remote
Note: We introduced the remote backend in Terraform v0.11.13 and Terraform Enterprise v201809-1. As of Terraform v1.1.0 and Terraform Enterprise v202201-1, we recommend using HCP Terraform's built-in cloud
integration instead. The cloud
option supports an improved user experience and more features, such as structured run output mode, which displays your plan and apply results in a human-readable format.
The remote backend is unique among all other Terraform backends because it can both store state snapshots and execute operations for HCP Terraform's CLI-driven run workflow. It used to be called an "enhanced" backend.
When using full remote operations, operations like terraform plan
or terraform apply
can be executed in Terraform
Cloud's run environment, with log output streaming to the local terminal. Remote plans and applies use variable values from the associated HCP Terraform workspace.
You can also use HCP Terraform with local operations, in which case only state is stored in the HCP Terraform backend.
Command Support
The remote backend supports the following Terraform commands:
apply
console
(supported in Terraform >= v0.11.12)destroy
fmt
get
graph
(supported in Terraform >= v0.11.12)import
(supported in Terraform >= v0.11.12)init
output
plan
providers
show
state
(supports all sub-commands: list, mv, pull, push, rm, show)taint
untaint
validate
version
workspace
Workspaces
The remote backend can work with either a single remote HCP Terraform workspace, or with multiple similarly-named remote workspaces (like networking-dev
and networking-prod
). The workspaces
block of the backend configuration
determines which mode it uses:
To use a single remote HCP Terraform workspace, set
workspaces.name
to the remote workspace's full name (likenetworking-prod
).To use multiple remote workspaces, set
workspaces.prefix
to a prefix used in all of the desired remote workspace names. For example, setprefix = "networking-"
to use HCP Terraform workspaces with names likenetworking-dev
andnetworking-prod
. This is helpful when mapping multiple Terraform CLI workspaces used in a single Terraform configuration to multiple HCP Terraform workspaces.
The backend configuration requires either name
or prefix
. Omitting both or
setting both results in a configuration error.
If previous state is present when you run terraform init
and the corresponding
remote workspaces are empty or absent, Terraform will create workspaces and
update the remote state accordingly. However, if your workspace requires variables or a specific version of Terraform for remote operations, we
recommend that you create your remote workspaces on HCP Terraform before
running any remote operations against them.
Workspace Names
Terraform uses shortened names without the common prefix to interact with workspaces on the command line. For example, if prefix = "networking-"
, use terraform workspace select prod
to switch to the Terraform CLI workspace prod
within the current configuration. However, remote Terraform operations such as plan
and apply
for that Terraform CLI workspace will take place in the HCP Terraform workspace networking-prod
.
Because of this, the terraform.workspace
interpolation expression produces different results depending on whether a remote workspace is configured to perform operations locally or remotely. For example, in a remote workspace called networking-prod
created with prefix = "networking-"
the expression produces the following:
- For local operations,
terraform.workspace
=prod
- For remote operations,
terraform.workspace
=networking-prod
Prior to Terraform version 1.1.0, HCP Terraform workspaces used only the single default
Terraform CLI workspace internally. So if a Terraform configuration used terraform.workspace
to return dev
or prod
, remote runs in HCP Terraform would always evaluate it as default
, regardless of
which workspace you set with the terraform workspace select
command. Therefore, we do not recommend using terraform.workspace
in Terraform configurations that use Terraform 1.0.x or earlier and run remote operations against HCP Terraform workspaces.
Determining Run Environment
If you need to determine whether a run is local or remote in your Terraform configuration, we recommend using HCP Terraform run environment variables. The example below uses HCP_TERRAFORM_RUN_ID
.
Example Configurations
Note: We recommend omitting the token from the configuration, and instead using
terraform login
or manually configuring
credentials
in the CLI config file.
Basic Configuration
Using CLI Input
Backend configuration file:
Running terraform init
with the backend file:
Data Source Configuration
Configuration Variables
Warning: We recommend using environment variables to supply credentials and other sensitive data. If you use -backend-config
or hardcode these values directly in your configuration, Terraform will include these values in both the .terraform
subdirectory and in plan files. Refer to Credentials and Sensitive Data for details.
The following configuration options are supported:
hostname
- (Optional) The remote backend hostname to connect to. Defaults to app.terraform.io.organization
- (Required) The name of the organization containing the targeted workspace(s).token
- (Optional) The token used to authenticate with the remote backend. We recommend omitting the token from the configuration, and instead usingterraform login
or manually configuringcredentials
in the CLI config file.workspaces
- (Required) A block specifying which remote workspace(s) to use. Theworkspaces
block supports the following keys:name
- (Optional) The full name of one remote workspace. When configured, only the default workspace can be used. This option conflicts withprefix
.prefix
- (Optional) A prefix used in the names of one or more remote workspaces, all of which can be used with this configuration. The full workspace names are used in HCP Terraform, and the short names (minus the prefix) are used on the command line for Terraform CLI workspaces. If omitted, only the default workspace can be used. This option conflicts withname
.
Note: You must use the name
key when configuring a terraform_remote_state
data source that retrieves state from another HCP Terraform workspace. The prefix
key is only
intended for use when configuring an instance of the remote backend.
Command Line Arguments
For configurations that include a backend "remote"
block, commands that
make local modifications to Terraform state and then push them back up to
the remote workspace accept the following option to modify that behavior:
-ignore-remote-version
- Override checking that the local and remote Terraform versions agree, making an operation proceed even when there is a mismatch.Normally state-modification operations require using a local version of Terraform CLI which is compatible with the Terraform version selected for the remote workspace as part of its settings. This is to avoid the local operation creating a new state snapshot which the workspace's remote execution environment would then be unable to decode.
Overriding this check can result in an HCP Terraform workspace that is no longer able to complete remote operations, so we recommend against using this option.
Excluding Files from Upload with .terraformignore
Version note: .terraformignore
support was added in Terraform 0.12.11.
When executing a remote plan
or apply
in a CLI-driven run,
an archive of your configuration directory is uploaded to HCP Terraform. You can define
paths to ignore from upload via a .terraformignore
file at the root of your configuration directory. If this file is not present, the archive will exclude the following by default:
.git/
directories.terraform/
directories (exclusive of.terraform/modules
)
The .terraformignore
file can include rules as one would include in a
.gitignore
file
- Comments (starting with
#
) or blank lines are ignored - End a pattern with a forward slash
/
to specify a directory - Negate a pattern by starting it with an exclamation point
!
Note that unlike .gitignore
, only the .terraformignore
at the root of the configuration
directory is considered.