Scale to zero refuses to work

[DaviPtrs]

Report

Can somebody tell me what the heck am I doing wrong? I just tried the same with KEDA 2.3.0 to make sure it isn't a 2.4.0 bug but still not working, so seems like this is my fault. In this example, I'm trying with a StatefulSet but I already tried with Deployment and nothing different happens.

This is my ScaleObject YAML

apiVersion: keda.sh/v1alpha1
kind: ScaledObject
metadata:
  name: arim
  namespace: enonic-review
spec:
  scaleTargetRef:
    apiVersion: apps/v1
    kind: StatefulSet
    name: arim
  pollingInterval: 10
  cooldownPeriod: 15
  minReplicaCount: 0
  maxReplicaCount: 1
  triggers:
    - type: prometheus
      metadata:
        serverAddress: http://prometheus.lens-metrics.svc.cluster.local
        metricName: container_network_receive_packets_total
        query: sum(rate(container_network_transmit_packets_total{namespace='enonic-review',pod_name=~"arim-.*"}[1m])) * 10
        threshold: '10'

Already tried with

  idleReplicaCount: 0
  minReplicaCount: 1
  maxReplicaCount: 1

and

  maxReplicaCount: 1

As you can see on HPA watch logs, it must be scaled to zero a few seconds after my HPA target goes to less than 10 (which is my Prometheus trigger threshold), but absolutely nothing happens.

keda-hpa-arim   StatefulSet/arim   7/10 (avg)           1         1         1          4m32s
keda-hpa-arim   StatefulSet/arim   50/10 (avg)          1         1         1          4m47s
keda-hpa-arim   StatefulSet/arim   103/10 (avg)         1         1         1          5m2s
keda-hpa-arim   StatefulSet/arim   115/10 (avg)         1         1         1          5m17s
keda-hpa-arim   StatefulSet/arim   79/10 (avg)          1         1         1          5m32s
keda-hpa-arim   StatefulSet/arim   26/10 (avg)          1         1         1          5m47s
keda-hpa-arim   StatefulSet/arim   6/10 (avg)           1         1         1          6m2s
keda-hpa-arim   StatefulSet/arim   5/10 (avg)           1         1         1          6m17s
keda-hpa-arim   StatefulSet/arim   6/10 (avg)           1         1         1          6m32s
keda-hpa-arim   StatefulSet/arim   5/10 (avg)           1         1         1          6m47s
keda-hpa-arim   StatefulSet/arim   6/10 (avg)           1         1         1          7m18s

Logs from KEDA operator

The operator logs keep repeating those following entries forever like nothing different happens

2021-08-19T22:37:48.168Z	INFO	controllers.ScaledObject	Reconciling ScaledObject	{"ScaledObject.Namespace": "enonic-review", "ScaledObject.Name": "arim"}
2021-08-19T22:37:48.168Z	DEBUG	controllers.ScaledObject	Parsed Group, Version, Kind, Resource	{"ScaledObject.Namespace": "enonic-review", "ScaledObject.Name": "arim", "GVK": "apps/v1.StatefulSet", "Resource": "statefulsets"}
2021-08-19T22:37:48.180Z	DEBUG	controllers.ScaledObject	ScaledObject is defined correctly and is ready for scaling	{"ScaledObject.Namespace": "enonic-review", "ScaledObject.Name": "arim"}
2021-08-19T22:37:48.191Z	DEBUG	controller	Successfully Reconciled	{"reconcilerGroup": "keda.sh", "reconcilerKind": "ScaledObject", "controller": "scaledobject", "name": "arim", "namespace": "enonic-review"}
2021-08-19T22:37:51.810Z	DEBUG	scalehandler	Scaler for scaledObject is active	{"Metrics Name": "prometheus-http---prometheus-lens-metrics-svc-cluster-local-container_network_receive_packets_total"}
2021-08-19T22:38:01.831Z	DEBUG	scalehandler	Scaler for scaledObject is active	{"Metrics Name": "prometheus-http---prometheus-lens-metrics-svc-cluster-local-container_network_receive_packets_total"}

KEDA Version

2.4.0

Kubernetes Version

1.21

Platform

Amazon Web Services

Scaler Details

Prometheus

[DaviPtrs]

[SOLVED] Turns out that KEDA only scales to zero when the target is exactly zero, I thought that it could scale if the target was above the threshold. I think it can be more clarified on documentation but that's ok, after I figured it out my ScaledObject worked like a charm.

[tomkerkhove]

Would you be open to submit a PR to improve the docs?