Garden includes a Terraform provider that you can use to automatically validate and provision infrastructure as part of your project. This guide walks through how to configure and use the provider.

It's strongly recommended that you learn about Terraform (if you haven't already) before using it with Garden.

How it works

Under the hood, Garden simply wraps Terraform, so there's no magic involved. Garden just automates its execution and makes stack outputs available to your Garden providers and modules.

The terraform provider can both provision a Terraform stack when initializing Garden, or through terraform modules that are deployed like other services in your stack.

The former, having a single Terraform stack for your whole project, is most helpful if other provider configurations need to reference the outputs from your Terraform stack, or if most/all of your services depend on the infrastructure provisioned in your Terraform stack. A good example of this is the terraform-gke example project, which provisions a GKE cluster that the kubernetes provider then runs on, along with the services in the project. The drawback is that Garden doesn't currently watch for changes in those Terraform files, and you need to restart to apply new changes, or apply them manually.

The latter method, using one or more terraform modules, can be better if your other providers don't need to reference the stack outputs but your services, tasks and tests do. In this style, you can basically create small Terraform stacks that are part of your Stack Graph much like other services. A good example would be deploying a database instance, that other services in your project can then connect to.

You can also use a combination of the two if you'd like. Below we'll walk through how each of these work.

Configuring the provider

First off, you need to enable the provider in your project configuration. This is as simple as placing it in your list of providers:

kind: Project
name: my-project
providers:
  - name: terraform
  - name: kubernetes
  ...

If you'd like to apply the stack when starting Garden, and then reference the stack outputs in other providers (or modules), you need to add a couple of more flags. Here's the project config from the aforementioned terraform-gke example:

kind: Project
name: terraform-gke
providers:
  - name: terraform
    # This must be set if we want to resolve a stack as part of the provider initialization.
    initRoot: "."
    # You can either replace these with your own values, or delete these and provide your own in a
    # terraform.tfvars file in the project root.
    variables:
      gcp_project_id: garden-gke-tf-1
      gcp_region: europe-west1
  - name: kubernetes
    kubeconfig: ${providers.terraform.outputs.kubeconfig_path}
    context: gke
    defaultHostname: terraform-gke-${local.username}.dev-2.sys.garden
    buildMode: cluster-docker

The initRoot parameter tells Garden that there is a Terraform working directory at the specified path. If you don't specify this, Garden doesn't attempt to apply a stack when initializing the provider.

Notice also that we're providing an output value from the stack to the kubernetes provider. This can be very powerful, and allows you to fully codify your full project setup, not just the services running in your environment. Any Garden module can also reference the provider outputs in the exact same way, so you can easily provide your services with any information they need to operate.

Planning and applying

Garden will not automatically apply the Terraform stack, unless you explicitly set the autoApply flag on the config for the stack. Instead, Garden will warn you if the stack is out of date.

To manually plan and apply stacks, we provide the following commands:

garden --env=<env-name> plugins terraform apply-root                  # Runs `terraform apply` for the provider root stack.
garden --env=<env-name> plugins terraform apply-module <module-name>  # Runs `terraform apply` for the specified terraform module.
garden --env=<env-name> plugins terraform plan-root                   # Runs `terraform plan` for the provider root stack.
garden --env=<env-name> plugins terraform plan-module <module-name>   # Runs `terraform plan` for the specified terraform module.

Each command automatically applies any variables configured on the provider/module in question. Any additional arguments you specify for the command are passed directly to the terraform CLI command, but you need to place them after a -- so that they aren't parsed as Garden options. For example, to apply the root stack with -auto-approve:

garden --env=<env-name> plugins terraform apply-root -- -auto-approve

Terraform modules

You can also define terraform modules as part of your project, which act much like other Garden modules. A terraform module maps to a single service, that you can define as a runtime dependency for any of your other services and tasks. You can also reference the stack outputs of a terraform module using runtime output template strings. For example:

kind: Module
type: terraform
name: tf
autoApply: true
---
kind: Module
type: container
name: my-container
services:
  - name: my-service
    # Important! You must declare the terraform service as a dependency, for the runtime template string to work.
    dependencies: [tf]
    env:
      DATABASE_URI: ${runtime.services.tf.outputs.my-database-uri}

Here we imagine a Terraform stack that has a my-database-uri output, that we then supply to my-service via the DATABASE_URI environment variable.

Much like other modules, you can also reference Terraform definitions in other repositories using the repositoryUrl key. See the Remote Sources guide for details.

Next steps

Check out the terraform-gke example project. Also take a look at the Terraform provider reference and the Terraform module type reference for details on all the configuration parameters.

If you're having issues with Terraform itself, please refer to the official docs.