Update: Due to Network restrictions we are switching to use popcorn for the most part. Thus, this document is going to be archived as for the most part we aren't going to use Kubernetes and are instead creating a https://en.wikipedia.org/wiki/Single_system_image
What we need:
- A laptop (serving as the master node), preferably running OSX or Linux
- Multiple single board computers (we are using Raspberry Pi's running Raspbian Linux since it is optimised for the pis themselves)
- HDMI Cables
- Power cords
Connect to nodes over SSH or through plugging it into a monitor (Repeat the following for each Pi)
ssh pi@ipaddress
sudo systemctl --now enable sshd # start and enable SSH on reboot
Check whether you are on Raspbian v10 . lsb_release -a
For each pi, upgrade it and install vim
sudo apt update && sudo apt upgrade
sudo apt install vim
The images on the pis are unfortunately 32bit. Just edit /boot/config.txt
sudo vim /boot/config.txt
And add this line to it
arm_64bit=1
Then run sudo systemctl reboot.
Once you are sshed back into the pi run uname -a. If it outputs aarch64 then you are running the 64 bit kernel.
Make sure the hostname of the pi is "node" followed by a number of your choice. For simplicity, we have node1, node2, node3 and node0 (master).
Edit /boot/cmdline.txt in your editor of choice (mine being vim):
sudo vim /boot/cmdline/txt
Add cgroup_enable=cpuset cgroup_memory=1 cgroup_enable=memory to the END of the line. Do NOT create a new line!
At this stage you must restart the pis for the changes to take affect. sudo reboot
On the laptop
curl -sfL https://get.k3s.io | sh -
sudo k3s kubectl get node
The second command should show that there is one "node", which is "node0". Kubernetes has conveniently given us a token. We must use this to connect our slave nodes to the master node.
sudo cat /var/lib/rancher/k3s/server/node-token
For each slave node (adjust the environmental variables as needed):
curl -sfL https://get.k3s.io | K3S_URL=https://node1:6443 K3S_TOKEN=token_from_earlier sh -
Now hold up for a couple of minutes for it to register the slave node.
Hop back over to the laptop:
sudo k3s kubectl get node
If the registration is crapping out on you run sudo service k3s-agent restart - nothing the old restart trick can't fix.
This should show that the slave nodes are registered to the master node:
Now what we want is to have like a GUI to click buttons and have stuff done for us in order to make it easier for everyone:
On the laptop run the following
sudo kubectl apply -n portainer -f https://raw.githubusercontent.com/Krish-sysadmin/k8s/master/deploy/manifests/portainer/portainer.yaml
Now the nice web GUI will magically appear in a couple of seconds:
Go to http://node1:20777
Click on Kubernetes (the "container orchestration system", in simple words what allows us to well make the cluster work) and click save:
We'll see the cluster on the home page (woohoo) with "4 nodes" writen besides it.
- Click on the cluster and "Applications + Add". We'll spin up multiple (data won't be shared across nodes/containers so we'll probs have to setup a central volume) Global instance of whatever app we want to test whether redundancy works (like a website for instance or online VSCode thing - we can decide with Marcus then). Select reasonabe CPU/memory amounts and container port. Click start.
Hope it is working so far. Now with this the possibilities for us are endless. YOU pick what you want to do!