Terraform Composition

Repositories

Each Terraform AVM module will have its own GitHub Repository in the Azure GitHub Organization as per SNFR19.

This repo will be created by the Module Owners and the AVM Core team collaboratively, including the configuration of permissions as per SNFR9

Directory and File Structure

Below is the directory and file structure expected for each AVM Terraform repository/module.
See template repo here.

• tests/ - (for unit tests and additional tests if required - e.g. tflint etc.)
  • unit/ - (optional, may use further sub-directories if required)
• modules/ - (for sub-modules only if used)
• examples/ - (all examples must deploy without successfully without requiring input - these are customer facing)
  • <at least one folder> - (at least one example that uses the variable defaults minimum/required parameters/variables only)
  • <other folders for examples as required>
• /... - (Module files that live in the root of module directory)
  • _header.md - (required for documentation generation)
  • _footer.md - (required for documentation generation)
  • main.tf
  • locals.tf
  • variables.tf
  • outputs.tf
  • terraform.tf
  • README.md (autogenerated)
  • main.resource1.tf (If a larger module you may chose to use dot notation for each resource)
  • locals.resource1.tf

Directory and File Structure

/ root
│
├───.github/
│   ├───actions/
│   │   ├───avmfix/
│   │   │   └───action.yml
│   │   ├───docs-check/
│   │   │   └───action.yml
│   │   ├───e2e-getexamples/
│   │   │   └───action.yml
│   │   ├───e2e-testexamples/
│   │   │   └───action.yml
│   │   ├───linting/
│   │   │   └───action.yml
│   │   └───version-check/
│   │       └───action.yml
│   ├───policies/
│   │   ├───avmrequiredfiles.yml
│   │   └───branchprotection.yml
│   ├───workflows/
│   │   ├───e2e.yml
│   │   ├───linting.yml
│   │   └───version-check.yml
│   ├───CODEOWNERS
│   └───dependabot.yml
├───.vscode/
│   ├───extensions.json
│   └───settings.json
├───examples/
│   ├───<example_folder>/
│   │   ├───README.md
│   │   ├───_footer.md
│   │   ├───_header.md
│   │   ├───main.tf
│   │   └───variables.tf
│   ├───.terraform-docs.yml
│   └───README.md
├───modules/
│   └───README.md
├───tests/
│   └───README.md
├───.gitignore
├───.terraform-docs.yml
├───CODE_OF_CONDUCT.md
├───LICENSE
├───Makefile
├───README.md
├───SECURITY.md
├───SUPPORT.md
├───_footer.md
├───_header.md
├───avm
├───avm.bat
├───locals.tf
├───main.privateendpoint.tf
├───main.telemetry.tf
├───main.tf
├───outputs.tf
├───terraform.tf
└───variables.tf

Code Styling

This section points to conventions to be followed when developing a module.

Casing

Use snake_casing as per TFNFR3.

Input Parameters and Variables

Make sure to review all specifications of Category: Inputs/Outputs within the Terraform specification pages.

Resources

Resources are primarily leveraged by resource modules to declare the primary resource of the main resource type deployed by the AVM module.

Make sure to review all specifications covering resource properties and usage.

Outputs

Make sure to review all specifications of Category: Inputs/Outputs within the Terraform specification pages.

Interfaces

To meet RMFR4 and RMFR5 AVM resource modules must leverage consistent interfaces for all the optional features/extension resources supported by the AVM module primary resource.

Please refer to the Terraform Interfaces page.

Telemetry

To meet SFR3 & SFR4.
We use a telemetry provider modtm.
This is a lightweight telemetry provider that sends telemetry data to Azure Application Insights via a HTTP POST front end service.

The telemetry provider is included in the module by default and is enabled by default.
You do not need to change the configuration included in the template repo.

You must make sure to have the modtm provider in your required_providers.
The linter will check this for you.
However, inside your terraform.tf file please make sure you have this entry:

terraform {
  required_providers {
    # .. other required providers as needed
    modtm = {
      source = "Azure/modtm"
      version = "~> 0.3"
    }
  }
}

Eventual Consistency

When creating modules, it is important to understand that the Azure Resource Manager (ARM) API is sometimes eventually consistent.
This means that when you create a resource, it may not be available immediately.
A good example of this is data plane role assignments.
When you create such a role assignment, it may take some time for the role assignment to be available.
We can use an optional time_sleep resource to wait for the role assignment to be available before creating resources that depend on it.

# In variables.tf...
variable "wait_for_rbac_before_foo_operations" {
  type: object({
    create  = optional(string, "30s")
    destroy = optional(string, "0s")
  })
  default     = {}
  description = <<DESCRIPTION
This variable controls the amount of time to wait before performing foo operations.
It only applies when `var.role_assignments` and `var.foo` are both set.
This is useful when you are creating role assignments on the bar resource and immediately creating foo resources in it.
The default is 30 seconds for create and 0 seconds for destroy.
DESCRIPTION
}

# In main.tf...
resource "time_sleep" "wait_for_rbac_before_foo_operations" {
  count = length(var.role_assignments) > 0 && length(var.foo) > 0 ? 1 : 0
  depends_on = [
    azurerm_role_assignment.this
  ]
  create_duration  = var.wait_for_rbac_before_foo_operations.create
  destroy_duration = var.wait_for_rbac_before_foo_operations.destroy

  # This ensures that the sleep is re-created when the role assignments change.
  triggers = {
    role_assignments = jsonencode(var.role_assignments)
  }
}

resource "azurerm_foo" "this" {
  for_each = var.foo
  depends_on = [
    time_sleep.wait_for_rbac_before_foo_operations
  ]
  # ...
}