Home

Awesome

🚨 PLEASE NOTE 🚨: This repository is no longer maintained.

New users: Use its successor Official Upbound Terraform Provider instead.

Existing users: Follow the migration guide here.

This repository will receive bugfix-only critical changes until the end of transition period on 1st of Feb 2023.

provider-terraform

An experimental Crossplane provider for Terraform. Use this provider to define new Crossplane Composite Resources (XRs) that are composed of a mix of 'native' Crossplane managed resources and your existing Terraform modules.

The Terraform provider adds support for a Workspace managed resource that represents a Terraform workspace. The configuration of each workspace may be either fetched from a remote source (e.g. git), or simply specified inline.

apiVersion: tf.crossplane.io/v1alpha1
kind: Workspace
metadata:
  name: example-inline
  annotations:
    # The terraform workspace will be named 'coolbucket'. If you omitted this
    # annotation it would be derived from metadata.name - i.e. 'example-inline'.
    crossplane.io/external-name: coolbucket
spec:
  forProvider:
    # For simple cases you can use an inline source to specify the content of
    # main.tf as opaque, inline HCL.
    source: Inline
    module: |
      // All outputs are written to the connection secret.  Non-sensitive outputs
      // are stored as string values in the status.atProvider.outputs object.
      output "url" {
        value       = google_storage_bucket.example.self_link
      }

      resource "random_id" "example" {
        byte_length = 4
      }

      // The google provider and remote state are configured by the provider
      // config - see examples/providerconfig.yaml.
      resource "google_storage_bucket" "example" {
        name = "crossplane-example-${terraform.workspace}-${random_id.example.hex}"
      }
  writeConnectionSecretToRef:
    namespace: default
    name: terraform-workspace-example-inline

apiVersion: tf.crossplane.io/v1alpha1
kind: Workspace
metadata:
  name: example-remote
  annotations:
    crossplane.io/external-name: myworkspace
spec:
  forProvider:
    # Use any module source supported by terraform init -from-module.
    source: Remote
    module: https://github.com/crossplane/tf
    # Variables can be specified inline, or loaded from a ConfigMap or Secret.
    vars:
    - key: region
      value: us-west-1
    varFiles:
    - source: SecretKey
      secretKeyRef:
        namespace: default
        name: terraform
        key: example.tfvar.json
  # All Terraform outputs are written to the connection secret.
  writeConnectionSecretToRef:
    namespace: default
    name: terraform-workspace-example-inline

Installation

We highly encourage to use a declarative way of provider installation:

kubectl apply -f examples/install.yaml

Notice that in this example Provider resource is referencing ControllerConfig with debug enabled.

You can also setup the Terraform Provider using AWS IAM Roles for Service Accounts (IRSA). For more information, check out the example setup, the process is similar to what you would use for the provider-aws.

Private Git repository support

To securely propagate git credentials create a git-credentials secret in [git credentials store] format.

cat .git-credentials
https://<user>:<token>@github.com

kubectl create secret generic git-credentials --from-file=.git-credentials

Reference it in ProviderConfig.

apiVersion: tf.crossplane.io/v1alpha1
kind: ProviderConfig
metadata:
  name: default
spec:
  credentials:
  - filename: .git-credentials # use exactly this filename
    source: Secret
    secretRef:
      namespace: crossplane-system
      name: git-credentials
      key: .git-credentials
...

Standard .git-credentials filename is important to keep so provider-terraform controller will be able to automatically pick it up.

Terraform Output support

Non-sensitive outputs are mapped to the status.atProvider.outputs section as strings so they can be referenced by the Composition. Strings, numbers and booleans can be referenced directly in Compositions and can be used in the convert transform if type conversion is needed. Tuple and object outputs will be available in the corresponding JSON form. This is required because undefined object attributes are not specified in the Workspace CRD and so will be sanitized before the status is stored in the database.

That means that any output values required for use in the Composition must be published explicitly and individually, and they cannot be referenced inside a tuple or object.

For example, the following terraform outputs:

      output "string" {
        value = "bar"
        sensitive = false
      }
      output "number" {
        value = 1.9
        sensitive = false
      }
      output "object" {
        // This will be a JSON string - the key/value pairs are not accessible
        value = {"a": 3, "b": 2}
        sensitive = false
      }
      output "tuple" {
        // This will be a JSON string - the elements will not be accessible
        value = ["foo", "bar"]
        sensitive = false
      }
      output "bool" {
        value = false
        sensitive = false
      }
      output "sensitive" {
        value = "SENSITIVE"
        sensitive = true
      }

Appear in the corresponding outputs section as:

  status:
    atProvider:
      outputs:
        bool: "false"
        number: "1.9"
        object: '{"a":3,"b":2}'
        string: bar
        tuple: '["foo", "bar"]'

Note that the "sensitive" output is not included in status.atProvider.outputs

Terraform CLI Command Arguments

Additional arguments can be passed to the Terraform plan, apply, and destroy commands by specifying the planArgs, applyArgs and destroyArgs options.

