Kubernetes in 5 mins

Clone https://github.com/betabandido/minikube-provisioner and follow the instructions there to setup a Kubernetes cluster using minikube.

You might want to use tmux to have multiple virtual terminals multiplexed into a single real one. Ctrl+B C creates a new window. Ctrl+B <number> switches to the specified window (using its index). Ctrl+B [ enters scroll mode, and q goes back to normal mode.

sudo minikube addons enable heapster
sudo minikube addons enable metrics-server

Run the following command to list all the different resources in the cluster:

kubectl get all --all-namespaces

cd ~/k8s-workshop/hello
sudo docker build -t hello:0.1 .

Deploy the application on Kubernetes:

kubectl apply -f k8s/deployment.yaml

List the deployments and pods:

kubectl get deployments
kubectl get pods

Display some more information:

kubectl describe deployment hello
kubectl describe pod hello

Show the application logs:

kubectl logs -l app=hello

# It is also possible to get the logs for a particular pod, given its name:
kubectl logs <pod-name>

Connecting to a pod:

kubectl exec <pod-name> -it -- sh

Update k8s/deployment.yaml so that the deployment uses two replicas:

# ...
spec:
  # ...
  replicas: 2
  # ...

And then apply the configuration again:

kubectl apply -f k8s/deployment.yaml

Listing the pods must now show two pods for application hello:

kubectl get pods -l app=hello

kubectl apply -f k8s/service.yaml

List the services in the cluster:

kubectl get services

Find some more information about the created service:

kubectl describe service hello

In a traditional Kubernetes cluster the external IP would be shown. In Minikube, however, the external IP must be obtained by running:

minikube service hello --url

Use curl to verify you can access the application:

curl $(minikube service hello --url)

Use the external address of the EC2 instance (not the one for the service) to access the application from your browser: http://ec2-<external-IP>.eu-west-1.compute.amazonaws.com:<port>

Get the dashboard's port running the following command:

minikube dashboard --url

Then use that port together with the external address of the EC2 instance to access the dashboard.

Explore the different resources in the cluster (pods, deployments, services, etc.).

Change main.py so that it returns another message, and then create a new docker image with a new tag:

sudo docker build -t hello:0.2 .

Now, update deployment.yaml so that it uses image hello:0.2. After that, update the deployment using:

kubectl apply -f k8s/deployment.yaml

Use your browser to check whether the message has changed.

Change main.py so that the healthz endpoint returns an error:

from flask import Flask, abort

# ...

@app.route('/healthz')
def healthz():
    abort(500)

Create a new docker image:

sudo docker build -t hello:0.3 .

Update the deployment.yaml file to use the new image, and update the deployment:

kubectl apply -f k8s/deployment.yaml

Watch pod status:

watch kubectl get pods

To rollback the deployment execute:

kubectl rollout undo deployment/hello

Create and deploy the cpu application (this application performs a CPU-intensive operation when it receives a request):

cd ~/k8s-workshop/cpu
sudo docker build -t cpu:0.1 .
kubectl apply -f k8s

Run the following command to send traffic to the application:

while true; do curl $(minikube service cpu --url) ; done

After some seconds you should see an increase in CPU utilization. Then, the auto-scale mechanism will trigger, creating new replicas. Run the following command to see the new replicas being created:

watch kubectl get pods

To create a full-fledged Kubernetes cluster have a look at kops.