k3s Install Process step-by-step

Notes:

• This guide is spawned both because of a friend and this Rancher k3s GitHub issue.
• This guide is made from a fresh install of a Ubuntu 18.04 Server in VirtualBox.
• The Ubuntu 18.04 Server (hence forth just 'server') only has OpenSSH pre-installed during installation.
  • Docker was not pre-installed as it is a snap version - it does not work well with k3s - DO NOT PRE-INSTALL IT

Requirements before you get started

• Before you start them make sure your computer is able to go on the Internet (aka ping google.com -c1)
• If needed: Install Docker using the proper installation method and DO NOT use the snap version provided by default via apt/apt-get - it is not working properly with k3s (note only install Docker if you need it - it can be omitted)

Installing k3s

• Open a terminal and run the following command as your own user with sudo rights: curl -sfL https://get.k3s.io | sudo sh -
  • The installation will now procede and once finished you'll get your command prompt back.
• Test that k3s is operational using the command: sudo kubectl get nodes

Communicating/working with k3s

The command line tool kubectl is the way to communicate with k3s or any Kubernetes installation if you do not have any other 3rd party tools to do so.

k3s comes with its own built in kubectl command that is by default symlinked to /usr/local/bin/kubectl. This version of kubectl is built into k3s and uses the /etc/rancher/k3s/k3s.yaml file which is root owned. To use that version run sudo chmod o+r /etc/rancher/k3s/k3s.yaml and you can run kubectl <commands> without the need for sudo.

Alternative: use the official kubectl

You can install the official kubectl and work without compromising the security of the above k3s.yaml file.

To use the official Kubernetes kubectl version then either go to https://kubernetes.io/docs/tasks/tools/install-kubectl/ to see the installation procedure for your distribution. Or do the following

Once you've installed the proper kubectl for your distribution then run the following 3 command: You should now be above to work with k3s using kubectl the way it is designed. If you did the chmod above of the k3s.yaml file then re-run it with o-r instead of o+r to reset the permissions on the file to 600 or read/write only for root.

A quick offical kubectl install/upgrade procedure

If you are on a amd64bit Intel or AMD Debian/Ubuntu or RHEL/CentOS like distribution then this can be done for install and upgrade of the official kubectl command:

cd /tmp 
curl -LO https://storage.googleapis.com/kubernetes-release/release/`curl -s https://storage.googleapis.com/kubernetes-release/release/stable.txt`/bin/linux/amd64/kubectl
sudo rm /usr/local/bin/kubectl
chmod +x kubectl 
sudo mv kubectl /usr/local/bin/kubectl

Regardless of which official kubectl installation method you choose then do this once The office kubectl uses the ~/.kube/config file to get its configuration to Kubernetes or k3s', so make a copy of the k3s.yaml` file in your HOME dir like this:

mkdir ~/.kube
sudo cat /etc/rancher/k3s/k3s.yaml > ~/.kube/config
sudo rm /usr/local/bin/kubectl

Your first deployment

Remember the IP address from the Requirements section? Now you need the IP address If you don't then you can always get it via the following command

sudo kubectl get services --namespace kube-system traefik --output jsonpath='{.status.loadBalancer.ingress[0].ip}' 

The IP address is needed for you to setup one or more DNS entries (A records or CNAME's) pointing to it or local /etc/hosts entries eg

x.x.x.x     www.example.com nginx.example.com example.com

(replace x.x.x.x with the IP from the above command)

This will enable you to setup an Ingress object so you can have multiple services running on the same IP normal FQDN's like "nginx.example.com" and the like.

Setting up a Nginx service

In this repo there is a directory named nginx-deployment where there are 3 files in it.

The 3 files are

• nginx-deployment.yml
• nginx-service.yml
• nginx-ingress.yml

You need to update the nginx-ingress.yml file with the FQDN you setup on your DNS or in /etc/hosts unless you configured it for nginx.example.com.

Download the 3 files and apply them to k3s in a new namespace (namespaces are virtual "rooms" that separate applications from each other like directories do for files)

Create a new namespace: nginx

kubectl create namespace nginx

# Shortform - some parameters to kubectl have shortform eg ns for namespace, svc for service etc.
kubectl create ns nginx

Apply the 3 files to the nginx namespace

kubectl create ns nginx
kubectl -n nginx create -f nginx-deployment.yml
kubectl -n nginx create -f nginx-service.yml 
kubectl -n nginx create -f nginx-ingress.yml 

For each of the commands you should get one line back from kubectl ... which is

• namespace/nginx created when creating the namespace
• deployment.apps/nginx created when creating the Deployment
• service/nginx created when creating the Service
• ingress.extensions/nginx created when creating the Ingress object

After a little while you'll be able to see the setup using the following commands:

kubectl --namespace nginx get deployments
kubectl --namespace nginx get pods
kubectl --namespace nginx get ingresses
kubectl --namespace nginx get services

Each of these will show you one or more lines. Each line explains the current state of the Deployment, Pod(s), Service and Ingress controller.

Check if Nginx is working

curl http://nginx.example.com

replace nginx.example.com with whatever you setup

Congratulations

You have now setup Kubernets on your desktop/server and have deployed nginx (though some files are missing... we'll get to that in README-volumes.md)

Kubectl tips and tricks

Command completion for kubectl

If you use the BASH shell (most likely you do) then run the following command and logout and then login again

kubectl completion bash | tee -a ~/.profile
source ~/.profile

You can now use completion for most commands and pod lookups just like on the file system eg

# get a list of namespaces
kubectl -n <tab><tab>

# get a list of what you can get from the namespace (it is a lot)
kubectl -n nginx get <tab><tab>

# get logs for a particular pod (if there are multiple they will be shown so you can chose the next alphanumeric character of the pods name you want to get logs from)
kubectl -n nginx lo<tab> <tab><tab>

If you use an alternate shell then run kubectl completion -h to get help to setup completion.