» Providers

While resources are the primary construct in the Terraform language, the behaviors of resources rely on their associated resource types, and these types are defined by providers.

Each provider offers a set of named resource types, and defines for each resource type which arguments it accepts, which attributes it exports, and how changes to resources of that type are actually applied to remote APIs.

Most of the available providers correspond to one cloud or on-premises infrastructure platform, and offer resource types that correspond to each of the features of that platform.

Providers usually require some configuration of their own to specify endpoint URLs, regions, authentication settings, and so on. All resource types belonging to the same provider will share the same configuration, avoiding the need to repeat this common information across every resource declaration.

» Provider Configuration

A provider configuration is created using a provider block:

provider "google" {
  project = "acme-app"
  region  = "us-central1"

The name given in the block header ("google" in this example) is the name of the provider to configure. Terraform associates each resource type with a provider by taking the first word of the resource type name (separated by underscores), and so the "google" provider is assumed to be the provider for the resource type name google_compute_instance.

The body of the block (between { and }) contains configuration arguments for the provider itself. Most arguments in this section are specified by the provider itself; in this example both project and region are specific to the google provider.

The configuration arguments defined by the provider may be assigned using expressions, which can for example allow them to be parameterized by input variables. However, since provider configurations must be evaluated in order to perform any resource type action, provider configurations may refer only to values that are known before the configuration is applied. In particular, avoid referring to attributes exported by other resources unless their values are specified directly in the configuration.

There are also two "meta-arguments" that are defined by Terraform itself and available for all provider blocks:

Unlike many other objects in the Terraform language, a provider block may be omitted if its contents would otherwise be empty. Terraform assumes an empty default configuration for any provider that is not explicitly configured.

» Initialization

Each time a new provider is added to configuration -- either explicitly via a provider block or by adding a resource from that provider -- Terraform must initialize the provider before it can be used. Initialization downloads and installs the provider's plugin so that it can later be executed.

Provider initialization is one of the actions of terraform init. Running this command will download and initialize any providers that are not already initialized.

Providers downloaded by terraform init are only installed for the current working directory; other working directories can have their own installed provider versions.

Note that terraform init cannot automatically download providers that are not distributed by HashiCorp. See Third-party Plugins below for installation instructions.

For more information, see the terraform init command.

» version: Provider Versions

Providers are plugins released on a separate rhythm from Terraform itself, and so they have their own version numbers. For production use, you should constrain the acceptable provider versions via configuration, to ensure that new versions with breaking changes will not be automatically installed by terraform init in future.

When terraform init is run without provider version constraints, it prints a suggested version constraint string for each provider:

The following providers do not have any version constraints in configuration,
so the latest version was installed.

To prevent automatic upgrades to new major versions that may contain breaking
changes, it is recommended to add version = "..." constraints to the
corresponding provider blocks in configuration, with the constraint strings
suggested below.

* provider.aws: version = "~> 1.0"

To constrain the provider version as suggested, add the version meta-argument to the provider configuration block:

provider "aws" {
  version = "~> 1.0"

  region     = "us-east-1"

This meta-argument applies to all providers. The terraform providers command can be used to view the specified version constraints for all providers used in the current configuration.

The version argument value may either be a single explicit version or a version constraint string. Constraint strings use the following syntax to specify a range of versions that are acceptable:

  • >= 1.2.0: version 1.2.0 or newer
  • <= 1.2.0: version 1.2.0 or older
  • ~> 1.2.0: any non-beta version >= 1.2.0 and < 1.3.0, e.g. 1.2.X
  • ~> 1.2: any non-beta version >= 1.2.0 and < 2.0.0, e.g. 1.X.Y
  • >= 1.0.0, <= 2.0.0: any version between 1.0.0 and 2.0.0 inclusive

When terraform init is re-run with providers already installed, it will use an already-installed provider that meets the constraints in preference to downloading a new version. To upgrade to the latest acceptable version of each provider, run terraform init -upgrade. This command also upgrades to the latest versions of all Terraform modules.

» alias: Multiple Provider Instances

You can optionally define multiple configurations for the same provider, and select which one to use on a per-resource or per-module basis. The primary reason for this is to support multiple regions for a cloud platform; other examples include targeting multiple Docker hosts, multiple Consul hosts, etc.

To include multiple configurations for a given provider, include multiple provider blocks with the same provider name, but set the alias meta-argument to an alias name to use for each additional configuration. For example:

# The default provider configuration
provider "aws" {
  region = "us-east-1"

# Additional provider configuration for west coast region
provider "aws" {
  alias  = "west"
  region = "us-west-2"

The provider block without alias set is known as the default provider configuration. When alias is set, it creates an additional provider configuration. For providers that have no required configuration arguments, the implied empty configuration is considered to be the default provider configuration.

» Referring to Alternate Providers

When Terraform needs the name of a provider configuration, it always expects a reference of the form <PROVIDER NAME>.<ALIAS>. In the example above, aws.west would refer to the provider with the us-west-2 region.

