complete

Kubernetes workshop with Fairwinds

Complete Workshop Walkthrough

Below are the steps that we'll be using to deploy a complete web app. Note that this demo will deploy a broken application and we'll be fixing the Kubernetes yaml documents to make the deployment successful.

Deploying Redis Database

First we'll be:

• Deploying a new Namespace for our application
• Deploying a redis server into the new namespace
• Reviewing the objects that we've deployed with Kubernetes (Deployments, Pods)
• Getting Logs from the deployed resources via kubectl

1. Create namespace in Kubernetes
   • kubectl apply -f namespace.yml Uses kubectl to apply the file (-f) with the Namespace definition. Note that the namespace we're deploying into is defined in the yaml documents in 01_redis/.
2. Deploy Redis Master and Replica
   • kubectl apply -f 01_redis/ Uses kubectl to apply all yaml definitions in the folder 01_redis/.
   • kubectl get deployments Gets the deployment resources from Kubernetes in the default namespace. You will not see the redis deployments since we have deployed them to a different namespace.
   • kubectl get pods Gets the pods in the default namespace. You won't see the redis pods since they've been created in another namespace.
   • kubectl get --namespace k8s-workshop deployments,pods Gets the deployments and pods with kubectl in the k8s-workshop namespace. You should now see that adding --namespace k8s-workshop will scope your get request to the k8s-workshop namespace.
   • kubectl describe --namespace k8s-workshop pod <pod name> Describes the current state of the <pod name> pod in Kubernetes. You can copy paste the name of the pod where you see <pod name>.
   • kubectl get --namespace k8s-workshop services This lists all the service definitions in the k8s-workshop namespace.
   • kubectl logs --namespace k8s-workshop <pod name> This prints the pods (containers) logs to your terminal. You can paste any pod name in the k8s-workshop namespace to replace <pod name>.

We should now see a healthy Redis Master and Replica in the k8s-workshop namespace in Kubernetes.

Deploying the Web App

Next we'll be:

• Deploying the web app into the default namespace (A different namespace from the one we created above)
• Watching the Cloud Load Balancer create external access to the web app
• CURLing the newly deployed web app to test manually
• Looking at the logs of the web app
• Fixing the broken web app deployment be deploying to the correct namespace
• CURL the web app again to test

1. Deploy the basic webapp
   • kubectl apply -f 02_webapp/ This deploys all the yaml definitions in the 02_webapp/ folder. Note that we are not defining the namespace in the metadata of the yaml, so it will default to the default namespace configured with kubectl.
   • kubectl get services Gets all the service definitions in the default namespace. We are going to wait until External IP has an IP Address.
2. Test the web app
   • curl [external_ip] This should show a hello world type response.
   • curl [external_ip]/asdf/1234 When you curl this website, you should experience an error connecting to Redis.
   • kubectl logs <webapp pod name> Let's look at the errors in the logs from the pod... Cannot connect to redis because it's default namespace DNS resolution won't work from a different namespace.
3. Fix the broken deployment
   • kubectl delete -f 02_webapp/ This will delete all the objects we created so that we can deploy correctly into the k8s-workshop namespace.
   • kubectl apply -f 02_webapp/ --namespace k8s-workshop This overrides the unset default namespace and deploys all the yaml files into the k8s-workshop namespace.
   • kubectl get services --namespace k8s-workshop We need to see the new service we've created, so the Load Balancer IP will have changed.
4. Test the new deployment
   • curl [external_ip]/asdf/1234 Now we see that the web app correctly saves the key and value to the database.
   • curl [external_ip]/asdf We can also query the value from the database.

Scaling the Application based on CPU

This section allows us to load test the web app and see the cluster responding by scaling the pod count up.

• Apply some load ./load.sh
• Open a new terminal or tab in your shell
• Look at the hpa occasionally to see the cpu usage go up kubectl get hpa -n k8s-workshop
• The number of pods should scale up to meet the new demand kubectl get deployment, po -n k8s-workshop

Optional Ingress (Separate Controller Required)

The 03_ingress sub-directory contains an optional Ingress definition. The kubernetes.io/ingress.class annotation may need to be changed to match that of your Ingress Controller. There is a second v1.18 yaml file for Kubernetes versions v1.18.0 and earlier, which do not support the networking.k8s.io/v1 APIVersion of Ingress.

Note that an Ingress Controller must be installed for Ingress resources to function. See the Ingress Nginx installation instructions for more information.