[BUG] `terramate create --all-terragrunt` error with nested stacks

[ohmer]

Describe the bug

Terramate does not exclude implicit nested stacks from after field generating invalid stack dependency graph.

To Reproduce

Steps to reproduce the behavior:

1. Have a code base with nested stacks
2. Run command terramate create --all-terragrunt. Terramate adds parent in after list in stack.tm.hcl
3. terramate list --run-order throws Invalid stack configuration > cycle detected

Expected behavior Produce a valid dependency graph exclucing parent stack from Terramate dependency graph since its implicit.

Log Output Add logs from Terramate to help debug your problem.

Environment

• OS: MacOS Sonoma
• OS Version 14.5
• Git Version 2.45.2
• Terramate Version 0.9.0

Additional context I love Terramate, happy to help! Thanks for putting such a great tool out there!

[soerenmartius]

Thanks for reporting this. We will look at this and get back to you asap!

[ohmer]

Thanks!

Can't share code to reproduce easily, sorry. As I am working on a mono repo with 1298 stacks, I can provide serious testing.

[ohmer]

Example of invalid deps graph below

Child referencing parent:

Parent referencing child:

[soerenmartius]

@i4ki @snakster could you guys take a look at this please?

[i4ki]

Hey @ohmer

can you check if the branch in this PR fixes the issue? I managed to reproduce it in parent-child case but maybe it's not the same case you are getting.

To install the branch, you can do:

go install github.com/terramate-io/terramate/cmd/terramate@i4k-fix-tg-after-cycle

You can also export GOBIN=<place> to install it in a separate directory.

Let me know if it works.

[ohmer]

Let me know if it works

Awesome, will do ASAP!

[ohmer]

Looks better, but not a fix I believe. How I tested:

GOBIN=~/bin/ go install github.com/terramate-io/terramate/cmd/terramate@i4k-fix-tg-after-cycle
time ~/bin/terramate create --all-terragrunt
...
Created stack /support/global/default
Created stack /support/us-east-1/default
Error: failed to initialize Terragrunt modules
> validating stack fields: field after has duplicate "/prod/eu-west-3/data-biron/aurora" element

________________________________________________________
Executed in  509.13 secs    fish           external
   usr time  668.92 secs  222.00 micros  668.92 secs
   sys time  110.65 secs  881.00 micros  110.65 secs

> cycle detected: /preprod/eu-west-3/data -> /preprod/eu-west-3/data/kms -> /preprod/eu-west-3/data: checking node id "/preprod/eu-west-3/data"

FYI

find . -type f -name terragrunt.hcl -prune -not -path '*/.terragrunt-cache/*' | wc -l
    1202

On the cyclic dependency:

❯ grep -A1 'dependency "'  preprod/eu-west-3/data/terragrunt.hcl
dependency "web" {
  config_path = "../web"
--
dependency "ecom_private" {
  config_path = "../ecom-private"
--
dependency "erp" {
  config_path = "../erp"
--
dependency "support_resources" {
  config_path = "../../../support/eu-west-3/default/resources"
--
dependency "kong" {
  config_path = "../kong"
--
dependency "control" {
  config_path = "../../../support/eu-west-3/control/sns"
--
dependency "kms" {
  config_path = "./kms"

❯ cat preprod/eu-west-3/data/stack.tm.hcl
stack {
  name        = "data"
  description = "data"
  after       = ["/preprod/eu-west-3/data/kms", "/preprod/eu-west-3/ecom-private", "/preprod/eu-west-3/erp", "/preprod/eu-west-3/kong", "/preprod/eu-west-3/web", "/support/eu-west-3/control/sns", "/support/eu-west-3/default/resources"]
  id          = "ea79feb0-4c94-412d-af50-489fea414d60"
}

Since /preprod/eu-west-3/data/kms is nested under /preprod/eu-west-3/data, preprod/eu-west-3/data/stack.tm.hcl should not add any of its children to the after list as it contradicts the implicit before in the parent, correct?