In this hands-on we will work using k3s, it’s a lightweight Kubernetes distribution easy to install.
We will not work with vanilla kubernetes as its installation could be challenging.
Using k3s should not change your experience since we will stay at a beginer level and asics of this distribution work as a vanilla kubernetes.
Setup a VM
We will use Vagrant to spawn a VM a work with k3s.
Vagrant.configure("2") do |config|
config.vm.box = "ubuntu/bionic64"
config.vm.hostname = "kubemaster"
config.vm.network :private_network, ip: "10.0.0.11"
end
Start the VM and ssh into it
vagrant up
vagrant ssh
Install k3s
To install k3s from the install scripty (not recommended in production) :
curl -sfL https://get.k3s.io | sh -
Wait for the script to finish then watch nodes through kubernetes until the node is Ready :
sudo kubectl get nodes -w
Congratulation you have now a working kubernetes cluster ready to work on.
If you want to learn more about Kubernetes components read the doc.
Pods
In kubernetes world, containers running are named pods, there are more than just running container and I invite you to read the official documentation about it.
Let’s start by starting a Job with a pod from a busybox that just print something and sleep.
Create a file basic.yml with content :
apiVersion: batch/v1
kind: Job
metadata:
name: hello
spec:
template:
# This is the pod template
spec:
containers:
- name: hello
image: busybox
command: ['sh', '-c', 'echo "Hello, Kubernetes!" && sleep 3600']
restartPolicy: OnFailure
# The pod template ends here
And apply it by running
sudo kubectl apply -f basic.yml
Congratulation, you have now launched a Job in Kubernetes ! Let’s have a look at that.
Let’s start by looking at our pods by running
sudo kubectl get pod
You should seen a pod named hello-xxxxx, that’s your pod 👏.
We could look at its logs by running.
sudo kubectl logs <pod-name>
If I run it I got :
vagrant@kubemaster:~$ sudo kubectl logs hello-7kn7h
Hello, Kubernetes!
And that’s what we want.
Exercice
Create a job to run a busybox printing in the log a message configured by environment variable.
Search in the documentation of kubernetes what need to be modify based on basic.yml file.
Deployments
If pods are the basic unit in kubernetes, Deployment are the main unit for pushing deployment units. In the previous section to run a pod we start a Job, a Job is a one of pod execution, Deployments are a deployment unit to maintain a set of pod up & running.
Let’s create a simple deployment without a file :
sudo kubectl create deployment kubernetes-bootcamp --image=gcr.io/google-samples/kubernetes-bootcamp:v1
Let’s have a look at it :
sudo kubectl get deployments
We see our deployment 👍
Search associated pod name.
When you have it let’s se what we deployed. There is a kubectl port-forward that permit to redirect network form a port on the host to a pod (or a service) let’s have a look.
sudo kubectl port-forward <pod-name> 8080:8080 --address=0.0.0.0
This command redirect the traffic from your host on 8080 listening to 0.0.0.0 (as from everywhere) into the port 8080 of the pod configured.
Now we have this command running you can open your browser on your machine and go to http://10.0.0.11:8080 and you should see waht the application print.
Exercice
Create a deployment.yml file that create a deployment of image “gcr.io/google-samples/kubernetes-bootcamp:v1” like we just do.
Services
An abstract way to expose an application running on a set of Pods as a network service.
Take some time to read the official documentation
Next, let’s list the current Services from our cluster:
sudo kubectl get services
Let’s expose our previous deployment.
sudo kubectl expose deployment/kubernetes-bootcamp --type="NodePort" --port 8080
We just created a service of type NodePort (aka route all the traffic on a port from the host into deployment kubernetes-bootcamp).
We cand look at our service by running :
sudo kubectl get service
On my test run it look like
sudo kubectl get service
NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE
kubernetes ClusterIP 10.43.0.1 <none> 443/TCP 148m
kubernetes-bootcamp NodePort 10.43.246.196 <none> 8080:30325/TCP 2m21s
We know have a service that expose our deployment as a port, in my example : 30325. Open now a browser on your machine and opne http://10.0.0.11/30325 (change with your own port) and your should see the application.
Let’s look at how the service choose to which pod it send the traffic to.
sudo kubectl describe svc kubernetes-bootcamp
You should see a line Selector: app=kubernetes-bootcamp. It’s this line that define how a service choose the pods to send to network.
By doing a selection based on their labels. To be sure about this, look as the description of the kuberneets-bootcamp pod to see that the label app exist with the value kubernetes-bootcamp.
Exercice
Create a service.yml file that create the same service that we just created.
Scale
Let’s scale our deployment.
sudo kubectl scale deployments/kubernetes-bootcamp --replicas=4
If you look at the pods, now the kubernetes-bootcamp is deployed with 4 pods each running the application.
Let’s look at how the service manage this. let’s run a for loop and a curl to reuse the service (don’t forget to set your port):
for i in {1..10}; do curl http://localhost:30325/; done
We see that the service load balance between the pods the calls and adapted to our scale. This is possible du to the selection of the pods with labels.
Update
We can edit the deployment to change all the parameters. And we could also change the image to run.
Let’s try that and what happen by watching the pods :
sudo kubectl set image deployments/kubernetes-bootcamp kubernetes-bootcamp=jocatalin/kubernetes-bootcamp:v2
sudo kubectl get pods -w
We see that the pods are created then the old one are terminated.
If we run a curl http://localhost:30325/; we should see that the server respond a v=2 text.
Exercice
Define a complete.yml file that define a deployment with 3 replicas of gcr.io/google-samples/kubernetes-bootcamp:v1 and a associated NodePort service (in the same file).
Deploy it check that everything run
Update the file to set the replicas to 5 and the image to jocatalin/kubernetes-bootcamp:v2 and apply the file to see the modifications applied