Kubernetes HPA (Horizontal Pod Autoscaling) on KubeSphere

This document describes how to configure Horizontal Pod Autoscaling (HPA) on KubeSphere.

The Kubernetes HPA feature automatically adjusts the number of Pods to maintain average resource usage (CPU and memory) of Pods around preset values. For details about how HPA functions, see the official Kubernetes document.

This document uses HPA based on CPU usage as an example. Operations for HPA based on memory usage are similar.

Prerequisites

• You need to enable the Metrics Server.
• You need to create a workspace, a project and a user (for example, project-regular). project-regular must be invited to the project and assigned the operator role. For more information, see Create Workspaces, Projects, Users and Roles.

Create a Service

1. Log in to the KubeSphere web console as project-regular and go to your project.

2. Choose Services in Application Workloads on the left navigation bar and click Create on the right.

3. In the Create Service dialog box, click Stateless Service.

4. Set the Service name (for example, hpa) and click Next.

5. Click Add Container, set Image to mirrorgooglecontainers/hpa-example and click Use Default Ports.

6. Set the CPU request (for example, 0.15 cores) for each container, click √, and click Next.

   Note

   • To use HPA based on CPU usage, you must set the CPU request for each container, which is the minimum CPU resource reserved for each container (for details, see the official Kubernetes document). The HPA feature compares the average Pod CPU usage with a target percentage of the average Pod CPU request.
   • For HPA based on memory usage, you do not need to configure the memory request.

7. Click Next on the Storage Settings tab and click Create on the Advanced Settings tab.

Configure Kubernetes HPA

1. Select Deployments in Workloads on the left navigation bar and click the HPA Deployment (for example, hpa-v1) on the right.

2. Click More and select Edit Autoscaling from the drop-down menu.

3. In the Horizontal Pod Autoscaling dialog box, configure the HPA parameters and click OK.

   • Target CPU Usage (%): Target percentage of the average Pod CPU request.
   • Target Memory Usage (MiB): Target average Pod memory usage in MiB.
   • Minimum Replicas: Minimum number of Pods.
   • Maximum Replicas: Maximum number of Pods.

   In this example, Target CPU Usage (%) is set to 60, Minimum Replicas is set to 1, and Maximum Replicas is set to 10.

   Note

   Ensure that the cluster can provide sufficient resources for all Pods when the number of Pods reaches the maximum. Otherwise, the creation of some Pods will fail.

Verify HPA

This section uses a Deployment that sends requests to the HPA Service to verify that HPA automatically adjusts the number of Pods to meet the resource usage target.

Create a load generator Deployment

1. Select Workloads in Application Workloads on the left navigation bar and click Create on the right.

2. In the Create Deployment dialog box, set the Deployment name (for example, load-generator) and click Next.

3. Click Add Container and set Image to busybox.

4. Scroll down in the dialog box, select Start Command, and set Command to sh,-c and Parameters to while true; do wget -q -O- http://<Target Service>.<Target project>.svc.cluster.local; done (for example, while true; do wget -q -O- http://hpa.demo-project.svc.cluster.local; done).

5. Click √ and click Next.

6. Click Next on the Storage Settings tab and click Create on the Advanced Settings tab.

View the HPA Deployment status

1. After the load generator Deployment is created, go to Workloads in Application Workloads on the left navigation bar and click the HPA Deployment (for example, hpa-v1) on the right. The number of Pods displayed on the page automatically increases to meet the resource usage target.

2. Choose Workloads in Application Workloads on the left navigation bar, click on the right of the load generator Deployment (for example, load-generator-v1), and choose Delete from the drop-down list. After the load-generator Deployment is deleted, check the status of the HPA Deployment again. The number of Pods decreases to the minimum.

Edit HPA Configuration

You can repeat steps in Configure HPA to edit the HPA configuration.

Cancel HPA

1. Choose Workloads in Application Workloads on the left navigation bar and click the HPA Deployment (for example, hpa-v1) on the right.

2. Click on the right of Autoscaling and choose Cancel from the drop-down list.

Thanks for the feedback. If you have a specific question about how to use KubeSphere, ask it on Slack. Open an issue in the GitHub repo if you want to report a problem or suggest an improvement.