Keda 2.14 Prometheus Scaler Divides by Number of replicas of pod

[doverk96]

Report

I am using Keda 2.14, externally hosted Prometheus for triggering the deployment running in EKS.
For every query provided, it always divides the value by the number of replicas.
Example: - count(container_cpu_cfs_throttled_periods_total{cluster="sample-prod", container="sample-web.*", namespace="prod"}) by (container)
This gives 4 in output, whereas if i run : - kubectl describe hpa sample-consumer -n qa, over there it displays value as
Metrics: ( current / target )
"s0-prometheus" (target average value): 1334m / 21
I have 3 pods running of sample-consumer in my EKS, and therefore I can see it is dividing 4 by 3 to get 1334m as value.

Expected Behavior

I was not expecting this inherent division by the number of replicas

Actual Behavior

It do inherent division by the number of replicas

Steps to Reproduce the Problem

1. Just use the query and see it divides the value by the number of replicas.

Logs from KEDA operator

example

KEDA Version

2.14.0

Kubernetes Version

1.28

Platform

Amazon Web Services

Scaler Details

Prometheus

Anything else?

No response

[JorTurFer]

Hello,
It's the expected behavior for AverageValue metrics. You can change the metric type to Value to not have the average:

To change the metric type, you just need to add metricType: Value in your trigger at the same level of type or metadata.

triggers:
- type: prometheus
  metricType: Value
  metadata:
    # Required fields:

You can get more information about how the HPA formula is calculated here: https://kubernetes.io/docs/tasks/run-application/horizontal-pod-autoscale/#algorithm-details

[doverk96]

Thanks