Are the scaling from 0 ASG tags optional?

[ykostsik]

Which component are you using?:

/area cluster-autoscaler

What version of the component are you using?:

Component version: 1.34.0

What did you expect to happen?:

When scheduling a pod, it is in a permanent pending state since there is no k8s.io/cluster-autoscaler/node-template/label/myNodeSelectorLabel tag for ASG

What happened instead?:

Autoscaler successfully scaled up and down a node from the unlabeled ASG

How to reproduce it (as minimally and precisely as possible):

It's rather a question. While configuring the HowTo docs I accidentally skipped the tags that are marked as important for CA.

#ASG
aws autoscaling describe-auto-scaling-groups   --auto-scaling-group-names eks-web-asg   --query 'AutoScalingGroups[0].Tags[].{Key:Key,Value:Value,PropagateAtLaunch:PropagateAtLaunch}' --output table

|                   Key                   | PropagateAtLaunch  |             Value              |
+-----------------------------------------+--------------------+--------------------------------+
|  eks:cluster-name                       |  True              |  eks-dev                       |
|  eks:nodegroup-name                     |  True              |  web-eks-dev                   |
|  k8s.io/cluster-autoscaler/eks-dev      |  True              |  owned                         |
|  k8s.io/cluster-autoscaler/enabled      |  True              |  true                          |
|  kubernetes.io/cluster/eks-dev          |  True              |  owned                         |

# EKS MNG
aws eks describe-nodegroup \
>   --cluster-name eks-dev \
>   --nodegroup-name web-eks-dev
{
    "nodegroup": {
        "scalingConfig": {
            "minSize": 0,
            "maxSize": 5,
            "desiredSize": 0
        },
        "labels": {
            "spotInstance": "true",
            "org.com/role": "WebNode"
        },
        "tags": {
            "EnvironmentType": "development",
            "Managed_by": "terraform",
            "Environment": "eks-dev",
            "k8s.io/cluster-autoscaler/eks-dev": "eks-dev",
            "k8s.io/cluster-autoscaler/enabled": "true",
            "Name": "web-eks-dev"
        }
    }
}

So there are no any k8s.io/cluster-autoscaler/node-template/label/ tags.

Next I created my test pod:

kubectl apply -f - <<EOF
> apiVersion: v1
> kind: Pod
> metadata:
>   name: node-selector-test
>   namespace: default
> spec:
>   nodeSelector:
>     org.com/role: WebNode
>   containers:
>     - name: test
>       image: busybox
>       command: ["sh", "-c", "sleep 3600"]
> EOF
pod/node-selector-test created

Pod events

Normal   TriggeredScaleUp  6s    cluster-autoscaler  pod triggered scale-up: [{web-eks-dev-redacted 0->1 (max: 5)}]

After killing the pod, CA also scales this node group down.

So, I'm wondering what tags/labels (if any) are mandatory now for CA? while running in EKS with managed node groups?