I am not able to configure auto-scaling for my existing VMSS node pools for my AKS cluster

Question

I am not able to configure auto-scaling for my existing VMSS node pools for my AKS cluster

cargobuddy 20

I have a private AKS cluster, with two VMSS node pools, one for system and one for user. The node pool, when created, is preconfigured with manual scaling. However, I want to leverage the auto-scaling feature. For that I applied the below command, where I got the error as shown:
Command:

az aks nodepool update --resource-group <<rg>> --cluster-name <<cluster-name>>--name <<pool-name>> --enable-cluster-autoscaler --min-count 2 --max-count 2
Error:
(OperationNotAllowed) .properties.nodeProvisioningProfile.mode cannot be Auto unless all AgentPools have property .properties.enableAutoScaling set to one of [false].

Code: OperationNotAllowed

Message: .properties.nodeProvisioningProfile.mode cannot be Auto unless all AgentPools have property .properties.enableAutoScaling set to one of [false].

Megha Ramakrishnan 325 Reputation points

2026-06-20T20:10:03.6433333+00:00

Hi @cargobuddy ,

This error occurs because your AKS cluster has been configured with AKS Automatic SKU features, or specifically Node Auto-Provisioning (NAP) (which relies on an Azure-managed Karpenter provider).

When ⁠.properties.nodeProvisioningProfile.mode⁠ is set to ⁠Auto⁠, Azure expects to control node creation and sizing completely dynamically based on pending pod requirements. It strictly prohibits you from enabling the traditional cluster autoscaler (⁠--enable-cluster-autoscaler⁠) on individual manual agent pools, resulting in the validation mismatch you encountered.

To resolve this issue, you have two paths depending on how you want your cluster to behave:

Option 1: Turn off Node Auto-Provisioning to use the Standard Autoscaler

If you want to manage traditional VMSS node pools manually and configure typical min/max bounds using the AKS cluster autoscaler, you must first switch the cluster's provisioning mode back to ⁠Manual⁠.

Change the Node Provisioning Mode to Manual:

az aks update --resource-group <<rg>> --name <<cluster-name>> --node-provisioning-mode Manual

Re-run your Node Pool Update Command:

Once the profile mode is reverted, your original command will be permitted:

az aks nodepool update --resource-group <<rg>> --cluster-name <<cluster-name>> --name <<pool-name>> --enable-cluster-autoscaler --min-count 2 --max-count 2

Option 2: Lean into Node Auto-Provisioning

If you intended to use the newer auto-provisioning framework (where you don't need to manually scale or pre-configure individual node pool boundaries), you don't need to run ⁠az aks nodepool update⁠ to turn on the autoscaler.

Under the ⁠Auto⁠ profile, Kubernetes workloads automatically dictate when new infrastructure is provisioned. Instead of setting explicit pool counts via CLI, you configure horizontal pod autoscalers (HPA) or resource requests in your deployment manifests, and Azure handles the underlying node scale-out seamlessly.

Please 'Upvote'(Thumbs-up) and 'Accept' as answer if the reply was helpful. This will be benefitting other community members who face the same issue.

Thank you!
Himanshu Shekhar 6,790 Reputation points Microsoft External Staff Moderator

2026-06-22T01:06:37.6966667+00:00
cargobuddy - please keep manual VMSS node pools + use the standard Cluster Autoscaler :

Always switch the cluster's provisioning mode back to Manual, then enable the autoscaler:

# 1. Revert provisioning mode to Manual az aks update --resource-group <rg> --name <cluster-name> --node-provisioning-mode Manual # 2. Re-run your node pool update az aks nodepool update --resource-group <rg> --cluster-name <cluster-name> --name <pool-name> \ --enable-cluster-autoscaler --min-count 2 --max-count 2

Note: NAP can only be disabled if there are no existing NAP‑managed nodes (kubectl get nodes -l karpenter.sh/nodepool) and all Karpenter NodePools have spec.limits.cpu set to 0. Clear those first or the switch to Manual will fail. - https://learn.microsoft.com/en-us/azure/aks/use-node-auto-provisioning?pivots=azure-cli

Option 2

Stay on Node Auto‑Provisioning (no autoscaler needed)

If NAP was intended, you don't run az aks nodepool update at all. Under Auto mode, scaling is driven by pending pod resource requests and just define your NodePool / AKSNodeClass CRDs (and HPA / resource requests on your workloads) and NAP provisions right‑sized nodes automatically. - https://learn.microsoft.com/en-us/azure/aks/node-auto-provisioning
Himanshu Shekhar 6,790 Reputation points Microsoft External Staff Moderator

2026-06-23T05:47:44.3433333+00:00

Just checking if provided response was helpful! please let me know if you have any queries.
cargobuddy 20 Reputation points

2026-06-23T14:35:41.4633333+00:00

Hello @Himanshu Shekhar / @Megha Ramakrishnan

By looking at the JSON template, I saw the below:

Does it mean, my cluster is already having auto-scaling enabled with NAP pool? If yes, then what are the minimum and maximum values for the nodes? Can I control that or it is completely managed by Azure? How can I test the auto-scaling is effective?

