Skip to main content
When the instances in a virtual machine scale set barely use their CPU, memory and network, Azure Advisor recommends shutting the scale set down rather than resizing it. nOps imports these recommendations, checks them against your bill, and shows the estimated monthly saving.

Why it matters

A scale set that does almost no work still bills every instance for every hour it runs.

How nOps finds it

This recommendation comes from Azure Advisor, which looks at the scale set’s use over its lookback period and recommends shutting it down. nOps imports Advisor’s recommendations daily and checks each one against your own Azure bill for the last 7 complete days: the resource must still be billed in the last 3 days of that window, and Advisor’s estimate is capped at what the resource actually cost (Advisor prices at pay-as-you-go rates, so you may pay less through discounts, reservations or savings plans). Recommendations worth less than $1 a month are left out.

Before you start

Confirm with the owner that nothing depends on the scale set, including scheduled or occasional jobs.

How to implement

  1. Confirm with the owner what the scale set runs and whether it is still needed.
  2. To keep it but stop paying for compute, scale it to zero instances (az vmss scale --resource-group RESOURCE_GROUP --name SCALE_SET --new-capacity 0) or deallocate it (az vmss deallocate).
  3. If it is no longer needed, delete it (az vmss delete) and remove it from Terraform, Bicep or ARM templates.
  4. For AKS node pools, scale or delete the node pool through AKS instead.

Risks and considerations

  • Managed disks keep billing while instances are deallocated; only deleting removes them.
  • Autoscale rules can scale the set back up. Adjust or disable them.
  • Deleting cannot be undone; scaling to zero can.