Terramate and Terraform: Managing Cross-Stack Variable Propagation

Answer

Use Terraform’s native terraform_remote_state data source to read outputs from a base stack, not Terramate template variables. Terramate should only orchestrate file generation; passing values through its variables creates tight coupling and can cause circular dependencies during terramate generate.

Confirmed Facts

• Terramate v0.x generates Terraform configuration but does not manage state.
• Template variables are resolved before Terraform runs, so a downstream stack that reads a variable from an upstream stack creates a build‑time dependency that can loop.
• The terraform_remote_state data source reads the state after terraform init, decoupling generation from consumption.

Steps to Implement (only what is needed for this case)

1. Declare outputs in the base stack’s outputs.tf (or wherever you define them).
2. Configure a remote backend (e.g., S3, Consul, Terraform Cloud) for the base stack and note the state path/workspace.
3. In each downstream stack, add a terraform_remote_state data source:

   data "terraform_remote_state" "network" {
     backend = "s3"
     config = {
       bucket = "my-tfstate-bucket"
       key    = "network/terraform.tfstate"
       region = "us-east-1"
     }
   }

4. Reference the needed output:

   locals {
     vpc_id = data.terraform_remote_state.network.outputs.vpc_id
   }

5. Ensure generation order: add a depends_on (or use Terramate’s stack ordering) in the downstream stack’s terramate.tm.hcl to list the base stack, so terramate generate processes the base first.
6. Run the workflow:

   terramate generate
   terraform init
   terraform plan

   Verify that the downstream plan shows the expected values without errors.

When to Ask for a Missing Detail

If you are using a Terramate version where depends_on semantics have changed (e.g., pre‑0.12 vs. 0.12+), confirm the exact version and whether the declared ordering is respected before relying on this pattern.