Creating a dynamic “storage_data_disk” block for Azure VM in Terraform

Question

Within AzureRM, one or more Storage data disk(s) can be created empty and attached to the VM, e.g.:

variable "extra_disks" {
  type = list(object({
    name              = string
    disk_size_gb      = number
    managed_disk_type = string
    lun               = number
  }))
  default=[]
}

resource "azurerm_virtual_machine" "linuxvm" {
  name                  = var.vm_name 
  location              = var.location
  resource_group_name   = var.resource_group_name

  ...
  dynamic "storage_data_disk" {
    for_each = var.extra_disks 
    content {
      name              = storage_data_disk.value.name
      disk_size_gb      = storage_data_disk.value.disk_size_gb
      managed_disk_type = storage_data_disk.value.managed_disk_type
      lun               = storage_data_disk.value.lun
      caching           = "ReadWrite"
      create_option     = "Empty"
    }
  }
...
}

And I can define an array to create extra_disks dynamically, e.g. in terraform.tfvars:

extra_disks = [
  {
    name              = "data1" # logs, backup, etc.
    disk_size_gb      = 40
    managed_disk_type = "Standard_LRS"
    lun               = 0
  }
] 

I can also attach existing managed disks while creating a VM. In this case the data structure would be something like this:

variable "extra_disks" {
  type = list(object({
    name              = string
    #disk_size_gb      = null # existing data
    #managed_disk_type = null # existing data
    lun               = number
  }))
  default=[]
}

The parameters such as disk_size_gb and managed_disk_type would not be relevant in this scenario.

extra_disks = [
  {
    name              = "db-data-opt-03-manageddisk" 
    disk_size_gb      = 0 #this will not be used 
    managed_disk_type = "Standard_LRS" #this will not be used 
    lun               = 0
  }
]

And with the following terraform code, I can perform a lookup query:

data "azurerm_managed_disk" "data_disks" {
  for_each            = var.extra_disks
  name                = each.value.name
  resource_group_name = var.resource_group_name
}

Is there a way to create a dynamic block that can be used to support both these types of storage configurations?

What I have tried is the following:

variable "attach_existing_managed_data_disks" {
  default = 0
} 

dynamic "storage_data_disk" {
    for_each = var.extra_disks 
    content {
      name              = (var.attach_existing_managed_data_disks == 0 ? storage_data_disk.value.name:  
        data.azurerm_managed_disk.data_disks[*].id)
      disk_size_gb      = (var.attach_existing_managed_data_disks == 0 ? storage_data_disk.value.disk_size_gb : null)
      managed_disk_type = (var.attach_existing_managed_data_disks == 0 ? storage_data_disk.value.managed_disk_type : null)
      lun               = storage_data_disk.value.lun
      caching           = "ReadWrite"
      create_option     = (var.attach_existing_managed_data_disks == 0 ? "Empty" : "Attach" )
    }
  }

The problem is how should I map the name to associate with the data values returned in the query of data_disks while attaching existing managed disks?

      name              = (var.attach_existing_managed_data_disks == 0 ? storage_data_disk.value.name:  
        data.azurerm_managed_disk.data_disks[*].id)

Here is the error I am currently getting:

on ../terraform-azurerm-vm/main.tf line 61, in data "azurerm_managed_disk" "data_disks":

The reason is that I am creating 3 VMs using the same module and not all of the VMs have extra_disks.

The following code with count parameter would not work, as we can't have count and for_each in the same code block:

data "azurerm_managed_disk" "data_disks" {
  #count = var.attach_existing_managed_data_disks == "" ? 0 : 1

  for_each            = var.extra_disks
  name                = each.value.name
  resource_group_name = var.resource_group_name
}

If I can resolve these issues, not sure if the code block with "data_disk[*].id" would correctly index the data disks in the dynamic block:

    name = (var.attach_existing_managed_data_disks == 0 ? storage_data_disk.value.name:  
    data.azurerm_managed_disk.data_disks[*].id)

Answer 1

Creating a dynamic “storage_data_disk” block for Azure VM in Terraform.

I have followed GithubDoc as a reference to create a Azure VM with dynamic storage data disk and acttached to same VM using below terraform code.

provider "azurerm" {
  features {}
}
resource "azurerm_resource_group" "venkatrg" {
  name     = "demo-rg"
  location = "westus2"
}

resource "azurerm_virtual_network" "vnet" {
  name                = "vm-vnet"
  address_space       = ["10.0.0.0/24"]
  location            = azurerm_resource_group.venkatrg.location
  resource_group_name = azurerm_resource_group.venkatrg.name
}

resource "azurerm_subnet" "subnet" {
  name                 = "vm-subnet"
  resource_group_name  = azurerm_resource_group.venkatrg.name
  virtual_network_name = azurerm_virtual_network.vnet.name
  address_prefixes     = ["10.0.2.0/24"]
}

resource "azurerm_network_interface" "nic" {
  name                = "vm-nic"
  location            = azurerm_resource_group.venkatrg.location
  resource_group_name = azurerm_resource_group.venkatrg.name

  ip_configuration {
    name                          = "internal"
    subnet_id                     = azurerm_subnet.subnet.id
    private_ip_address_allocation = "Dynamic"
  }
}


resource "azurerm_virtual_machine" "vm-dbnode" {
  name                          = "sample-vm"
  location                      = azurerm_resource_group.venkatrg.location
  resource_group_name           = azurerm_resource_group.rg.name
  primary_network_interface_id  = azurerm_network_interface.nic.id
  network_interface_ids         = [azurerm_network_interface.nic.id]
  vm_size                       = "Standard_D8s_v3"
  delete_os_disk_on_termination = "true"

  storage_os_disk {
    name              = "vm-os-disk"
    caching           = "ReadWrite"
    managed_disk_type = "Standard_LRS"
    create_option     = "FromImage"
    disk_size_gb      = 64
  }

  storage_image_reference {
    publisher = "SUSE"
    offer     = "sles-sap-12-sp5"
    sku       = "gen1"
    version   = "latest"
  }

  dynamic "storage_data_disk" {
    iterator = disk
    for_each = range(5)
    content {
      name                      = join("-", ["data", "disk", disk.key])
      caching                   = "None"
      create_option             = "Empty"
      managed_disk_type         = "Standard_LRS"
      disk_size_gb              = 512
      write_accelerator_enabled = false
      lun                       = disk.key
    }
  }

  os_profile {
    computer_name  = "venakt-vm"
    admin_username = "Admin"
    admin_password = "Pa$$word@123$"
  }

  os_profile_linux_config {
    disable_password_authentication = false
  }
}

Terraform Apply:

Once the Terraform code is run, a virtual machine is created with five 512GB data disks, which are also attached to the VM