Manual Scaling and HorizontalPodAutoscaler

Application Deployment / Guide 22 of 44 / Intermediate

• Kubernetes1.36.4
• Cluster4 nodes
• Runtimecontainerd 2.2.6
• CNICalico v3.32.1
• TimeAbout 18 min

--cpu-percent is deprecated and the captured output says so - the current flag is --cpu 50%. Note also that adding requests did not fix <unknown> immediately: it still read unknown 45 seconds later, because metrics-server scrapes on an interval and the Pods had just been recreated by the rollout.

This guide includes

Use this when replicas have to follow load rather than a number you typed. This matters because an HPA divides usage by the CPU request, so a Deployment that declares no requests gives it no denominator and it never scales anything at all.

• scaling by hand with kubectl scale, which always works
• creating an HPA with no denominator and watching it sit unable to decide anything
• adding requests with kubectl set resources so the percentage becomes computable
• putting real load on it and watching the replica count climb
• taking the load away, and seeing how long scale-down deliberately waits

Before you start

• A cluster with metrics-server. Without it kubectl top fails and an HPA can never read a value.
• The session creates namespace ckad-scale, a Deployment of registry.k8s.io/hpa-example, an HPA targeting 50% CPU, and a busybox Pod generating load in a loop.

1. 1

   Scaling by hand

2. 2

   An HPA with no denominator

3. 3

   Give it something to divide by

4. 4

   Put it under load

5. 5

   Taking the load away

Official sources

• Horizontal Pod Autoscaling
• HPA Walkthrough
• Resource metrics pipeline