Cluster for AKS service no autoscaling anymore

Hello @Prabjyot Kaur

Thanks for the details — losing autoscaling right after a 1.33 upgrade is frustrating, and the good news is this is almost always quick to recover. Let me explain the likely cause and the fastest fix.

About the "Customer Initiated" label In Azure Resource Health, "Customer Initiated" just means the state change was triggered by a customer action — and a cluster upgrade is a customer-initiated action. So this label is most likely reflecting your 1.33 upgrade, not pointing to anything else you've done. The most common reason autoscaling stops right after an upgrade is the node pool entering a backoff state after the scaling/upgrade operation, which makes the autoscaler appear "unavailable."

Step 1 — Reset the backoff (most likely fix) Disabling and re-enabling autoscaling clears the backoff state and forces a fresh scaling attempt. This is the documented reset for both node-pool backoff and the upstream "Failed to fix node group sizes" deadlock.

Per node pool:

az aks nodepool update -g <rg> --cluster-name <cluster> -n <nodepool> \     --disable-cluster-autoscaler  az aks nodepool update -g <rg> --cluster-name <cluster> -n <nodepool> \     --enable-cluster-autoscaler --min-count <min> --max-count <max>

If autoscaling is managed at the cluster level rather than per node pool, use

az aks update --disable-cluster-autoscaler

then

az aks update --enable-cluster-autoscaler --min-count <min> --max-count <max>

instead — just don't mix the two scopes.

Backoff can take up to ~30 minutes to clear on its own, but the disable/re-enable above resets it immediately.

Ref: https://learn.microsoft.com/en-us/azure/aks/cluster-autoscaler-overview#common-issues-and-mitigation-recommendations

Step 2 — Confirm the actual reason in the logs So we're certain rather than guessing, check the cluster autoscaler logs (Log Analytics) or Diagnose & Solve Problems. If you see Failed to fix node group sizes, that's a known upstream race condition and the Step 1 disable/re-enable is exactly the documented fix.

Ref: https://learn.microsoft.com/en-us/troubleshoot/azure/azure-kubernetes/create-upgrade-delete/cluster-autoscaler-fails-to-scale

Step 3 — Confirm the cluster/node pool isn't in a Failed provisioning state
Run

az aks show -g <rg> -n <cluster> --query provisioningState 

and check the node pool state. If a partial upgrade left it Failed, autoscaler operations are blocked until reconciled

 az aks update -g <rg> -n <cluster> 

(no other args) reconciles it.

Step 4 — Rule out blockers and capacity limits If Steps 1–3 don't restore scaling, check:

• Resource locks / Azure Policy on the cluster or the managed node resource group (MC_<rg>_<cluster>_<region>). Remove locks; locked write operations block the autoscaler. Please don't modify the managed node RG directly — it's unsupported and breaks autoscaler operations. Ref: Write operations locked.
• Capacity limits: core quota, subnet IP exhaustion (Azure CNI), the node pool's max-count, or 429 throttling — any of these pushes the autoscaler into exponential backoff.

Could you run Steps 1–3 and share the autoscaler log output (especially any Failed to fix node group sizes, quota, subnet, or 429 messages) plus the provisioningState? That will let me confirm whether this was upgrade backoff, the upstream deadlock, or a lock/quota block.

Thanks,
Manish.

Hello

I wanted to check if my last response made sense. I’d be glad to assist further or explain anything in more detail and please accept as Yes and upvote if the answer is helpful so that it can help others in the community.