This set of demos and lab goes along with the coursera course: Network Principles in Practice: Linux Networking. This particular module was on Kubernetes and Kubernetes networking.

You are welcome to run this code and I try to make it as self explanatory as possible, but some of the explanation will be in the videos for the course.

NOTE: Mac M1/M2 users should refer to the guidance here

A Vagrantfile is provided that will create a Ubuntu 22.04 VM, and install the needed software on the VM.

This was tested using vagrant VirtualBox running on Windows 11.

vagrant up

Configure your ssh client with the following. I use MobaXterm.

Hostname/IP address: 127.0.0.1
Port number: 2222
Username: vagrant
Private Key: <path/to/private_key>
Note: you’ll want x11 forwarding on

To get the location of the private key:

vagrant ssh-config

When you want to stop the VM, you can either run vagrant suspend to save the state so you can resume it later with vagrant up, or vagrant halt to shut the VM down.

We use Kubernetes in Docker (KinD) to create a Kubernetes cluster. Provided are a few sample configuration files (under cluster-config directory), but kind is simple to use.

kind create cluster --config ./cluster-configs/1master2worker.yaml

Also provided are some sample Pod, Service, and Deployment configurations.

kubectl apply -f pod-configs/simple-nginx.yaml

The core of the material centers around creating a CNI network plugin. We do this in bash, and it applies what we covered in the rest of the modules. There are two main files - a configuration file, which gets put in /etc/cni/net.d, and an executable, which gets put in /opt/cni/bin. We start with a skeleton - which won't work, but is enough to see it getting called by looking at the log file. Then, we have a full example (nppnet that is very basic).

Below is a walkthrough of one possible way to install and run the network plugin.

First, create directories that will be mounted in the docker containers created with kind (representing the Kubernetes nodes).

run ./make_dirs.sh

Then, create the cluster with kind. This config will spin up 1 control plane node and 2 workers, and will mount a directory for each.

kind create cluster --config ./cluster-configs/1master2workerMount.yaml

Next, you need to edit nppnet-install.sh to set the IP addresses - instructions are in the file. Then run it. It's useful to look at the script to see what it is doing - you can even do each of the commands manually to really see how it's working.

./nppnet-install.sh

At this point, we have a network plugin installed, so you can run some pods.

kubectl apply -f pod-configs/forexec1-node1.yaml
kubectl apply -f pod-configs/forexec2-node1.yaml
kubectl apply -f pod-configs/forexec3-node2.yaml

Get the IP addresses of the pods

kubectl get pods -o wide

Ping each other pod. e.g., if forexec1 was 10.244.2.2 and you wanted to ping from forexec3, do this:

kubectl exec -it forexec3 -- ping 10.244.2.2

To get inside of worker 1:

docker exec -it kind-worker /bin/bash

Then you can do things like:

cat /var/log/nppnet.log   #(to see the output of the nppnet plugin
ls /opt/cni/bin/   #(you should see nppnet)
ls /etc/cni/net.d/ #(you should see 09-nppnet.conf)
ip netns
crictl ps

To delete a pod:

kubectl delete --force pod forexec1

(you should see the DEL command show up in the /var/log/nppnet/log - which we didn't do anything for, other than print a message)

To delete the cluster:

kind delete cluster

For all files in this repo, we follow the MIT license. See LICENSE file.