I currently develop web applications on GKE.

I set up pods autoscaling to the environment.

When loading these pods, it scaled out as expected.(pods:1 -> 5) After that I stopped loading and scaled down as expected.(pods:5 -> 1) Then the page no longer appears.

Although it pings to the external IP of the load balancer, it does not respond to the request to the 80 port, so it seems not to be a routing problem of the load balancer.

I checked the load of the container, but the load was not particularly high.

After that, I manually scaled out pod to respond to requests.

Why does it stop responding to requests?

AutoScale setting file below.

apiVersion: autoscaling/v1
kind: HorizontalPodAutoscaler

metadata:
  name: autoscaler
  namespace: default
spec:
  scaleTargetRef:
    apiVersion: apps/v1beta1
    kind: Deployment
    name: frontend
  minReplicas: 1
  maxReplicas: 10
  targetCPUUtilizationPercentage: 50