This post will demonstrate how to create a segregated Azure Cloud environment (a resource group and service connection from Azure DevOps) that can be used by development teams to provision their own infrastructure and deploy their compiled code. All in an automated and secure manner.

Teams can then setup their own release pipelines using the provisioned resource group and service connection. The service connection created will only have permission to a single resource group within Azure.

Specifically, this post will demonstrate:

  • How to provision a resource group for an application or service; and
  • How to create a service connection from Azure DevOps to Azure Cloud with contributor access to that resource group.

The above will be done with Terraform wrapped in a Python script. The reference code for this post is available on this GitHub repository.

Pre-prerequisites and Setup

In order to automate the above process Terraform has two providers: an Azure RM provider and an Azure DevOps provider. Both of these will be utilised in order to create a resource group with a service connection.

The first step is authenticating to an Azure DevOps instance and an Azure Cloud Subscription. To authenticate to Azure Cloud the Azure CLI is needed.

% az login
The default web browser has been opened at Please continue the login in the web browser. If no web browser is available or if the web browser fails to open, use device code flow with `az login --use-device-code`.

There are several other ways to authenticate to Azure Cloud and they are listed on the Terraform website.

To authenticate into Azure DevOps, A Personal Access Token (PAT) is needed with the scope Read, query, & manage on Service Connections. After generating the PAT, Terraform requires certain environment variables to be set:

% export AZDO_PERSONAL_ACCESS_TOKEN={a-secure-access-token}
% export AZDO_ORG_SERVICE_URL={a-secure-org}

Initialising Terraform

Terraform relies on a state file to perform resource deployments. For the purposes of this post and to keep things simple, local system state will be used. For production work-loads remote state should be used.

The Python script, initialises Terraform state dynamically by assigning a different state file according to the name of the resource group desired to be provisioned. This means that the same script can be used for multiple resource groups by changing the argument --resource-group-name.

For example: if the script is run with the below command:

% ./ --resource-group-name "a-new-resource-group"  --azure-devops-project-name main

The resulting terraform init command will be

% terraform init  -backend-config='path=.terraform-state-a-new-resource-group/a-new-resource-group.tfstate' -reconfigure --input=false

Terraform is able to dynamically change state via the -backend-config flag.

So, if the script is ran twice with two different resource groups like the below snippet:

% ./ --resource-group-name "a-new-resource-group" --azure-devops-project-name main
% ./ --resource-group-name "a-new-resource-group-2" --azure-devops-project-name main

Two folders will be created in the working directory: .terraform-state-a-new-resource-group and .terraform-state-a-new-resource-group-2. Both of which have a tfstate file.

This pattern can be achieved for any Terraform remote backend and will allow the executor to use the same Terraform templates to create multiple instances of the resources declared in the Terraform template. Note it is important to treat Terraform state files as sensitive data (passwords, tokens). See this Terraform page for more information.

Creating the Service principal

A service principal is needed to create the service connection between Azure DevOps and Azure Cloud. This service principal is provisioned with a randomly generated password by Terraform, which will be used by the service connection.

The resource used to generate this is random_password. A snippet of the code is shown below and the full code is available on GitHub.

resource "azuread_service_principal" "service-principal" {
  application_id               =

resource "azuread_service_principal_password" "service-principal-password" {
  service_principal_id =
  description          = "Managed by Terraform"
  value                = random_password.service-principal-password.result
  end_date             = "2099-01-01T01:02:03Z"

resource "random_password" "service-principal-password" {
  length           = 16
  special          = true
  override_special = "_%@"

Creating the Resource Group

The new resource group needs to be declared with the Terraform resource type azurerm_role_assignment with Contributor access given to the previously declared service principal. The snippet of the code is shown below and the full code is available on GitHub

resource "azurerm_resource_group" "resource-group" {
  name     = var.resource-group-name
  location = "Australia East"

resource "azurerm_role_assignment" "resource-group-service-principal-role-assignment" {
  scope                =
  role_definition_name = "Contributor"
  principal_id         =

Creating the Service Connection

The final resource to provision is the service connection from Azure DevOps to Azure Cloud. At this stage, a resource group has been provisioned and a service principal has been provisioned with a password. To create this service connection the resource azuredevops_serviceendpoint_azurerm is used. Access to the password is available as a resource attribute, so this can be used to provision the service connection on Azure DevOps. The snippet of the code is show below and the full code is available on GitHub

resource "azuredevops_serviceendpoint_azurerm" "serviceendpoint-azure" {
  project_id            =
  service_endpoint_name = "${var.resource-group-name}-azure-service-connection"
  description = "Managed by Terraform" 
  credentials {
    serviceprincipalid  =
    serviceprincipalkey = random_password.service-principal-password.result
  azurerm_spn_tenantid      = data.azurerm_subscription.current.tenant_id
  azurerm_subscription_id   = data.azurerm_subscription.current.subscription_id
  azurerm_subscription_name = data.azurerm_subscription.current.display_name


The above setup creates a resource group and a service principal with Contributor access to that resource group. It also creates a service connection from Azure DevOps that uses the provisioned service principal that can then be used in Azure Release pipelines. This means that this service connection can be used by Azure DevOps pipelines to automate the deployments of resources and compiled code into the provisioned resource group ensuring an automated, isolated, and secure release process into Azure Cloud.

If the script is run as follows:

% ./ --resource-group-name "an-example-resource-group" --azure-devops-project-name main
Apply complete! Resources: 7 added, 0 changed, 0 destroyed.

The result would be a resource group an-example-resource-group provisioned on Azure Cloud with the service principal an-example-resource-group-spn having Contributor access to that resource group.

Resource Groups Provisioned with Terraform

As well as a service connection from Azure DevOps into Azure Cloud that can be used by release pipelines.

Resource Groups Provisioned with Terraform

Finally, thank you for reading this post. I hope it has been beneficial for you.