This is one of a suite of Terraform related actions - find them at dflook/terraform-github-actions.
This action applies a Terraform plan.
The default behaviour is to apply the plan that has been added to a PR using the dflook/terraform-plan
action.
If the plan is not found or has changed, then the apply
action will fail.
This is to ensure that the action only applies changes that have been reviewed by a human.
You can instead set auto_approve: true
which will generate a plan and apply it immediately, without looking for a plan attached to a PR.
This a demo of the process for apply a Terraform change using the dflook/terraform-plan
and dflook/terraform-apply
actions.
To make best use of this action, require that the plan is always reviewed before merging the PR to approve. You can enforce this in github by going to the branch settings for the repo and enable protection for the main branch:
- Enable 'Require pull request reviews before merging'
- Check 'Dismiss stale pull request approvals when new commits are pushed'
- Enable 'Require status checks to pass before merging', and select the job that runs the plan.
- Enable 'Require branches to be up to date before merging'
These input values must be the same as any dflook/terraform-plan
for the same configuration. (unless auto_approve: true)
-
path
Path to the Terraform root module to apply
- Type: string
- Optional
- Default: The action workspace
-
workspace
Terraform workspace to run the apply in
- Type: string
- Optional
- Default:
default
-
label
A friendly name for the environment the Terraform configuration is for. This will be used in the PR comment for easy identification.
It must be the same as the
label
used in the correspondingdflook/terraform-plan
action.- Type: string
- Optional
-
variables
Variables to set for the Terraform plan. This should be valid Terraform syntax - like a variable definition file.
with: variables: | image_id = "${{ secrets.AMI_ID }}" availability_zone_names = [ "us-east-1a", "us-west-1c", ]
Variables set here override any given in
var_file
s.- Type: string
- Optional
-
var_file
List of tfvars files to use, one per line. Paths should be relative to the GitHub Actions workspace
with: var_file: | common.tfvars prod.tfvars
- Type: string
- Optional
-
backend_config
List of Terraform backend config values, one per line.
with: backend_config: token=${{ secrets.BACKEND_TOKEN }}
- Type: string
- Optional
-
backend_config_file
List of Terraform backend config files to use, one per line. Paths should be relative to the GitHub Actions workspace
with: backend_config_file: prod.backend.tfvars
- Type: string
- Optional
-
replace
List of resources to replace, one per line.
with: replace: | kubernetes_secret.tls_cert_public kubernetes_secret.tls_cert_private
- Type: string
- Optional
-
target
List of resources to apply, one per line. The apply operation will be limited to these resources and their dependencies.
with: target: | kubernetes_secret.tls_cert_public kubernetes_secret.tls_cert_private
- Type: string
- Optional
-
destroy
Set to
true
to destroy all resources.This generates and applies a plan in destroy mode.
- Type: boolean
- Optional
- Default: false
-
plan_path
Path to a plan file to apply. This would have been generated by a previous
dflook/terraform-plan
action.The default behaviour when this is not set is to generate a plan from the current configuration and compare it to the plan attached to the PR comment. If it is logically the same, the plan will be applied.
When this is set to a plan file, the plan will not be generated again. If it is the exact same plan as the one attached to the PR comment, it will be applied. This will be faster than generating a new plan.
There are downsides to applying a stored plan:
- The plan may contain sensitive information so must be stored securely, possibly outside of GitHub.
- It does not account for any changes that have occurred since it was generated, and may no longer be correct.
- Plans must be generated and applied in strict order. Multiple open PRs will cause conflicts if they are applied out of order.
- Plans are not portable between platforms.
- Terraform and provider versions must match between the plan generation and apply.
When
auto_approve
is set totrue
, the plan will be applied without checking if it is the same as the one attached to the PR comment.- Type: string
- Optional
-
auto_approve
When set to
true
, plans are always applied.The default is
false
, which requires plans to have been added to a pull request comment.- Type: bool
- Optional
- Default: false
-
parallelism
Limit the number of concurrent operations
- Type: number
- Optional
- Default: The terraform default (10)
-
json_plan_path
This is the path to the generated plan in JSON Output Format The path is relative to the Actions workspace.
This won't be set if the backend type is
remote
- Terraform does not support saving remote plans. -
text_plan_path
This is the path to the generated plan in a human-readable format. The path is relative to the Actions workspace. This won't be set if
auto_approve
is true while using aremote
backend. -
failure-reason
When the job outcome is
failure
, this output may be set. The value may be one of:apply-failed
- The Terraform apply operation failed.plan-changed
- The approved plan is no longer accurate, so the apply will not be attempted.state-locked
- The Terraform state lock could not be obtained because it was already locked.
