The two scripts doctl-create.sh and doctl-delete.sh are used to spin up
DigitalOcean VMs or "Droplets" based on Debian 10. The list of hosts to be
created/deleted by these shell scripts are in the file hosts. Ansible also uses this list as its inventory file, as mentioned in ansible.cfg
The create/delete scripts also adds/removes the VMs' IP addresses and hostname
to/from your /etc/hosts so you can SSH into VMs using just the name mentioned in the hosts file, e.g, ssh root@control-00 without fishing for Public IP addresses.
P.S If you want to follow along but don't have a DigitalOcean Account, consider signing up using this referral link. It gets you $100 worth of credits with them and also helps me out.
The kubeadm-install.yaml does all the configuration necessary for Kubernetes
to work. Including installing Kubernetes' packages, configuring boot parameters,
and kernel modules. The documentation to manually perform this can be found here. I use containerd as the container runtime.
The cleanup.yaml playbook deletes the private IPs on eth0 and eth1 just
because I wanted to simplify the underlying networking.
After both the playbooks have run across all your nodes and all the nodes have
been rebooted. Select one to work as your Control and run kubeadm init
inside it and follow the instructions printed in there to join other nodes to
the control plane. You can also copy /etc/kubernetes/admin.conf file to your
local machine at ~/.kube/config and use it to run kubectl from your end
without having to SSH into the control node.
After this you will need to install a Pod Networking solution.
P.S: Since I use Flannel for Pod Networking I run:
kubeadm init --pod-network-cidr 10.244.0.0/16 --ignore-preflight-errors=NumCPU,Mem
Followed by:
kubectl create -f https://raw.githubusercontent.com/coreos/flannel/master/Documentation/kube-flannel.yml
Because Kubeadm's default cidr doesn't work with Flannel and I use just 1vCPU and 1GB of memory for my test bench.
Once everything is done, run kubectl get nodes and kubectl get pods --all-namespaces to see if you cluster is healthy and up and running.