A very quick bodge done to get kubernetes going on vagrant with 3 nodes.
I've been messing too much with my own setup, so having a quick build is just nice to have.
The setup is basically identical to a certified kubeadm install for bare-metal infrastucture.
Full auto provisioning and deployment with
vagrant up
Comment out last section in Vagrantfile if you don't need full auto.
./vagrant-k8s.sh or ansible-playbook k8s.yaml
- intel osx 12.3.1 monterey (my host OS) - should easily run on your linux laptop
- vagrant 2.2.19
- virtualbox 6.1.32
- ansible 2.12.4
- ubuntu 20.4 LTS Focal (guest OS) - test with ubuntu 22.04 failed for k8s (reason unknown as of yet)
Kubernetes version here is 1.23.5-00. Reason being is to allow you to learn how to upgrade k8s with kubeadm.
Change in roles/k8s-install/defaults/main.yaml to what you need.
Add 127.0.0.1 kubernetes.default to /etc/hosts on your host if you wish to talk to the cluster from outside of control node.
Typically though this would have been achieved by another interface and EXTERNAL-IP config in your k8s
Typically in vagrant installs you will endup with worker's internal IPs being 10.0.2.15 This will cause inability to run kubectl exec. I've fixed this as below:
-
INTERNAL-IP for workers are based on
enp0s8network device as present on my version of virtualbox. Adjustinternal_ipv4variable inhosts.yamlif your setup is different. It should not be needed, really. -
Virtualbox on mac by default might not allow you to create required interfaces.
Try creating a following file/etc//etc/vbox/networks.conf* 10.0.0.0/8 192.168.0.0/16
You have a great chance to try them for yourself here. Otherwise try distros like microk8s.
6-9 minutes on linux i7-3930K and fairly slowish SSD and 50Mbit internet (and somehow faster on a i9 mac)
vagrant up --no-provision will be ignored and ansible will still execute (and should fail safely)