README.md

Scaling

Expected outcome

In this lab you will use both imperative and declarative methods to scale a ReplicaSet or Deployment. You will then learn to use the Horizontal Pod Autoscaler hpa to leverage heapster based metrics to scale a deployment.

How to

1. Deploy the kaur application

   Use evillgenius/kuar:1, create a deployment with the service exposed on port 80 and 2 replicas.

   kubectl create -f config\scaleapp.yaml

   Example

   Wait for the service to spin up and verify it is publicly accessible by going to the public IP in a browser.

2. Scale the deployment

   Scale the deployment to 4

   kubectl scale --replicas=4 --namespace=my-ns deployment/my-deploy

   While using the kubectl scale command is good for emergency type situations or one off scaling. In a declarative world, we make changes by editing the manifest file directly then applying the changes to the api:

   ...
   spec:
   replicas: 4
   ...

   Apply the original changes to scale the replica's back to 2

   kubectl apply -f config\scaleapp.yaml

3. Autoscale

   Set the Horizontal Pod Autoscaler to scale the deployment based on cpu. Use the kubectl autoscale deploy with a min, max and cpu-percent options to scale the service based on cpu usage above 80%. Set the minimum pods as 5 and max as 10.

   kubectl autoscale deployment my-deploy --namespace=my-ns --min=5 --max=10 --cpu-percent=80

   Get the status of the autoscaler.

   kubectl get hpa --namespace=my-ns

4. Clean Up

   kubectl delete -f config\scaleapp.yaml

Advanced areas to explore

1. Take a look scaling nodes in AKS.