The Horizontal Pod Autoscaler is a great way to automatically scale Kubernetes applications in response to demands in real time. In this lab, you will have the opportunity to practice your skills with the Horizontal Pod Autoscaler by autoscaling an existing Deployment. This will give you some hands-on experience with autoscaling applications in Kubernetes.
Successfully complete this lab by achieving the following learning objectives:
- Configure Resource Requests for the Deployment
plantzoneDeployment is in the
defaultnamespace. Add a resource request to the Deployment’s Pod template so that the Horizontal Pod Autoscaler will have a baseline for app’s expected CPU usage. Each Pod is expected to use around
50mCPU under normal conditions.
You can find a Deployment manifest for the
- Create an HPA to Autoscale the Deployment
Create a Horizontal Pod Autoscaler to automatically scale the
plantzoneDeployment. Make sure there is always a minimum of
2replicas, with the ability to scale up to a maximum of
8replicas. The target percentage for CPU utilization should be
- Simulate a Burst of CPU Load and Observe the Autoscaling Behavior
You can create load by making a specialized HTTP request to any of the Deployment’s Pods. There is a NodePort service in place to make it easy to reach these Pods. You can communicate with the Service using port
30080on the lab server node,
This command will generate approximately
250mCPU load for
curl localhost:30080/ConsumeCPU -d "millicores=250&durationSec=120"
After generating the load, observe the behavior of the autoscaler. It should scale the Deployment up. Then, about 5 minutes after the CPU load goes way, it should scale back down.