Setup a K3S cluster with one master and two workers on 3 VMs
By running this configurator you will have an highly available Kube cluster on AWS Lightsail using three VMs and a load balancer. K3s is set up with the default configurations but you can fork this code and customize in case you need to.
For kube CNI, in this project we follow the k3s recommendation to use the default Traefik which exposes one port for our ingress service (nginx)
As stated, on that port there's Nginx with a reverse proxy to be easily able to access the deployed services
K3S uses https://github.com/appliedblockchain/provisioner#k3s / provisioner/infra to setup the AWS Lightsail VMs
git clone https://github.com/appliedblockchain/provisioner
git checkout k3s
cd infra
rake
rake CMD=provision
Run rake to prepare and deploy
rake will run and check if k3s is already provisioned, and will perform a deployment using the docker-compose.yml for pods definition
you can choose to use the raw verison rake KOMPOSE=y (or rake KOMPOSE=1) which instead of stack deploy will use kompose (which actually also builds and automatically pushes the containers)
The default command used internally is: docker stack deploy --orchestrator=kubernetes
As simple as that!
rake
Your deploy will start!
It is recommended to close port 80 on the workers as k3s can (dynamically) assign it as a port
Use dockerhub (using a private docker registry needs extra setup infos)
provisioner/infra will set up those for you, but if you're configuring it from scratch, you can open the following ports:
- Master: 6443 TCP
- Workers: 8472 UDP
- implement dnsimple auto-domain regisering
- implement close port
- close port 80
- write troubleshooting guide
- TODO: write difference between kompose and stack deploy
- TODO: write readme about what to check when things go wrong
This project uses Ruby (v1.9+) without needing any gem dependency
#kubernetes, #k3s, #ruby, #ops, #devops, #compose, #docker, #kompose
- @makevoid