For example:

apiVersion: tf.crossplane.io/v1alpha1
kind: Workspace
metadata:
  name: example-args
spec:
  forProvider:
    # Run the terraform init command with -upgrade=true to upgrade any stored providers
    initArgs:
      - -upgrade=true
    # Run the terraform plan command with the -parallelism=2 argument
    planArgs:
      - -parallelism=2
    # Run the terraform apply command with the -target=specificresource argument
    applyArgs:
      - -target=specificresource
    # Run the terraform destroy command with the -refresh=false argument
    destroyArgs:
      - -refresh=false
    # Use any module source supported by terraform init -from-module.
    source: Remote
    module: https://github.com/crossplane/tf
  # All Terraform outputs are written to the connection secret.
  writeConnectionSecretToRef:
    namespace: default
    name: terraform-workspace-example-inline

This will cause the terraform init command to be run with the "-upgrade=true" argument, the terraform plan command to be run with the -parallelism=2 argument, the terraform apply command to be run with the -target=specificresource argument, and the terraform destroy command to be run with the -refresh=false argument.

Note that by default the terraform init command is run with the "-input=false", and "-no-color" arguments, the terraform apply and destroy commands are run with the "-no-color", "-auto-approve", and "-input=false" arguments, and the terraform plan command is run with the "-no-color", "-input=false", and "-detailed-exitcode" arguments. Arguments specified in applyArgs, destroyArgs and planArgs will be added to these default arguments.

Custom Entrypoint for Terraform Invocation

In some cases, you might want to initialize and apply terraform in the subdirectory of the repository checkout. It is most relevant for the cases when your terraform modules contain inline relative paths.

To enable it, the Workspace spec has an optional Entrypoint field.

Consider this example:

apiVersion: tf.crossplane.io/v1alpha1
kind: Workspace
metadata:
  name: relative-path-test
spec:
  forProvider:
    module: git::https://github.com/ytsarev/provider-terraform-test-module.git
    source: Remote
    entrypoint: relative-path-iam
    vars:
      - key: iamRole
        value: relative-path-test

In this case, the whole repository will be checked out but terraform will be initialized in the relative-path-iam subdirectory with the module that contains relative path reference to the iam module located in the root of the tree.

module "relative-path-iam" {
  source  = "../iam"
  iamRole = var.iamRole
}

Provider Plugin Cache(enabled by default)

Provider Plugin Cache is enabled by default to speed up reconciliation.

In case you need to disable it, set optional pluginCache to false in ProviderConfig:

apiVersion: tf.crossplane.io/v1alpha1
kind: ProviderConfig
metadata:
  name: default
spec:
  pluginCache: false
...

Enable External Secret Support

If you need to store the sensitive output to an external secret store like Vault, you can specify the --enable-external-secret-stores flag to enable it:

apiVersion: pkg.crossplane.io/v1alpha1
kind: ControllerConfig
metadata:
  name: terraform-config
  labels:
    app: crossplane-provider-terraform
spec:
  image: crossplane/provider-terraform-controller:v0.3.0
  args:
    - -d
    - --enable-external-secret-stores
  metadata:
    annotations:
      vault.hashicorp.com/agent-inject: "true"
      vault.hashicorp.com/agent-inject-token: "true"
      vault.hashicorp.com/role: "crossplane"
      vault.hashicorp.com/agent-run-as-user: "2000"

Prepare a StoreConfig for Vault:

apiVersion: tf.crossplane.io/v1alpha1
kind: StoreConfig
metadata:
  name: vault
spec:
  type: Vault
  defaultScope: crossplane-system
  vault:
    server: http://vault.vault-system:8200
    mountPath: secret/
    version: v2
    auth:
      method: Token
      token:
        source: Filesystem
        fs:
          path: /vault/secrets/token

Specify it in spec.publishConnectionDetailsTo:

apiVersion: apiextensions.crossplane.io/v1
kind: Composition
metadata:
  name: ...
  labels:
    feature: ess
spec:
  compositeTypeRef:
    apiVersion: ...
    kind: ...
  resources:
    - name: foo
      base:
        apiVersion: tf.crossplane.io/v1alpha1
        kind: Workspace
        metadata:
          name: foo
        spec:
          forProvider:
            ...
          publishConnectionDetailsTo:
            name: bar
            configRef:
              name: vault

At Vault side configuration is also needed to allow the write operation, see example here for inspiration.

A concrete provider terraform use case is also available here.

Known limitations: