How does the Kubernetes HPA(HorizontalPodAutoscaler) determine which pod's metrics should be us-CodePudding

suppose we have below HPA(HorizontalPodAutoscaler) deployed in the demo namespace, and multiple pods (POD-A,POD-B) in this demo namespace have the same metric "istio_requests_per_second", How does the HPA determine the metric "istio_requests_per_second" from which pod should be used? Or every POD with this metric will be evaluate against the HPA target?

apiVersion: autoscaling/v2beta2
kind: HorizontalPodAutoscaler
metadata:
  name: httpbin
spec:
  minReplicas: 1
  maxReplicas: 5
  metrics:
  - type: Pods
    pods:
      metric:
        name: istio_requests_per_second
      target:
        type: AverageValue
        averageValue: "10"
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: httpbin

test...

CodePudding user response：

If you're using prometheus then its the adapter thats correlating between k8's pod name and what metric value to return. Basically the HPA is asking the prometheus adapter for metric istio_requests_per_second. By calling /apis/custom.metrics.k8s.io/v1beta1/namespaces/myNamespace/pods/mypod the adapter takes that and looks at its rules configured for what it should query for.

https://github.com/kubernetes-sigs/prometheus-adapter/blob/master/docs/config-walkthrough.md

CodePudding user response：

Based on my test, I think HPA uses the 'scaleTargetRef' to determine which POD's metrics should be used, and pull these metrics from the metrics server and evaluate them against the target config.