Scaling Kubernetes Pods with External Metrics

Most workloads I’ve seen are IO bound and not CPU/memory bound.

Kubernetes’s horizontal pod autoscaler (HPA) can only scale based on CPU and memory via its built-in metrics.

However, using Prometheus/Grafana or Datadog, you can set your HPAs to scale your workloads off any external metric you want.

For example, you might want to increase your API service’s pods when the average response time is above a certain threshold.

Here are links to a blog post detailing how to set up scaling with Prometheus metrics and Datadog’s documentation on scaling with their metrics.

---