HTTP add-on external scaler ignores scaledownPeriod setting

[ITAdmin2026]

Report

The HTTP add-on external scaler is not respecting the scaledownPeriod setting from HTTPScaledObject. Pods are being terminated after ~10 minutes instead of the configured 30 minutes (1800 seconds).

spec
  hosts:
  - abc.com
  pathPrefixes:
  - /api/prepress-validation
  replicas:
    max: 2
    min: 0
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: prepress-validation
    port: 4000
    service: prepress-validation
  scaledownPeriod: 1800
  scalingMetric:
    requestRate:
      granularity: 1s
      targetValue: 15
      window: 1m0s

Expected Behavior

The HTTP add-on external scaler should respect the scaledownPeriod: 1800 (30 minutes) setting from HTTPScaledObject. When no HTTP requests are active, the http scaler should scale down only after 30 minutes of inactivity, keeping the pod running for the full cooldown period.

Actual Behavior

The external scaler when no HTTP requests are active, KEDA http add on scales down the deployment to 0 and terminate pods after approximately 10 minutes instead of the configured 30 minutes.

Steps to Reproduce the Problem

1. Create HTTPScaledObject with scaledownPeriod: 1800 (30 minutes)
2. Send HTTP traffic to trigger scale-up to 1 replica
3. Stop all HTTP traffic to the service
4. Observe pod termination after ~10 minutes instead of 30 minutes
5. Check KEDA operator logs showing scale-down every 5 minutes

Logs from KEDA HTTP operator

2025-11-21T05:19:37Z INFO scaleexecutor Successfully updated ScaleTarget {"scaledobject.Name": "prepress-validation", "scaledObject.Namespace": "qa", "scaleTarget.Name": "prepress-validation", "Original Replicas Count": 0, "New Replicas Count": 1}

2025-11-21T05:29:51Z INFO scaleexecutor Successfully set ScaleTarget replicas count to ScaledObject minReplicaCount {"scaledobject.Name": "prepress-validation", "scaledObject.Namespace": "eng-dev", "scaleTarget.Name": "prepress-validation", "Original Replicas Count": 1, "New Replicas Count": 0}

HTTP Add-on Version

0.11.1

Kubernetes Version

1.32

Platform

Amazon Web Services

Anything else?

No response

[Fedosin]

Hi @ITAdmin2026,

Thank you for reporting this issue. We've investigated the behavior and found that the http-add-on is working correctly - the scaledownPeriod value from your HTTPScaledObject is being properly passed to the KEDA ScaledObject's cooldownPeriod field.

You can verify this by checking the ScaledObject that was created:

kubectl get scaledobject prepress-validation -n qa -o yaml | grep cooldownPeriod

You should see cooldownPeriod: 1800 in the output.

The issue is with KEDA, not the http-add-on

The logs you provided are from KEDA's scaleexecutor component, which is responsible for actually scaling the deployments:

2025-11-21T05:19:37Z INFO scaleexecutor Successfully updated ScaleTarget ...
2025-11-21T05:29:51Z INFO scaleexecutor Successfully set ScaleTarget replicas count to ScaledObject minReplicaCount ...

The http-add-on correctly creates the ScaledObject with cooldownPeriod: 1800 and reports metrics to KEDA via the external scaler. But it does not control when scaling happens - that's entirely KEDA's responsibility based on the ScaledObject configuration.

If KEDA is really not respecting the cooldownPeriod setting, please file an issue with the KEDA project: https://github.com/kedacore/keda/issues

[Fedosin]

Keda issue: kedacore/keda#7204