Configured destroy entrypoint (and body) not being invoked

[crandall-chow-bl]

TL;DR

The configured destroy_cmd_entrypoint and destroy_cmd_body are not being invoked, even though the logs show the destroy was triggered and resource was removed from the .tfstate.

Expected behavior

On "destroy" (such as when TF needs to do a replace) the configured destroy_cmd_entrypoint and destroy_cmd_body should be invoked, similar to what happens on create, where the create_cmd_entrypoint and create_cmd_body are clearly invoked.

Observed behavior

Creates are doing as expected, calling the configured entrypoint, but destroys are not. The .tfstate data is indeed created and destroyed as expected, but the custom entrypoints are never getting invoked.

Terraform Configuration

terraform {
  required_version = ">= 1.0.0"

  required_providers {
    google = {
      source  = "hashicorp/google"
      version = "3.90.0"
    }
  }
}

provider "google" {
  region      = "us-central1"
  project     = "my-xxx"
  credentials = "my-xxx-account.json"
}

# comment out the module below after "create" to get TF to "destroy"
module "my_xxx_tester" {
  source  = "terraform-google-modules/gcloud/google"
  version = "3.1.0"

  create_cmd_endpoint = "gcloud"
  create_cmd_body        = "version"
  destroy_cmd_body      = "gcloud"
  destroy_cmd_body       = "version"
}

Terraform Version

Terraform v1.1.5
on darwin_amd64

Debug Output

https://gist.github.com/crandall-chow-bl/b5a427c276f6c2c6be0d402304affe53

Additional information

Steps to Reproduce

1. terraform apply with the TF file above to "create"
2. Comment out the module in the TF file
3. `terraform apply" with the TF file above to "destroy"
4. Note that the destroy entrypoint/body do not get invoked

[morgante]

What happens if you run terraform destroy instead of commenting the module out?

[crandall-chow-bl]

Will try that and reply but that's certainly not what I want to do. Commenting it out is just the test to show how the "destroy" can be called.

The actual use case is to pass a release number in the create_cmd_body. When I want to deploy a new version, I update the module stanza with the new version number. Calling terraform apply will see the change to the module, and have it do a "replace" (invoking my previous destroy_cmd_entrypoint and destroy_cmd_body saved in the .tfstate, then invoking the new create_cmd_entrypoint and body).

Also, notice in the debug output, some "destroy" is indeed being triggered, but my configured entrypoint/body are not called.

[morgante]

Will try that and reply but that's certainly not what I want to do. Commenting it out is just the test to show how the "destroy" can be called.

Yeah, I'm mostly trying to figure out why it wouldn't be invoked in your environment.

[crandall-chow-bl]

I attempted the terraform destroy and sure enough, doing THAT does indeed invoke the entrypoint and body (see gist below). So it's just the destroy on a normal apply path (e.g. delete the resource, or modify the resource causing a delete/create replace) where the endpoint/body aren't being invoked.

https://gist.github.com/crandall-chow-bl/b5a427c276f6c2c6be0d402304affe53#file-destroy-txt

This is on macOS 12.2.1 in case that matters.

[morgante]

Unfortunately it looks like this is a Terraform Core issue: https://github.com/hashicorp/terraform/issues/13549

[github-actions[bot]]

This issue is stale because it has been open 60 days with no activity. Remove stale label or comment or this will be closed in 7 days