-backed VMs, then provision them using Ansible to build a kubernetes cluster. The end result is 1 master node and 2 minion nodes, a flannel overlay network, a single etcd server on the master, and TLS encrypted communication between the services.
Credit goes to jameswmills for the SSL-related configuration and ansible plays.
Requirements
- CentOS or Fedora bare metal host
- vagrant
- vagrant-libvirt plugin
- ansible
Configuration
There are not many variables to set up with this project. In it's current form, there are no variables defined outside of the Vagrantfile for vagrant itself. The 3 VMs that are built have hardcoded hostnames and MAC addresses, the latter of which is used to create a second vNIC in the guest for public access. In my environment I have DNS set up to map these MAC addresses to specific IPs, however the ansible playbook also updates /etc/hosts in each VM with FQDN mapping.
On the ansible side there is currently a group_vars/all.yml file with a few variables in it. The only one that might be worth updating is the cluster IP range.
Running kube-fedora
In order to make sure ansible only runs once all nodes are online, it is necessary to use the --no-parallel flag with vagrant. Vagrant can be run like this, and will automatically start the ansible playbook:
[root@fedorahost kube-fedora]$ vagrant up --no-parallel