Terraform Meta-Arguments

Count

The 'count' attribute can be added to any resource and the value of this attribute determines the number of resource instances to be deployed. The 'count' attribute provides an 'index' property which can be use with variable interpolation to assign unique names to each resource created. The index begins at 0.

resource "azurerm_virtual_machine" "user_vm" {
  count = 3
  name = "${var.vm_name}-0${count.index}"
...
...
}

output vm_ids {
  value = azurerm_virtual_machine.user_vm[*].id
}

Terraform stores the list of resources created as an array in the terraform state and the splatting operator (an asterisk) can be used to refer to this array as a whole.

The 'count' meta-element can also be used with a list variable, allowing more tailored names to be given to the resources created:

variable "vm_names" {
  type = list(string)
  default = ["vm_harry", "vm_larry", "vm_barry"]
}

resource "azurerm_virtual_machine" "user_vm" {
  count = length(var.vm_names)
  name = "${var.vm_names[count.index]}"
...
...
}

output vm_ids {
  value = azurerm_virtual_machine.user_vm[*].id
}

For Each

The 'for_each' meta element can be used to iterate a map or set to provision multiple resources from a single resource block. 'for_each' provides a 'each.key' and an 'each.value' attribute to further tailor the properties of the resources created:

variable "user_machines" = {
  type = map(string)
  default = {
    "vm_harry" : "rg-env-dev",
    "vm_larry" : "rg-env-live",
    "vm_barry" : "rg-env-test"
  }
}

resource azurerm_virtual_machine "user_vm" {
  for_each = var.user_machines
  name = each.key
  resource_group_name = each.value
...
...
}

The 'for_each' meta-element can also be used to iterate at the property level: instead of using for_each to define multiple instance of the same resource type, for_each can be used to define multiple instances of a property on a singe resource:

resource "azurerm_app_service" "app_service" {
  name = var.app_service_name
  location = azurerm_resource_group.resource_group.location
  resource_group_name = azurerm_resource_group.resource_group.name
  app_service_plan_id = azurerm_app_service_plan.app_service_plan.id

  dynamic "connection_string" {
    for_each = var.connection_strings
      content {
        name = connection_string.name
        type = connection_string.type
        value = connection_string.value
      }
  }

  https_only = true
}

Property iterations use the keyword 'dynamic' along with the name of the property being iterated

Depends On

Terraform can identify resource dependancies automatically, and will create resources with no dependancies in parallel, whilst resources with dependancies are created sequentially. But where a resource depends on another resources behaviour rather than it's data, we have to explicitly declare the dependancy. A data dependancy is clear to see: where the value for an attribute is declared by reference to the value of an attribute of another resource. For example:

resource azurerm_virtual_machine "my_database_server" {
...
  resource_group_name = azurerm_resource_group.resource_group.name
...
}

Implicit dependancies must be declared as a list of references to other resources, using a depends_on block:

resource_azurerm_virtual_machine "my_app_server" {
...
  depends_on = [
      azurerm_virtual_machine.db_server.id,
      azurerm_virtual_machine.controller.id
  ]
}

A depends_on block can be included in any resource or module block.

Terraform creates a dependancy graph from the dependancy information to ensure that resources are created in the correct order.

Provider

Where multiple providers are being used in Terraform scripts, you can add a provider attribute to a resource to tell it to use the non-default provider

Lifecycle

The 'lifecycle' meta-argument can be used to change terraform's default behaviour when creating, updating or destroying resources

Template File

The template_file() function allows you to interpolate values into a script file at resource creation time. A simple template file might contain references to variables such as ${group} and ${user}. The template_file() function allows you to specify the values for the variables:

user_data = template_file("user_data.tftpl", { group = var.user_group, user = var.user_name })

The file() function does not interpolate values into the files contents, but allows you to read a file into your configuration without modification:

resource "aws_key_pair" "ssh_key" {
  key_name = "ssh_key"
  public_key = file("ssh_key.pub")
}