If the job fails for any other reason this will not be set. This can be used with the Actions expression syntax to conditionally run steps.
-
lock-info
When the job outcome is
failure
and the failure-reason isstate-locked
, this output will be set.It is a json object containing any available state lock information and typically has the form:
{ "ID": "838fbfde-c5cd-297f-84a4-d7578b4a4880", "Path": "terraform-github-actions/test-unlock-state", "Operation": "OperationTypeApply", "Who": "root@e9d43b0c6478", "Version": "1.3.7", "Created": "2023-01-28 00:16:41.560904373 +0000 UTC", "Info": "" }
-
run_id
If the root module uses the
remote
orcloud
backend in remote execution mode, this output will be set to the remote run id.- Type: string
-
Terraform Outputs
An action output will be created for each output of the Terraform configuration.
For example, with the Terraform config:
output "service_hostname" { value = "example.com" }
Running this action will produce a
service_hostname
output with the same value. See terraform-output for details.
-
GITHUB_TOKEN
The GitHub authorization token to use to fetch an approved plan from a PR. This must belong to the same user/app as the token used by the terraform-plan action. The token provided by GitHub Actions can be used - it can be passed by using the
${{ secrets.GITHUB_TOKEN }}
expression, e.g.env: GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
The token provided by GitHub Actions has default permissions at GitHub's whim. You can see what it is for your repo under the repo settings.
The minimum permissions are
pull-requests: write
. It will also likely needcontents: read
so the job can checkout the repo.You can also use any other App token that has
pull-requests: write
permission.You can use a fine-grained Personal Access Token which has repository permissions:
- Read access to metadata
- Read and Write access to pull requests
You can also use a classic Personal Access Token which has the
repo
scope.- Type: string
- Optional
-
TERRAFORM_ACTIONS_GITHUB_TOKEN
When this is set it is used instead of
GITHUB_TOKEN
, with the same behaviour. The GitHub Terraform provider also uses theGITHUB_TOKEN
so this can be used to make the github actions and the Terraform provider use different tokens.- Type: string
- Optional
-
GITHUB_DOT_COM_TOKEN
This is used to specify a token for GitHub.com when the action is running on a GitHub Enterprise instance. This is only used for downloading OpenTofu binaries from GitHub.com. If this is not set, an unauthenticated request will be made to GitHub.com to download the binary, which may be rate limited.
- Type: string
- Optional
-
TERRAFORM_CLOUD_TOKENS
API tokens for cloud hosts, of the form
<host>=<token>
. Multiple tokens may be specified, one per line. These tokens may be used with theremote
backend and for fetching required modules from the registry.e.g:
env: TERRAFORM_CLOUD_TOKENS: app.terraform.io=${{ secrets.TF_CLOUD_TOKEN }}
With other registries:
env: TERRAFORM_CLOUD_TOKENS: | app.terraform.io=${{ secrets.TF_CLOUD_TOKEN }} terraform.example.com=${{ secrets.TF_REGISTRY_TOKEN }}
- Type: string
- Optional
-
TERRAFORM_SSH_KEY
A SSH private key that Terraform will use to fetch git module sources.
This should be in PEM format.
For example:
env: TERRAFORM_SSH_KEY: ${{ secrets.TERRAFORM_SSH_KEY }}
- Type: string
- Optional
-
TERRAFORM_PRE_RUN
A set of commands that will be ran prior to
terraform init
. This can be used to customise the environment before running Terraform.The runtime environment for these actions is subject to change in minor version releases. If using this environment variable, specify the minor version of the action to use.
The runtime image is currently based on
debian:bullseye
, with the command run usingbash -xeo pipefail
.For example:
env: TERRAFORM_PRE_RUN: | # Install latest Azure CLI curl -skL https://aka.ms/InstallAzureCLIDeb | bash # Install postgres client apt-get install -y --no-install-recommends postgresql-client
- Type: string
- Optional
-
TERRAFORM_HTTP_CREDENTIALS
Credentials that will be used for fetching modules sources with
git::http://
,git::https://
,http://
&https://
schemes.Credentials have the format
<host>=<username>:<password>
. Multiple credentials may be specified, one per line.Each credential is evaluated in order, and the first matching credentials are used.
