Prerequisites

If you wish to follow along, you will need:

• A Google Kubernetes Engine (GKE) cluster enabled with Workload Identity
• Custom Metrics Adapter (new resource model) installed into cluster as per Autoscaling Deployments with Cloud Monitoring Metrics_; _ensure you also enable it to work with Workload Identity
• The download of the exporter project

Problem

Imagine that you have a workload that experiences predictable large spiky loads. Specifically, for most of the day, the traffic to your workload is fairly even with only gradual changes. But then, every morning on the half-hour the traffic to your workload begins to experience a brief spike of traffic; say it triples for several minutes.

You already have in place a horizontal pod autoscaler (HPA) for your workload that has been working fine up until the introduction of the morning load spikes.

#scaling #kubernetes #gke