Project that uses vagrant and ansible to create a local as also remote
kubernetes cluser with kubeadm in non High Availability mode, i.e. with a
single master that also incorporates the etcd database.
Provided that it works as expected, you can use it to set up locally a
non-single node (as provided by microk8s and minikube) kubernetes cluster;
vagrant up should give you a local kubernetes cluster with which you can
interact via kubectl
Make sure you override the if_name role variable with the interface name whose
IP you want to use for intra-node communication.
You can tweak the NODES variable in Vagrantfile to account for the number of
nodes you wish.
Once the provisioning is over, you should be able, after executing
kubectl proxy
on your host machine, to access the admin dashboard here
follow these instructions to create the necessary role/rolebinding and grab the token that will allow you admin access to the UI.
Creating a playbook and running against normal (i.e. non-vagrant) hosts should
work, provided that the inventory file used groups nodes in a nodes group as
also masters in a masters group. NOT TESTED YET
You should also tweak the resources used to spin up the vagrant VMs in the
Vagrantfile to match your host machine's capacity; as in ...
master.vm.provider "virtualbox" do |vb|
vb.cpus = "2"
vb.memory = "4096"
end
The role is not (yet at least) idempotent so re-runing it over an existing
cluster that you created using this functionality will most likely reset the
existing cluster (despite checks being performed on whether 6443 is in use
or a cluster is already running on the particular master)
- Test it against non vagrant hosts/vms etc.
- Add other networking options besides
flannel - Fail / Exit if cluster exists and / or
6443on master is in use - Add cluster health check at the end of the role
- [...] other