Flagger Canary not sync-ing KEDA ScaledObject from original to primary

[teochenglim]

Describe the bug

I am not sure this is not yet a feature or a bug.

While i try to create flagger canary with autoscalerRef.Kind=ScaledObject. The ScaledObject-primary successfully created. This is describe in the documentation https://docs.flagger.app/tutorials/keda-scaledobject

Then the problem, while i updating original ScaledObject settings doesn't get copy to the ScaledObject-primary.

I have tail the controller log and not action is taken while original ScaledObject changes.

To Reproduce

1. Follow https://docs.flagger.app/tutorials/keda-scaledobject guide to create resource.

apiVersion: flagger.app/v1beta1
kind: Canary
metadata:
  name: podinfo
  namespace: test
spec:
  provider: kubernetes
  # deployment reference
  targetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: podinfo
  # Scaler reference
  autoscalerRef:
    apiVersion: keda.sh/v1alpha1
    kind: ScaledObject
    # ScaledObject targeting the canary deployment
    name: podinfo-so
.
.

2. Change original scaleobject podinfo-so to some other values. cooldownPeriod from 20 to 15, minReplicaCount from 1 to 2.

apiVersion: keda.sh/v1alpha1
kind: ScaledObject
metadata:
  name: podinfo-so
  namespace: test
spec:
  scaleTargetRef:
    name: podinfo
  pollingInterval: 10
  cooldownPeriod: 20 -> 15
  minReplicaCount: 1 -> 2
  maxReplicaCount: 3
  triggers:
  - type: prometheus
    metadata:
      name: prom-trigger
      serverAddress: http://flagger-prometheus.flagger-system:9090
      metricName: http_requests_total
      query: sum(rate(http_requests_total{ app="podinfo" }[30s]))
      threshold: '5'

Expected behavior

The flagger controller shall monitor podinfo-so and update podinfo-so-primary. Why? This is simply more elegant and less moving parts to maintain when we thinking to use flagger.

Keda is a controller, Flagger is also a controller. So in the above cases, Flagger actually miss the "original deployment" replicas count, it is now control by keda

podinfo -> count = 2 (very unexpected), expected 0 podinfo-primary -> count = 1

In facts, all settings shall be deep copy including deployment/service/ingress/ScaledObject and then overwrite values if needed.

Additional context

• Flagger version: 1.27.0
• Kubernetes version: 1.22
• Service Mesh provider: na
• Ingress provider: nginx-ingress

[teochenglim]

seems like the pod count = 2 issue is solved in 1.28

[teochenglim]

found the annotation

$ kubectl get scaledobject/podinfo-so -o yaml  -n test
apiVersion: keda.sh/v1alpha1
kind: ScaledObject
metadata:
  annotations:
    autoscaling.keda.sh/paused-replicas: "0"
    
$ kubectl get -n test hpa/podinfo-so -o yaml
apiVersion: autoscaling/v1
kind: HorizontalPodAutoscaler
metadata:
  annotations:
    autoscaling.keda.sh/paused-replicas: "0"

[aryan9600]

The new changes in the primary ScaledObject will be reflected after you start a new canary analysis. This is documented here for HPAs, we should probably change the language to imply scalers in general.

[teochenglim]

does it work on Keda? I am not directly using HPA but Keda/HPA.