Installing Software in the Run Environment
Terraform relies on provider plugins to manage resources. In most cases, Terraform can automatically download the required plugins, but there are cases where plugins must be managed explicitly.
In rare cases, it might also be necessary to install extra software on the Terraform worker, such as a configuration management tool or cloud CLI.
Installing Terraform Providers
The mechanics of provider installation changed in Terraform 0.13, thanks to the introduction of the Terraform Registry for providers which allows custom and community providers to be installed via terraform init
. Prior to Terraform 0.13, Terraform could only automatically install providers distributed by HashiCorp.
Terraform 0.13 and later
Providers From the Terraform Registry
The Terraform Registry allows anyone to publish and distribute providers which can be automatically downloaded and installed via terraform init
.
Terraform Enterprise instances must be able to access registry.terraform.io
to use providers from the public registry; otherwise, you can install providers using the terraform-bundle
tool.
In-House Providers
If you have a custom provider that you'd rather not publish in the public Terraform Registry, you have a few options:
Add the provider binary to the VCS repo (or manually-uploaded configuration version). Place the compiled
linux_amd64
version of the plugin atterraform.d/plugins/<SOURCE HOST>/<SOURCE NAMESPACE>/<PLUGIN NAME>/<VERSION>/linux_amd64
, relative to the root of the directory.The source host and namespace will need to match the source given in the
required_providers
block within the configuration, but can otherwise be arbitrary identifiers. For instance, if yourrequired_providers
block looks like this:HCP Terraform will be able to use your compiled provider if you place it at
terraform.d/plugins/my-host/my-namespace/custom/1.0.0/linux_amd64/terraform-provider-custom
.Use a privately-owned provider registry service which implements the provider registry protocol to distribute custom providers. Be sure to include the full source address, including the hostname, when referencing providers.
Terraform Enterprise only: Use the
terraform-bundle
tool to add custom providers.
Note: Using a network mirror to host custom providers for installation is not currently supported in HCP Terraform, since the network mirror cannot be activated without a provider_installation
block in the CLI configuration file.
Terraform 0.12 and earlier
Providers Distributed by HashiCorp
HCP Terraform can automatically install providers distributed by HashiCorp. Terraform Enterprise instances can do this as well as long as they can access releases.hashicorp.com
.
If that isn't feasible due to security requirements, you can manually install providers. Use the terraform-bundle
tool to build a custom version of Terraform that includes the necessary providers, and configure your workspaces to use that bundled version.
Custom and Community Providers
To use community providers or your own custom providers with Terraform versions prior to 0.13, you must install them yourself.
There are two ways to accomplish this:
Add the provider binary to the VCS repo (or manually-uploaded configuration version) for any workspace that uses it. Place the compiled
linux_amd64
version of the plugin atterraform.d/plugins/linux_amd64/<PLUGIN NAME>
(as a relative path from the root of the working directory). The plugin name should follow the naming scheme and the plugin file must have read and execute permissions. (Third-party plugins are often distributed with an appropriate filename already set in the distribution archive.)You can add plugins directly to a configuration repo, or you can add them as Git submodules and symlink the executable files into
terraform.d/plugins/
. Submodules are a good choice when many workspaces use the same custom provider, since they keep your repos smaller. If using submodules, enable the "Include submodules on clone" setting on any affected workspace.Terraform Enterprise only: Use the
terraform-bundle
tool to add custom providers to a custom Terraform version. This keeps custom providers out of your configuration repos entirely, and is easier to update when many workspaces use the same provider.
Installing Additional Tools
Avoid Installing Extra Software
Whenever possible, don't install software on the worker. There are a number of reasons for this:
- Provisioners are a last resort in Terraform; they greatly increase the risk of creating unknown states with unmanaged and partially-managed infrastructure, and the
local-exec
provisioner is especially hazardous. The Terraform CLI docs on provisioners explain the hazards in more detail, with more information about the preferred alternatives. (In summary: use Packer, use cloud-init, try to make your infrastructure more immutable, and always prefer real provider features.) - We don't guarantee the stability of the operating system on the Terraform build workers. It's currently the latest version of Ubuntu LTS, but we reserve the right to change that at any time.
- The build workers are disposable and are destroyed after each use, which makes managing extra software even more complex than when running Terraform CLI in a persistent environment. Custom software must be installed on every run, which also increases run times.
Only Install Standalone Binaries
HCP Terraform does not allow you to elevate a command's permissions with sudo
during Terraform runs. This means you cannot install packages using the worker OS's normal package management tools. However, you can install and execute standalone binaries in Terraform's working directory.
You have two options for getting extra software into the configuration directory:
Include it in the configuration repository as a submodule. (Make sure the workspace is configured to clone submodules.)
Use
local-exec
to download it withcurl
. For example:
When downloading software with local-exec
, try to associate the provisioner block with the resource(s) that the software will interact with. If you use a null resource with a local-exec
provisioner, you must ensure it can be properly configured with triggers. Otherwise, a null resource with the local-exec
provisioner will only install software on the initial run where the null_resource
is created. The null_resource
will not be automatically recreated in subsequent runs and the related software won't be installed, which may cause runs to encounter errors.
Note: Terraform Enterprise instances can be configured to allow sudo
commands during Terraform runs. However, even when sudo
is allowed, using the worker OS's package tools during runs is still usually a bad idea. You will have a much better experience if you can move your provisioner actions into a custom provider or an immutable machine image.