The Azure VM Scale Set (VMSS) autoscaler is not supported for use with AKS cluster node pools

Question

In the diagnose and solve problems tab of our AKS cluster, we are seeing this critical error under Reliability, the error message is
Your cluster is using an unsupported nodepool autoscaler.

The Azure VM Scale Set (VMSS) autoscaler is not supported for use with AKS cluster node pools and can lead to unexpected issues and failures, especially if used in conjunction with the AKS cluster autoscaler. AKS recommends that you should only use the AKS cluster autoscaler to scale your node pools.

VMSS auto-scale operations: userpool - Shrink (93). Please disable the VM Scale Set autoscaler and use the AKS cluster autoscaler.

After we understood this message, we have disabled the VMSS autoscaling on both agent pool and user pool, we went into vmss tab and scaling and changed scaling method back to manual and after that from azure cli, we have executed below commands to check the autoscale settings and deleted them.
az monitor autoscale list --resource-group <MC-resource-group-name> --query "[].{Name:name, Target:targetResourceUri}"

when we executed above command we got 2 autoscale settings one for agent pool and one for user pool

az monitor autoscale delete --resource-group <MC-resource-group-name> --name <autoscale-settings-name>

so we used above command to delete both autoscale settings and after that we again executed above list command and we got null output now.

Our cluster autoscaler are already enabled in aks cluster node pools tab, we have set scaling method as autoscale and min-max node count for agentpool as 1-5 and for userpool as 1-10

we confirmed they are still valid by executing below command in azure cli

az aks show --resource-group <resource-group-name> --name <cluster-name> --query "agentPoolProfiles[].{Name:name, EnableAutoScaling:enableAutoScaling, MinCount:minCount, MaxCount:maxCount}" --output table

we got the expected output

but even after waiting 24 hours the diagnose and solve problems tab in aks cluster still shows the same critical error message as mentioned above.

I want help in understanding am I still missing something or is this still showing error from cache, how to resolve it from that diagnose tab

Answer 1

hi there,

thanks for dropping this question, its super detailed and helps a ton ))

ok so u did everything right on paper,disabled vmss autoscaling, deleted the autoscale settings, and confirmed aks cluster autoscaler is running. but that pesky error won’t leave?

So according that microsoft’s docs are really clear vmss autoscaler and aks cluster autoscaler do not play nice together. like, at all. AKS cluster autoscaler docs.

Diagnose tab lags sometimes. it’s not real-time, more like a cached health check. u already did the heavy lifting, so this might just be azure being slow to update. Please check the vmss settings – sometimes the ui or cli doesn’t sync instantly. run az vmss show --resource-group <mc-rg> --name <vmss-name> --query "automaticRepairs" to see if anything’s stuck. restart the aks node pools. yes, sounds dumb, but a quick az aks nodepool restart --resource-group <rg> --cluster-name <aks> --name <nodepool> can kick the system into realizing the change.

............sometimes the backend needs a manual nudge to clear cached errors. but... I did not tell u )))))))

hope this helps,

Best regards,

Alex

and "yes" if you would follow me at Q&A - personaly thx.
P.S. If my answer help to you, please Accept my answer
PPS That is my Answer and not a Comment

https://ctrlaltdel.blog/

Answer 2

Hi harshith-8625

I undersand your issue you have added an external VMSS to your AKS cluster,this external VMSS did not support autoscaling,you disabled autoscaling for the external VMSS

You enabled autoscaling on the main AKS node pool which initially worked,But after 24 hours

you got the error:

Your cluster is using an unsupported nodepool autoscaler

AKS caches the autoscaler configuration, even after it’s removed.

External VMSSs are not officially supported for cluster autoscaling only AKS-managed node pools will support autoscaling cluster autoscaler in Azure Kubernetes Service

If autoscaler settings were once applied to the external VMSS, this configuration may still exist in AKS backend metadata.

Check your current node pools and autoscaling configuration.

az aks nodepool list --resource-group <your-rg> --cluster-name <your-cluster>

Manually Clear Autoscaler on External VMSS

az vmss update \
  --name <vmss-name> \
  --resource-group <your-rg> \
  --set upgradePolicy.automaticOSUpgradePolicy.enableAutomaticOSUpgrade=false \
  --set virtualMachineProfile.extensionProfile=null

Enable in aks cluster node pools

az aks nodepool update \
  --resource-group <your-rg> \
  --cluster-name <your-cluster> \
  --name <nodepool-name> \
  --enable-cluster-autoscaler \
  --min-count <your-min-count> \
  --max-count <your-max-count>

Reapplying this forces AKS to re-sync node pool metadata and clear stale configurations from backend.

Restart the Control Plane because when you remove autoscaler settings or modify node pools, AKS backend sometimes keeps stale state,

Scale down agentpools temporarily

az aks nodepool scale \
  --resource-group <your-resource-group> \
  --cluster-name <your-cluster-name> \
  --name agentpool \
  --node-count 1  

Then scale back to previous count

az aks nodepool scale \
  --resource-group <your-resource-group> \
  --cluster-name <your-cluster-name> \
  --name agentpool \
  --node-count 10

Do not use VMSS autoscaler with AKS, it should only be used in AKS-managed node pools with AKS autoscaler.

If you have any further queries, let me know. If the information is helpful, please click on Upvote.

Thank you.