Ctrl+C leaves orphaned GCS state lock - SIGKILL sent before OpenTofu can release lock

[bradenwright]

Describe the bug

When pressing Ctrl+C during a Terraform/OpenTofu operation with a GCS backend, Terragrunt logs "Gracefully shutting down..." but sends SIGKILL to the subprocess within ~3ms. This doesn't give OpenTofu enough time to release the remote state lock, leaving it orphaned in GCS.

The same test with a local backend works correctly - OpenTofu receives the signal and shuts down gracefully.

Steps To Reproduce

main.tf:

terraform {
  required_providers {
    time = { source = "hashicorp/time", version = "~> 0.9" }
  }
}

resource "time_sleep" "wait" {
  create_duration = "10s"
}

terragrunt.hcl:

terraform { source = "./" }

remote_state {
  backend = "gcs"
  config = {
    bucket   = "your-bucket"
    prefix   = "test"
    location = "us-east1"
  }
  generate = { path = "backend.tf", if_exists = "overwrite_terragrunt" }
}

Commands:

terragrunt init
terragrunt apply -auto-approve
# Press Ctrl+C once while "time_sleep.wait: Creating..." is shown
terragrunt apply -auto-approve  # Fails with lock error

Expected behavior

After pressing Ctrl+C once, OpenTofu should have enough time to release the GCS state lock before being terminated. The subsequent terragrunt apply should not fail with a lock error.

Must haves

• Steps for reproduction provided.

Nice to haves

• Terminal output

Terminal output:

23:35:48.882 STDOUT tofu: time_sleep.wait: Creating...
23:35:50.423 STDOUT tofu: Interrupt received.
23:35:50.423 INFO   Interrupt signal received. Gracefully shutting down...
23:35:50.423 STDOUT tofu: Please wait for OpenTofu to exit or data loss may occur.
23:35:50.424 STDOUT tofu: Gracefully shutting down...
23:35:50.424 STDOUT tofu: Stopping operation...
23:35:50.427 ERROR  error occurred:
* Failed to execute "tofu apply -auto-approve" in .terragrunt-cache-test/...
  signal: killed

Note: Only ~3ms between "Gracefully shutting down" and "signal: killed"

Subsequent apply fails:

23:35:58.343 STDERR tofu: │ Error: Error acquiring the state lock
23:35:58.344 STDERR tofu: │ Lock Info:
23:35:58.344 STDERR tofu: │   ID:        1764830148129739
23:35:58.344 STDERR tofu: │   Path:      gs://starwatch-terraform-backend/test-signal-handling/test/default.tflock
23:35:58.344 STDERR tofu: │   Operation: OperationTypeApply

Versions

• Terragrunt version: v0.93.12
• OpenTofu/Terraform version: OpenTofu v1.10.7
• Environment details: macOS 15 (darwin/arm64)

Additional context

• With a local backend, the same test works correctly - OpenTofu gracefully shuts down and no state corruption occurs.
• The issue appears to be that Terragrunt sends SIGKILL to the subprocess too quickly after receiving SIGINT, not giving OpenTofu time to release the GCS lock.

[bradenwright]

This maybe a duplicate of: #4839