/k8s

Local K8s Cluster

Credits

Simple set up ansible scripts based on a DO tutorial.

Deviation

While generally the same, I had updated for Ubuntu 20.04 (minimal changes), and use my own username. The largest changes were:

  • that I am using the latest version(s) of the tools vs pinned at v1.14
  • an updated flannel URL for later versions.
  • username change
  • dropped the use of ansible_user in the hosts file.

Use

  • deploy 4 Ubuntu VM's (or hosts)
  • All commands are on your local host unless otherwise specified.
  • clone this repo
  • create a hosts file (root of this repo) in the format of:
[masters]
master ansible_host=master1_ip

[workers]
worker1 ansible_host=worker1_ip
worker2 ansible_host=worker2_ip
worker3 ansible_host=worker3_ip

[all:vars]
ansible_python_interpreter=/usr/bin/python3
  • create your user and add ssh keys. ansible-playbook -i hosts init.yml
  • install Docker, k8s tools and deps. ansible-playbook -i hosts k8s-deps.yml
  • [OPTIONAL] You can edit --pod-network-cidr=10.244.0.0/16 in the master.yml if you prefer a different pod network.
  • Init the master/control-plane, setup Flannel, setup admin.conf (on master), etc. ansible-playbook -i hosts master.yml
  • ssh into the master (From Host: ssh user@master_ip) and check that it is ready (On master node: kubectl get nodes). Wait until you see the node is ready with something like:
$ kubectl get nodes
NAME   STATUS   ROLES                  AGE   VERSION
k8c    Ready    control-plane,master   56m   v1.20.2
  • If all good, you can exit the ssh session back to your local host and continue.
  • Add the worker nodes to the cluster. ansible-playbook -i hosts workers.yml
  • Verify the cluster - ssh into the master (From Host: ssh user@master_ip) and check that it is ready (On master node: kubectl get nodes). Wait until you see the nodes are ready with something like:
$ kubectl get nodes
NAME   STATUS   ROLES                  AGE   VERSION
k81    Ready    <none>                 17m   v1.20.2
k82    Ready    <none>                 17m   v1.20.2
k83    Ready    <none>                 17m   v1.20.2
k8c    Ready    control-plane,master   56m   v1.20.2
  • If all of your nodes have the value Ready for STATUS, it means that they’re part of the cluster and ready to run workloads.
  • You are now ready to deploy a workload via kubectl or similar tools.