How to scale down to replicas=0

[sd2302]

How can I scale down my deployment to replicas=0 for node maintenance ?

My deployment is managed via Keda

I tried to manually edit the scaled object but

kubectl edit scaledobjects.keda.k8s.io cloud-ingestionuseractivitysvc-scaler

spec:
maxReplicaCount: 10
minReplicaCount: 3

both to 0 but then it immediately changes back (below is the full desc of keda scaled object)

Error: required flag(s) "replicas" not set
sysdev@imabvlsys-496:~/.kube$ kubectl scale ScaledObject cloud-ingestionuseractivitysvc-scaler --replicas=0
Error from server (NotFound): the server could not find the requested resource (patch scaledobjects cloud-inge

apiVersion: keda.k8s.io/v1alpha1
kind: ScaledObject
metadata:
creationTimestamp: "2020-04-16T22:21:47Z"
finalizers:

• finalizer.keda.k8s.io
  generation: 3
  labels:
  app: cloud-ingestionuseractivitysvc
  deploymentName: cloud-ingestionuseractivitysvc
  name: cloud-ingestionuseractivitysvc-scaler
  namespace: cloud
  resourceVersion: "25130625"
  selfLink: /apis/keda.k8s.io/v1alpha1/namespaces/cloud/scaledobjects/cloud-ingestionuseractivitysvc-scaler
  uid: 429390f5-e86c-4bdd-9acf-1ef6b001dbf0
  spec:
  maxReplicaCount: 10
  minReplicaCount: 3
  scaleTargetRef:
  deploymentName: cloud-ingestionuseractivitysvc
  triggers:
• metadata:
  connection: EventBusSettings__ConnectionString
  queueLength: "5"
  subscriptionName: UserActivityIngester
  topicName: eventbus
  type: azure-servicebus
  status:
  externalMetricNames:
• queueLength
  lastActiveTime: "2020-06-29T14:41:44Z"

[ahmelsayed]

I tried to manually edit the scaled object but

kubectl edit scaledobjects.keda.k8s.io cloud-ingestionuseractivitysvc-scaler

spec:
maxReplicaCount: 10
minReplicaCount: 3

both to 0 but then it immediately changes back (below is the full desc of keda scaled object)

This is a bit strange. KEDA itself doesn't update the spec of the crd, only the status. do you have gitops or something else thats enforcing a desired state for the cluster that might be reupdating the crd?

[clancytime]

As a separate option, we scale down the keda-operator pod before scaling down deployments during maintenance windows. Then scale up the operator when done and let it scale back up specific deploys as needed.

[isarns]

If someone finds it self with this issue use this annotation on your scaleobject:

  annotations:
    autoscaling.keda.sh/paused-replicas: "0"

https://keda.sh/docs/2.13/concepts/scaling-deployments/#pause-autoscaling