The scaled workload does not function properly when the minimum replica count is set to 0

Report

I’m using the Kafka scaler (Confluent) to monitor Kafka lags and manage the scaling of my workloads based on the defined lag threshold. My configuration includes setting minReplicaCount to 0 to allow workloads to scale down to zero when there is no lag.

However, the expected behavior doesn’t occur—KEDA consistently keeps at least one replica running even when there are no new messages and the lag is zero.

Additionally, when minReplicaCount is set to 0, the HPA metric displays an unusual value of -9223372036854775808m (likely an int64 conversion issue). This happens despite the external metric value in GCP being correct.

Expected Behavior

KEDA should downscale the workload to zero replicas when there is no lag.
The HPA should display the correct metric value when minReplicaCount is set to 0.

Actual Behavior

KEDA maintains at least one replica even when there is no lag and minReplicaCount is set to 0.
The HPA displays an incorrect metric value, such as -9223372036854775808m, when minReplicaCount is set to 0. This happens despite the external metric from GCP reporting the correct value.

Steps to Reproduce the Problem

ScaledObject configuration: `apiVersion: keda.sh/v1alpha1 kind: ScaledObject metadata: name: sync-animations-commerciales-scaledobject namespace: api spec: scaleTargetRef: name: sync-animations-commerciales pollingInterval: 10
minReplicaCount: 0
triggers:
- type: kafka metadata: bootstrapServers: "xxxxxx" consumerGroup: "int-sync-animations-commerciales" lagThreshold: "100"
  excludePersistentLag: "true" scaleToZeroOnInvalidOffset: "true"
  sasl: "plaintext" tls: "enable"
  authenticationRef: name: "kafka-trigger-auth"`
HPA reports a negative value, and the replicas remain at one despite scaling conditions

Logs from KEDA operator

example

KEDA Version

2.16.0

Kubernetes Version

1.30

Platform

Google Cloud

Scaler Details

Apache kafka

Anything else?

No response

kedacore / keda