One of the most power features of orchestration tools such as Kubernetes is the ability to automcatically scale resource allocation in response to real-time changes in resource usage. In the context of continuous deployment, this provides a great deal of stability, with less need for human intervention. In this lesson, you will learn the basics of autoscaling in Kubernetes by creating a simple Horizontal Pod Autoscaler which will create and destroy pod replicas in response to CPU utilization.
Learning Objectives
Successfully complete this lab by achieving the following learning objectives:
- Install the Kubernetes metrics API in the cluster.
To accomplish this, do the following:
- Clone the Kubernetes metrics repo.
git clone https://github.com/kubernetes-incubator/metrics-server.git - Apply the standard configurations to install the Metrics API.
cd metrics-server/ git checkout ed0663b3b4ddbfab5afea166dfd68c677930d22e kubectl create -f deploy/1.8+/ - Wait a few seconds for the metrics server pods to start. You can see their status with
kubectl get pods -n kube-system.
- Clone the Kubernetes metrics repo.
- Configure a Horizontal Pod Autoscaler to autoscale the train schedule app.
Check the
example-solutionbranch of the source code repo for an example of the code changes needed in the train-schedule-kube.yml file: https://github.com/linuxacademy/cicd-pipeline-train-schedule-autoscaling/blob/example-solution/train-schedule-kube.yml.To complete this task, you will need to do the following:
- Create a fork of the source code at https://github.com/linuxacademy/cicd-pipeline-train-schedule-autoscaling.
- Add a CPU resource request in train-schedule-kube.yml for the pods created by the train-schedule deployment. Check the example solution if you need to need to know where to add this.
resources: requests: cpu: 200m- Define a HorizontalPodAutoscaler in train-schedule-kube.yml to autoscale in response to CPU load.
--- apiVersion: autoscaling/v2beta1 kind: HorizontalPodAutoscaler metadata: name: train-schedule namespace: default spec: scaleTargetRef: apiVersion: apps/v1 kind: Deployment name: train-schedule-deployment minReplicas: 1 maxReplicas: 4 metrics: - type: Resource resource: name: cpu targetAverageUtilization: 50- Generate some load on the app to see the autoscaler in action!
