cluster-autoscaler: Exoscale fails to get node group

[owngr]

Which component are you using?:

/area cluster-autoscaler

What version of the component are you using?:

Component version: 9.53.0

What k8s version are you using (kubectl version)?:

kubectl version Output

$ kubectl version

Client Version: v1.34.2
Kustomize Version: v5.7.1
Server Version: v1.34.2

What environment is this in?:

Exoscale

What did you expect to happen?:

cluster-autoscaler scaling down unneeded nodes

What happened instead?:

cluster-autoscale fails to get nodes groups:

I1205 08:23:11.326874       1 static_autoscaler.go:270] Starting main loop
I1205 08:23:11.327193       1 log.go:36] exoscale-provider: refreshing node groups cache
I1205 08:23:11.327214       1 log.go:32] exoscale-provider: cluster-autoscaler is disabled: no node groups found
I1205 08:23:11.328046       1 log.go:36] exoscale-provider: looking up node group for node ID REDACTED
I1205 08:23:12.279320       1 filter_out_schedulable.go:65] Filtering out schedulables
I1205 08:23:12.279352       1 filter_out_schedulable.go:122] 0 pods marked as unschedulable can be scheduled.
I1205 08:23:12.279378       1 filter_out_schedulable.go:85] No schedulable pods
I1205 08:23:12.279386       1 filter_out_daemon_sets.go:47] Filtered out 0 daemon set pods, 0 unschedulable pods left
I1205 08:23:12.279436       1 static_autoscaler.go:520] No unschedulable pods
I1205 08:23:12.279457       1 static_autoscaler.go:559] Calculating unneeded nodes
I1205 08:23:12.279479       1 log.go:36] exoscale-provider: looking up node group for node ID REDACTED
I1205 08:23:12.733911       1 static_autoscaler.go:602] Scale down status: lastScaleUpTime=2025-12-05 07:18:01.453406785 +0000 UTC m=-3583.024926371 lastScaleDownDeleteTime=2025-12-05 07:18:01.453406785 +0000 UTC m=-3583.024926371 lastScaleDownFailTime=2025-12-05 07:18:01.453406785 +0000 UTC m=-3583.024926371 scaleDownForbidden=false scaleDownInCooldown=false
I1205 08:23:12.733970       1 static_autoscaler.go:624] Starting scale down: no scale down candidates. skipping...
I1205 08:23:12.733997       1 log.go:36] exoscale-provider: looking up node group for node ID REDACTED

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

1. Create a Kubernetes cluster with metrics server in Exoscale
2. Create an IAM Role in exoscale with the following policy

{
  "default-service-strategy": "deny",
  "services": {
    "compute": {
      "type": "rules",
      "rules": [
        {
          "expression": "operation == 'get-instance'",
          "action": "allow"
        },
        {
          "expression": "operation == 'get-instance-pool'",
          "action": "allow"
        },
        {
          "expression": "operation == 'get-operation'",
          "action": "allow"
        },
        {
          "expression": "operation == 'get-quota'",
          "action": "allow"
        },
        {
          "expression": "operation == 'list-sks-clusters'",
          "action": "allow"
        },
        {
          "expression": "operation == 'scale-sks-nodepool'",
          "action": "allow"
        },
        {
          "expression": "operation == 'evict-sks-nodepool-members'",
          "action": "allow"
        }
      ]
    }
  }
}

2. Create a key for this IAM Role and put it in a secret with the zone of the cluster

apiVersion: v1
data:
  api-key: REDACTED
  api-secret: REDACTED
  api-zone: 'ch-dk-2'
kind: Secret
metadata:
  name: cluster-autoscaler-exoscale-cluster-autoscaler
type: Opaque

3. Install cluster-autoscaler with the following values:

cluster-autoscaler:
  cloudProvider: exoscale
  autoDiscovery:
    clusterName: gitlab-runner
  extraArgs:
    scale-down-unneeded-time: 15m

4. Scale workload on it to trigger the scaling

Anything else we need to know?:

It started happening the 4th December 2025 at 13:00 UTC time, I think it might be related to an Exoscale API change:
https://openapi-v2.exoscale.com/compare/main..e63a3347-d83b-4c4d-b2f8-acda36f96348

I was on 9.52.1 and I tried updating to 9.53.0 but it didn't change anything