superseb / multipass-k3s

Use multipass instances to create your k3s cluster

k3s cluster on multipass instances

This script will create a configurable amount of instances using multipass, install k3s server(s) (HA using embedded etcd), and add the agent instances to the cluster.

Requirements

• multipass (See multipass: Install Multipass)

This is tested on MacOS and Ubuntu Linux 22.04.

Running it

Clone this repo, and run the script:

bash multipass-k3s.sh

This will (defaults):

• Generate random name for your cluster (configurable using NAME)
• Create init-cloud-init file for server to install the first k3s server with embedded etcd (contains --cluster-init to activate embedded etcd)
• Create one instance for the first server with 2 CPU (SERVER_CPU_MACHINE), 10G disk (SERVER_DISK_MACHINE) and 1G of memory (SERVER_MEMORY_MACHINE) using Ubuntu focal (IMAGE)
• Create cloud-init file for server to install additional k3s servers with embedded etcd.
• Create one instance for additional server (configurable using SERVER_COUNT_MACHINE)
• Create cloud-init file for agent to join the cluster.
• Create one machine (configurable using AGENT_COUNT_MACHINE) with 1 CPU (AGENT_CPU_MACHINE), 3G disk (AGENT_DISK_MACHINE) and 512M of memory (AGENT_MEMORY_MACHINE) using Ubuntu focal (IMAGE)
• Wait for the nodes to be joined to the cluster
• Optionally merge the generated kubeconfig with the existing $KUBECONFIG (MERGE_KUBECONFIG)

Quickstart Ubuntu 22.04 droplet

sudo snap install multipass
wget https://raw.githubusercontent.com/superseb/multipass-k3s/master/multipass-k3s.sh
bash multipass-k3s.sh
curl -Lo /usr/local/bin/kubectl "https://dl.k8s.io/release/$(curl -L -s https://dl.k8s.io/release/stable.txt)/bin/linux/amd64/kubectl"
chmod +x /usr/local/bin/kubectl
kubectl --kubeconfig *-kubeconfig.yaml get nodes

Clean up

The files that are created are:

• $NAME-agent-cloud-init.yaml
• $NAME-cloud-init.yaml
• $NAME-kubeconfig.yaml
• $NAME-kubeconfig-orig.yaml
• $NAME-kubeconfig-backup.yaml (if MERGE_KUBECONFIG is set)
• $NAME-kubeconfig-merged.yaml (if MERGE_KUBECONFIG is set)

You can clean up the instances by running multipass delete k3s-server-$NAME-1 --purge and multipass delete k3s-agent-$NAME-{1,2,3} or (WARNING this deletes and purges all instances): multipass delete --all --purge

multipass-k3s's Issues

[Feature Request] Add an option to append the kubeconfig to default kubeconfig file

Thanks for creating such a great script, it works perfectly on my m1 mbp.
Is there possible to add an option to append the kubeconfig to the default kubeconfig file (e.g. ~/.kube/config on macOS)?

script fails

No server token given, generated server token: iizytxzu5etj4r3irtgn
No agent token given, generated agent token: akud5gjvscig6fu9f6cl
No name given, generated name: ext13m
Creating cluster ext13m with 1 server(s) and 1 agent(s)
Cloud-init is created at ext13m-init-cloud-init.yaml
Creating initial server instance: k3s-server-ext13m
Running multipass launch --cpus 2 --disk 5G --mem 1G focal --name k3s-server-ext13m --cloud-init ext13m-init-cloud-init.yaml
error: No such file: ext13m-init-cloud-init.yaml
There was an error launching the instance

Ubuntu 20.04 fresh install BM

HA setup?

Hello

I just found your script and was reading that it supports HA with embedded etcd.
So I checked the script but got confused how the config is working

# How many additional server instances to create
SERVER_COUNT_MACHINE="0"
# How many agent instances to create
AGENT_COUNT_MACHINE="1"

I would like to spin up 3 masters and 3 workers.

Do I understand correct that by default it only spin up 1 master?
So if I put SERVER_COUNT_MACHINE="2", it will generate a total of 3 masters?

And for the agent, I need to set it as AGENT_COUNT_MACHINE="3" ?