Unable to scale with AmbiguousSelector message when there are multiple HPA

[KasyapG]

Report

• We have two HPA (1. KEDA-Kafka scaler 2. Native HPA) which were working as expeccted and scaling applications based on metrics.
• we recently got upgraded to kubernetes 1.27.7 version and now see scaling is not happening and the HPA description shows ScalingActive as False and throws message "pods by selector rtf.mulesoft.com/id=d20e76b6-02af-48ea-b9d1-d62ad3b020e6 are controlled by multiple HPAs: [namespace/keda-hpa-att-perf-msgrtr-e namespace/att-perf-msgrtr-e".
• keda is on 2.11 and compatibility matrix shows this compatible with 1.27.7.
• Just to add we are having seperately scalers to avoid single point of failure for Keda not working as expected scenarios.

Expected Behavior

Scaling should work as expected when there HPA and Keda Kafka separately as well.

Actual Behavior

Not scaling and thew description shows below error.

Steps to Reproduce the Problem

1. Implement Keda and apply a Kafka trigger scaledobject.
2. Then apply a HPA with CPU and Memory trigger seperately.

Logs from KEDA operator

example

KEDA Version

2.11.0

Kubernetes Version

1.27

Platform

Microsoft Azure

Scaler Details

Kafka

Anything else?

No response

[JorTurFer]

Hello, I'd say that you upgrade k8s from 1.25 or below to 1.27. Am I right? The error and message that you are facing with isn't from KEDA but from Kubernetes directly:

• PR: https://github.com/kubernetes/kubernetes/pull/112011
• Commit: https://github.com/kubernetes/kubernetes/commit/caddfdd0407d32c68176b52d05f3b65a8dd79f04

They introduced a check to validate that a workload is managed by a single HPA based on pod labels/selectors. There isn't anything that we can do. The only thing that I can do is suggesting reviewing your pod labels to avoid conflicts with selectors