Does Kubernetes HPA scale on 'limits' or 'requests'

5/26/2021

I am trying to implement the HPA, but I am struggling to understand if the HPA checks the avg cpu utilization of the 'limits' or the 'requests'.

I've defined the following resources:

django:
  limits:
    cpu: 400m
    memory: 700Mi
  requests:
    cpu: 200m
    memory: 350Mi

on the HPA I configured the following CPU Utilization targetCPUUtilizationPercentage: 35

Does this mean 35% of the requests 200m (scale up on > 70m) or 35% of the limits 400m (scale up on > 140m)

-- Jeroen Beljaars
autoscaling
google-kubernetes-engine
hpa
kubernetes

0 Answers