Скачать или смотреть Horizontal Pod Autoscaling (HPA) using Rafay

In this video, I will describe why “horizontal pod autoscaling (HPA)” is important for applications operating on Kubernetes. I will also demonstrate how developers can quickly configure HPA using the Rafay platform.

Resource utilization for an application can vary throughout the day. Therefore, it is important for applications to automatically adjust capacity based on demand. It is common to provision capacity for a baseline and autoscale for peaks.

Let us look at how you can configure this in Rafay.

My application is called “podscaling”. It has a container image called “stress”. It is configured with arguments to continuously increase its resource demands. The image is sourced from Progrium’s repository on Docker Hub.

For pod autoscaling, we have configured a ceiling of “3 replicas” and want autoscaling to be triggered if CPU utilization is higher than 30%. Now, let us publish the workload on a cluster on Rafay’s Sandbox.

Now, we will use Rafay’s built in debug/diagnostics capabilities to view the status of the pods in my application.

Note that the information presented looks identical to output if the user was using the “kubectl” utility to interact with the Kubernetes cluster. Also, notice that pod events are also directly accessible.

Let us now use Rafay’s built in interactive debug capabilities to “exec” into the pod running on the Kubernetes cluster and monitor memory its usage. Also note that we have a real time view into the logs being emitted by our container.

In a few seconds, you will see autoscaling has been triggered and “additional pods” have been scheduled on the cluster.

Rafay also provides an “application dashboard” where the pod’s resource utilization is visually presented to the developer. As you can see, we are now running 3 replicas because the autoscaling thresholds were breached.