These references are special expressions. Like references to other named entities (for example, var.image_id), they aren't strings and don't need to be quoted. But they are only valid in specific meta-arguments of resource, data, and module blocks, and can't be used in arbitrary expressions.

» Selecting Alternate Providers

By default, resources use a default provider configuration inferred from the first word of the resource type name. For example, a resource of type aws_instance uses the default (un-aliased) aws provider configuration unless otherwise stated.

To select an aliased provider for a resource or data source, set its provider meta-argument to a <PROVIDER NAME>.<ALIAS> reference:

resource "aws_instance" "foo" {
  provider = aws.west

  # ...

To select aliased providers for a child module, use its providers meta-argument to specify which aliased providers should be mapped to which local provider names inside the module:

module "aws_vpc" {
  source = "./aws_vpc"
  providers = {
    aws = aws.west

Modules have some special requirements when passing in providers; see Providers within Modules for more details. In most cases, only root modules should define provider configurations, with all child modules obtaining their provider configurations from their parents.

» Third-party Plugins

Anyone can develop and distribute their own Terraform providers. (See Writing Custom Providers for more about provider development.) These third-party providers must be manually installed, since terraform init cannot automatically download them.

Install third-party providers by placing their plugin executables in the user plugins directory. The user plugins directory is in one of the following locations, depending on the host operating system:

Operating system User plugins directory
Windows %APPDATA%\terraform.d\plugins
All other systems ~/.terraform.d/plugins

Once a plugin is installed, terraform init can initialize it normally.

Providers distributed by HashiCorp can also go in the user plugins directory. If a manually installed version meets the configuration's version constraints, Terraform will use it instead of downloading that provider. This is useful in airgapped environments and when testing pre-release provider builds.

» Plugin Names and Versions

The naming scheme for provider plugins is terraform-provider-<NAME>_vX.Y.Z, and Terraform uses the name to understand the name and version of a particular provider binary.

If multiple versions of a plugin are installed, Terraform will use the newest version that meets the configuration's version constraints.

Third-party plugins are often distributed with an appropriate filename already set in the distribution archive, so that they can be extracted directly into the user plugins directory.

» OS and Architecture Directories

Terraform plugins are compiled for a specific operating system and architecture, and any plugins in the root of the user plugins directory must be compiled for the current system.

If you use the same plugins directory on multiple systems, you can install plugins into subdirectories with a naming scheme of <OS>_<ARCH> (for example, darwin_amd64). Terraform uses plugins from the root of the plugins directory and from the subdirectory that corresponds to the current system, ignoring other subdirectories.

Terraform's OS and architecture strings are the standard ones used by the Go language. The following are the most common:

» Provider Plugin Cache

By default, terraform init downloads plugins into a subdirectory of the working directory so that each working directory is self-contained. As a consequence, if you have multiple configurations that use the same provider then a separate copy of its plugin will be downloaded for each configuration.

Given that provider plugins can be quite large (on the order of hundreds of megabytes), this default behavior can be inconvenient for those with slow or metered Internet connections. Therefore Terraform optionally allows the use of a local directory as a shared plugin cache, which then allows each distinct plugin binary to be downloaded only once.

To enable the plugin cache, use the plugin_cache_dir setting in the CLI configuration file. For example:

# (Note that the CLI configuration file is _not_ the same as the .tf files
#  used to configure infrastructure.)

plugin_cache_dir = "$HOME/.terraform.d/plugin-cache"

This directory must already exist before Terraform will cache plugins; Terraform will not create the directory itself.

Please note that on Windows it is necessary to use forward slash separators (/) rather than the conventional backslash (\) since the configuration file parser considers a backslash to begin an escape sequence.

Setting this in the configuration file is the recommended approach for a persistent setting. Alternatively, the TF_PLUGIN_CACHE_DIR environment variable can be used to enable caching or to override an existing cache directory within a particular shell session:

export TF_PLUGIN_CACHE_DIR="$HOME/.terraform.d/plugin-cache"

When a plugin cache directory is enabled, the terraform init command will still access the plugin distribution server to obtain metadata about which plugins are available, but once a suitable version has been selected it will first check to see if the selected plugin is already available in the cache directory. If so, the already-downloaded plugin binary will be used.

If the selected plugin is not already in the cache, it will be downloaded into the cache first and then copied from there into the correct location under your current working directory.

When possible, Terraform will use hardlinks or symlinks to avoid storing a separate copy of a cached plugin in multiple directories. At present, this is not supported on Windows and instead a copy is always created.

The plugin cache directory must not be the third-party plugin directory or any other directory Terraform searches for pre-installed plugins, since the cache management logic conflicts with the normal plugin discovery logic when operating on the same directory.

Please note that Terraform will never itself delete a plugin from the plugin cache once it's been placed there. Over time, as plugins are upgraded, the cache directory may grow to contain several unused versions which must be manually deleted.