windows_virtual_machine_scale_set SKU(Size) change cause whole VMSS resource redeployment

[SlavisaBakicOB]

Is there an existing issue for this?

• I have searched the existing issues

Community Note

• Please vote on this issue by adding a 👍 reaction to the original issue to help the community and maintainers prioritize this request
• Please do not leave "+1" or "me too" comments, they generate extra noise for issue followers and do not help prioritize the request
• If you are interested in working on this issue or have submitted a pull request, please leave a comment

Terraform Version

1.2.7

AzureRM Provider Version

3.19.1

Affected Resource(s)/Data Source(s)

azurerm_windows_virtual_machine_scale_set

Terraform Configuration Files

resource "azurerm_windows_virtual_machine_scale_set" "vmss_winsrv" {
    tags = merge(local.default_tags)
    resource_group_name         = var.resource_group_name
    name                        = local.vmss_name
    location                    = var.location
    provision_vm_agent          = true
    enable_automatic_updates    = true
    admin_password = var.admin_password
    admin_username = var.admin_username
    
    computer_name_prefix = "vmsstest"

    sku = var.vmss_sku  

    instances = var.instance_count

    upgrade_mode = "Manual"

    

    os_disk {
        caching              = lookup(var.storage_os_disk_config, "caching", "ReadWrite")
        storage_account_type = lookup(var.storage_os_disk_config, "storage_account_type", "Standard_LRS")
        disk_size_gb         = lookup(var.storage_os_disk_config, "disk_size_gb", 127)
        write_accelerator_enabled = false
    }

    source_image_reference {
        publisher      = lookup(var.vm_image, "publisher", "MicrosoftWindowsServer")
        offer          = lookup(var.vm_image, "offer", "WindowsServer")
        sku            = lookup(var.vm_image, "sku", "2019-Datacenter")
        version        = lookup(var.vm_image, "version", "latest")
    }

    network_interface {
         name    = "vm_scale_set_net"
         primary = true        
         ip_configuration {
             name      = "internal"
             primary   = true
             subnet_id = var.vmss_subnet_id
             load_balancer_backend_address_pool_ids = [azurerm_lb_backend_address_pool.vmss_lb_bep.id]

         }
     }

  lifecycle {
    ignore_changes = [
     tags,
     resource_group_name,
     name,
     provision_vm_agent,
     enable_automatic_updates,
     admin_password,
     admin_username,
     computer_name_prefix,
     instances,
     upgrade_mode,
     os_disk,
     source_image_reference,
     network_interface 
    ]
  }
}

Debug Output/Panic Output

azurerm_windows_virtual_machine_scale_set.vmss_winsrv: Modifications complete after 5m0s [id=/subscriptions/*ommited*/resourceGroups/*ommited*/providers/Microsoft.Compute/virtualMachineScaleSets/*ommited*]

Expected Behaviour

TF should just change size if only sku attribute is changed without redeploying whole VMSS resource.
Manually chaning VMSS Size via Portal is completed in less than 10 seconds and VMSS is not recreated.

Actual Behaviour

TF redeployed whole VMSS resource which can take up to 5 minutes instead of just changing VMSS sku which should not take more than 20 seconds.

Steps to Reproduce

No response

Important Factoids

No response

References

No response

[myc2h6o]

Hi @SlavisaBakicOB thanks for opening the issue! When upgrade_mode is set to Manual, changing sku on Azure Portal is quick since it doesn't update the sku for existing VM in the scale set. In Terraform however, it is designed to run without manual operation outside Terraform, thus even with the Manual upgrade mode, we still do an update to the VM inside the scale set.

terraform-provider-azurerm/internal/services/compute/virtual_machine_scale_set_update.go

Lines 59 to 60 in c7e7b2b

	if upgradeMode == compute.UpgradeModeManual {
	if err := metadata.upgradeInstancesForManualUpgradePolicy(ctx); err != nil {

[geglne]

This is unfortunate. As SlavisaBakicOB said above, when changing the sku - not unlike when manually changing it in the portal - it would be nice to preserve the OS disk.

For me, the speed of the deployment isn't an issue. It is our customers who have changed the VM sku (for whatever reason they have), and are surprised when their OS disk has been re-formated.

I realize that a VM and a VMSS are separate resources altogether, however the virtual_machine has a delete_os_disk_on_termination argument and it would be nice to have a similar flag like that.

Workaround

• Log into the portal
• Locate the VMSS
• Upgrade the Size of your VMSS
• Edit the sku in the azurerm_virtual_machine_scale_set to match the manually set size.
• Execute a terraform plan to validate the change:
  

[myc2h6o]

@geglne
I've been adding a feature flag to skip the reimage process during the upgrade when upgrade_mode is Manual with #22975.
By specifying its value to false, the OS disk shall be preserved after the update.

[github-actions]

I'm going to lock this issue because it has been closed for 30 days ⏳. This helps our maintainers find and focus on the active issues.
If you have found a problem that seems similar to this, please open a new issue and complete the issue template so we can capture all the details necessary to investigate further.