HPA in Production: What the Docs Don't Tell You
In Kubernetes, managing workloads efficiently is vital for maintaining performance under varying loads. The Horizontal Pod Autoscaler (HPA) exists to automatically adjust the number of pods in a deployment or stateful set based on real-time demand. This means your application can scale up during peak usage and scale down when demand decreases, optimizing resource usage and cost.
The HPA works by monitoring the average CPU utilization across all pods in a deployment. By default, it targets an average CPU utilization of 50%. You can configure the minimum and maximum number of replicas, with defaults set to 1 and 10, respectively. When the load increases, the HPA controller increases the number of replicas; conversely, it scales down if the load decreases and the number of pods exceeds the minimum. To set this up, you need to enable the Metrics Server, which collects resource metrics from your cluster and exposes them via the Kubernetes API.
In production, be aware that it may take a few minutes for the number of replicas to stabilize after scaling actions. Also, if there are no clients sending requests, the current CPU consumption may show as 0%, which can be misleading. Ensure your Kubernetes server is version 1.23 or later to utilize HPA effectively. Running this on a cluster with at least two nodes is recommended to avoid control plane host issues.
Key takeaways
- →Configure HPA with a target average CPU utilization of 50% for optimal performance.
- →Set minimum and maximum replicas to control scaling behavior effectively.
- →Enable the Metrics Server to collect and expose resource metrics for HPA.
- →Monitor the stabilization time after scaling actions to manage expectations.
- →Ensure your Kubernetes version is 1.23 or later for HPA functionality.
Why it matters
In production, effective autoscaling can lead to significant cost savings and improved application performance. By dynamically adjusting resources, you can handle traffic spikes without over-provisioning.
Code examples
kubectl autoscale deployment php-apache --cpu=50% --min=1 --max=10kubectl get hpakubectl run -i --tty load-generator --rm --image=busybox:1.28 --restart=Never -- /bin/sh -c "while sleep 0.01; do wget -q -O- http://php-apache; done"When NOT to use this
The official docs don't call out specific anti-patterns here. Use your judgment based on your scale and requirements.
Want the complete reference?
Read official docsUnified observability — logs, uptime monitoring, and on-call in one place. Used by 50,000+ engineering teams to ship faster and sleep better.
Try Better Stack free →Mastering Zonal Shift Support for EKS Auto Mode and Karpenter
Zonal Shift is a game-changer for maintaining application availability during zone failures in EKS. This feature automatically manages node provisioning and traffic routing, minimizing downtime. Dive in to understand how it works and what you need to implement it effectively.
Mastering Serverless with the Headlamp Plugin for Knative
Unlock the full potential of your Knative workloads with the Headlamp plugin. This tool allows you to manage KServices, traffic splitting, and autoscaling configurations all in one place. Dive in to see how it enhances your serverless experience.
Unlocking EKS Auto Mode: Speed and Efficiency for Kubernetes Nodes
EKS Auto Mode is a game changer for Kubernetes scaling, delivering faster node management and smarter resource allocation. With Karpenter's 43% faster scale-out, you can optimize your cluster like never before.
Get the daily digest
One email. 5 articles. Every morning.
No spam. Unsubscribe anytime.