tfc-apply-replace 1.0.1

Tool to perform dynamic terraform applies w/replacement on TFC using resource name keywords.

tfc-ws-apply-replace

What does it do?

This script targets a TFC workspace and runs the equivalent of a terraform apply -replace=[] on a dynamic set of resources determined by a provided keyword

Why?

In certain TFC implementations, people will use a tightly controlled organization and workspace to create child organizations and workspaces. This is commonly referred to as the workspace vending machine approach. In this model, the parent workspace is also responsible for managing the child workspace's credentials in the form of sensitive workspace variables. However, setting those variables once and never rotating them introduces some risk. Using this script, you can schedule the rotation of those credentials (or any other type of resource).

Usage

1. Set an environment variable TFC_TOKEN equal to a valid TFC token with adequate permissions
2. run python3 rotate.py <org name> <workspace ID> <resource name filter> where:

• org name: is the name of the org that contains your workspace
• workspace ID: is the ws-xxxxxxxxxxxxxxxx formatted ID of your target workspace
• resource name filter: is the beginning of the resource name string

Example

My parent workspace creates a child workspace and an IAM user specific to the child workspace. It then generates aws_iam_access_key credentials and assigns them to the child workspace as a sensitive variable:

resource "aws_iam_user" "iam_user" {
  name = "tfc_workspace_${tfe_workspace.workspace.name}"
}

resource "aws_iam_user_policy_attachment" "iam_user_policy_attach" {
  user       = aws_iam_user.iam_user.name
  policy_arn = var.workspace_iam_policy
}

resource "aws_iam_access_key" "workspace_creds" {
  user = aws_iam_user.iam_user.name
}

resource "tfe_variable" "aws_access_key_id" {
  key          = "AWS_ACCESS_KEY_ID"
  category     = "env"
  value        = aws_iam_access_key.workspace_creds.id
  workspace_id = tfe_workspace.workspace.id
}

resource "tfe_variable" "aws_secret_access_key" {
  key          = "AWS_SECRET_ACCESS_KEY"
  category     = "env"
  sensitive    = true
  value        = aws_iam_access_key.workspace_creds.secret
  workspace_id = tfe_workspace.workspace.id
}

If I wanted to replace/rotate the resource aws_iam_access_key (and by relation any resource that depends on it, i.e. the tfe_variable.aws_secret_access_key resource), I would use the resources given name workspace_creds

python3 rotate.py some-org ws-xxxxxxxxxxxxxxxx workspace_creds
-------------------------------------
TFC Organization Name: some-org
TFC Workspace ID: ws-xxxxxxxxxxxxxxxx
Resource Name FIlter: workspace_creds
-------------------------------------

Found 3 resources whose name(s) start with the keyword:

    module.ws_vending_machine["djs-ec2"].aws_iam_access_key.workspace_creds
    module.ws_vending_machine["djs-iam"].aws_iam_access_key.workspace_creds
    module.ws_vending_machine["djs-s3"].aws_iam_access_key.workspace_creds

Triggering a 'terraform apply -replace=' to rotate the found resources...

Link to run: https://app.terraform.io/app/some-org/workspaces/some-workspace/runs/run-xxxxxxxxxxxxxxxx

Status: pending...
Status: plan_queued...
Status: planning...
Status: cost_estimating...
Status: applying...
Status: applied...

While this workspace manages 23 resources in total, you can see below the script targeted only the resources needed to rotate the downstream workspace's access keys: This script can be scheduled to rotate the credentials limiting the risk of bad terraform code leaking the sensitive variables.