1 answer

Your answer

Himanshu Shekhar 6,790 Reputation points Microsoft External Staff Moderator

2026-06-22T01:06:37.6966667+00:00

cargobuddy - please keep manual VMSS node pools + use the standard Cluster Autoscaler :

Always switch the cluster's provisioning mode back to Manual, then enable the autoscaler:

# 1. Revert provisioning mode to Manual az aks update --resource-group <rg> --name <cluster-name> --node-provisioning-mode Manual # 2. Re-run your node pool update az aks nodepool update --resource-group <rg> --cluster-name <cluster-name> --name <pool-name> \ --enable-cluster-autoscaler --min-count 2 --max-count 2

Note: NAP can only be disabled if there are no existing NAP‑managed nodes (kubectl get nodes -l karpenter.sh/nodepool) and all Karpenter NodePools have spec.limits.cpu set to 0. Clear those first or the switch to Manual will fail. - https://learn.microsoft.com/en-us/azure/aks/use-node-auto-provisioning?pivots=azure-cli

Option 2

Stay on Node Auto‑Provisioning (no autoscaler needed)

If NAP was intended, you don't run az aks nodepool update at all. Under Auto mode, scaling is driven by pending pod resource requests and just define your NodePool / AKSNodeClass CRDs (and HPA / resource requests on your workloads) and NAP provisions right‑sized nodes automatically. - https://learn.microsoft.com/en-us/azure/aks/node-auto-provisioning
Himanshu Shekhar 6,790 Reputation points Microsoft External Staff Moderator

2026-06-23T05:47:44.3433333+00:00

Just checking if provided response was helpful! please let me know if you have any queries.
cargobuddy 20 Reputation points

2026-06-23T14:35:41.4633333+00:00

Hello @Himanshu Shekhar / @Megha Ramakrishnan

By looking at the JSON template, I saw the below:

Does it mean, my cluster is already having auto-scaling enabled with NAP pool? If yes, then what are the minimum and maximum values for the nodes? Can I control that or it is completely managed by Azure? How can I test the auto-scaling is effective?

Answer 1

hi cargobuddy, thx for sharing urs issue here at Q&A portal,

check whether the AKS cluster is using Node Auto-Provisioning/Auto mode. If yes, don’t enable the classic autoscaler on existing node pools. Pick one model NAP/Karpenter-style provisioning, or classic node pool autoscaler. Mixing both is what’s causing the OperationNotAllowed error, this error doesn’t look like a VMSS problem directly. It looks like ur AKS cluster has Node Auto-Provisioning / Auto mode enabled, and that conflicts with enabling the classic cluster autoscaler on existing agent pools.

Microsoft describes Node Auto-Provisioning as a newer AKS feature that automatically provisions and manages nodes based on pending pod needs, using Karpenter under the hood. That’s different from the classic AKS cluster autoscaler, which scales existing node pools. The key part of the error is this

nodeProvisioningProfile.mode cannot be Auto unless all AgentPools have enableAutoScaling set to false

AKS is saying, “u can’t use this Auto node provisioning mode while any normal node pool autoscaler is enabled.” Yeah, very Azure-ish wording technically correct, emotionally hostile.

I’d verify the cluster-level node provisioning mode

az aks show \

--resource-group <rg> \

--name <cluster-name> \

--query "nodeProvisioningProfile"

check the node pools

az aks nodepool list \

--resource-group <rg> \

--cluster-name <cluster-name> \

--query "[].{name:name, enableAutoScaling:enableAutoScaling, count:count, min:minCount, max:maxCount, mode:mode}"

If nodeProvisioningProfile.mode is Auto, then the fix is not to enable autoscaler on the existing VMSS node pools. U either keep Node Auto-Provisioning and leave enableAutoScaling false on all agent pools, or u disable Node Auto-Provisioning and then use the classic cluster autoscaler on the node pools.

One more thing this command doesn’t really make sense for autoscaling

--min-count 2 --max-count 2

That means the autoscaler has nowhere to scale. Min and max are both 2, so it’s basically manual scaling wearing a fake mustache. Use something like --min-count 2 --max-count 5 if u actually want scaling.

https://learn.microsoft.com/en-us/azure/aks/cluster-autoscaler

rgds,

Alex

&

If my answer was helpful pls mark it and additional thx if u follow me at Q&A portal