output_planfile: state has no output for key plan_content_json

[talset]

I tried to get the plan file locally with the following config:

resources:
  - name: tfstate
    type: terraform
    source:
      backend_config:
         path: "terraform.tfstate"
      backend_type: local
      env:
        AWS_ACCESS_KEY_ID: ((aws.access_key))
        AWS_DEFAULT_REGION: eu-west-1
        AWS_SECRET_ACCESS_KEY: ((aws.secret_key))
      env_name: test

     ...

resource_types:
  - name: terraform
    type: docker-image
    source:
      repository: ljfranklin/terraform-resource
      tag: 0.12.26-b11328e
jobs:
  - name: terraform-plan
    plan:
      - do:
          - get: git_terraform
            trigger: true
          - put: tfstate
            params:
              plan_only: true
              terraform_source: git_terraform/
            get_params:
              output_planfile: true

But after the end of the task with Successfully Ran Terraform Plan!
the put output say 2020/06/08 12:24:51 state has no output for key plan_content_json

[ljfranklin]

The local backend type isn't supported. Concourse requires that all state is persisted outside the container. Added an error message and docs to describe that here: db907cb. You'll have to use one of the other backend types that saves the statefile somewhere other than the local filesystem.

[talset]

@ljfranklin Agree that we need to have the tfstate persistent but since we have output_statefile and output_planfile available, a user can easily have a custom task to get/push the tfstate generated localy in the terraform resource.

It should be up to the user to give an external backend or provide terraform_source with the local tfstate. WDYT ?

[ljfranklin]

It should be up to the user to give an external backend or provide terraform_source with the local tfstate

Unfortunately no. To give more detail, a Concourse put cannot output any local files. The put must persist any required state somewhere outside the container. After each put a get automatically runs to download that state and output into a directory so that subsequent tasks can consume the outputs. There is no way for a Terraform put to write a planfile to terraform_source directory so that a subsequent task could read it.