Credentials that are used by git (
git::http://
,git::https://
) allow a path after the hostname. Paths are ignored byhttp://
&https://
schemes. For git module sources, a credential matches if each mentioned path segment is an exact match.For example:
env: TERRAFORM_HTTP_CREDENTIALS: | example.com=dflook:${{ secrets.HTTPS_PASSWORD }} github.com/dflook/terraform-github-actions.git=dflook-actions:${{ secrets.ACTIONS_PAT }} github.com/dflook=dflook:${{ secrets.DFLOOK_PAT }} github.com=graham:${{ secrets.GITHUB_PAT }}
- Type: string
- Optional
When applying a plan from a PR comment (auto_approve
is the default of false
), the workflow can be triggered by the following events:
- pull_request
- pull_request_review_comment
- pull_request_target
- pull_request_review
- issue_comment, if the comment is on a PR (see below)
- push, if the pushed commit came from a PR (see below)
- repository_dispatch, if the client payload includes the pull_request url (see below)
When auto_approve
is set to true
, the workflow can be triggered by any event.
This event triggers workflows when a comment is made in a Issue or a Pull Request. Since running the action will only work in the context of a PR, the workflow should check that the comment is on a PR before running.
Also take care to checkout the PR ref.
jobs:
apply:
if: ${{ github.event.issue.pull_request }}
runs-on: ubuntu-latest
env:
GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
steps:
- name: Checkout
uses: actions/checkout@v4
with:
ref: refs/pull/${{ github.event.issue.number }}/merge
- name: terraform apply
uses: dflook/terraform-apply@v1
with:
path: my-terraform-config
The pushed commit must have come from a Pull Request. Typically this is used to trigger a workflow that runs on the main branch after a PR has been merged.
This event can be used to trigger a workflow from another workflow. The client payload must include the pull_request api url of where the plan PR comment can be found.
A minimal example payload looks like:
{
"pull_request": {
"url": "https://api.github.com/repos/dflook/terraform-github-actions/pulls/1"
}
}
This example workflow runs for every push to main. If the commit came from a PR that has been merged, applies the plan from the PR.
name: Apply
on:
push:
branches:
- main
permissions:
contents: read
pull-requests: write
jobs:
apply:
runs-on: ubuntu-latest
name: Apply approved plan
env:
GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
steps:
- name: Checkout
uses: actions/checkout@v4
- name: terraform apply
uses: dflook/terraform-apply@v1
with:
path: my-terraform-config
This example workflow runs for every push to main. Changes are planned and applied.
name: Apply
on:
push:
branches:
- main
jobs:
apply:
runs-on: ubuntu-latest
name: Apply Terraform
steps:
- name: Checkout
uses: actions/checkout@v4
- name: terraform apply
uses: dflook/terraform-apply@v1
with:
path: my-terraform-config
auto_approve: true
This example workflow runs every morning and updates a TLS certificate if necessary.
name: Rotate certs
on:
schedule:
- cron: "0 8 * * *"
jobs:
apply:
runs-on: ubuntu-latest
name: Rotate certs
steps:
- name: Checkout
uses: actions/checkout@v4
- name: terraform apply
uses: dflook/terraform-apply@v1
with:
path: my-terraform-config
auto_approve: true
target: |
kubernetes_secret.tls_cert_public
kubernetes_secret.tls_cert_private
This workflow applies a plan on demand, triggered by someone
commenting terraform apply
on the PR. The plan is taken
from an existing comment generated by the dflook/terraform-plan
action.
name: Terraform Apply
on: [issue_comment]
jobs:
apply:
if: ${{ github.event.issue.pull_request && contains(github.event.comment.body, 'terraform apply') }}
runs-on: ubuntu-latest
name: Apply Terraform plan
env:
GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
steps:
- name: Checkout
uses: actions/checkout@v4
with:
ref: refs/pull/${{ github.event.issue.number }}/merge
- name: Terraform apply
uses: dflook/terraform-apply@v1
with:
path: my-terraform-config
This example retries the terraform apply operation if it fails.
name: Apply plan
on:
push:
branches:
- main
permissions:
contents: read
pull-requests: write
jobs:
plan:
runs-on: ubuntu-latest
name: Apply Terraform plan
env:
GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
steps:
- name: Checkout
uses: actions/checkout@v4
- name: terraform apply
uses: dflook/terraform-apply@v1
continue-on-error: true
id: first_try
with:
path: terraform
- name: Retry failed apply
uses: dflook/terraform-apply@v1
if: ${{ steps.first_try.outputs.failure-reason == 'apply-failed' }}
with:
path: terraform
auto_approve: true