Deposed resources do not call when = destroy provisioners

[arbourd]

When using create_before_destroy lifecycle hook, it appears that remote-exec provisioners are not called. Here's the pattern I'm trying to adapt (due to Equinix Metal not having a shutdown script function):

1. A new metal_device is brought up
2. A new null_resource is brought up to house the triggers
3. The deposed null_resource is destroyed, executing the remote-exec which ensures that the node leaves the consul pool
4. The deposed metal_device is destroyed

Terraform Version

Terraform v1.2.3
on darwin_arm64
+ provider registry.terraform.io/equinix/metal v3.3.0-alpha.4
+ provider registry.terraform.io/hashicorp/null v3.1.1

Terraform Configuration Files

resource "metal_device" "device" { 
  # ...

  lifecycle {
    create_before_destroy = true
  }
}

resource "null_resource" "device" {
  triggers = {
    name = metal_device.device.hostname
    host = metal_device.device.access_public_ipv4
    key  = var.metal_ssh_private_key
  }

  provisioner "remote-exec" {
    when = destroy

    inline = [
      "/opt/bin/consul leave"
    ]
    on_failure = continue

    connection {
      type        = "ssh"
      user        = "core"
      host        = self.triggers.host
      private_key = self.triggers.key
    }
  }

  lifecycle {
    create_before_destroy = true
  }
}

Debug Output

Nothing of immediately value popped up here.

Expected Behavior

The remote-exec script should run.

Actual Behavior

The remote-exec does not run.

metal_device.device: Creation complete after 10m9s [id=6d5385ee-128f-47f8-aaf5-338a3824790e]
null_resource.device: Creating...
null_resource.device: Creation complete after 0s [id=5852924778467424020]
null_resource.device (deposed object 1ec4e06d): Destroying... [id=7715824139324971453]
null_resource.device: Destruction complete after 0s
metal_device.device (deposed object 6307fae0): Destroying... [id=3c3711ac-18c6-4bd5-a174-be9ddd074b73]
metal_device.device: Destruction complete after 1s

There is no indication that the SSH script ran. Removing the lifecycle shows that it does:

null_resource.worker["da/0"]: Destroying... [id=5882924178467424020]
null_resource.worker["da/0"]: Provisioning with 'remote-exec'...
null_resource.worker["da/0"] (remote-exec): Connecting to remote host via SSH...
null_resource.worker["da/0"] (remote-exec):   Host: 1.1.1.1
null_resource.worker["da/0"] (remote-exec):   User: user
null_resource.worker["da/0"] (remote-exec):   Password: false
null_resource.worker["da/0"] (remote-exec):   Private key: true
null_resource.worker["da/0"] (remote-exec):   Certificate: false
null_resource.worker["da/0"] (remote-exec):   SSH Agent: false
null_resource.worker["da/0"] (remote-exec):   Checking Host Key: false
null_resource.worker["da/0"] (remote-exec):   Target Platform: unix
null_resource.worker["da/0"] (remote-exec): Connected!
null_resource.worker["da/0"] (remote-exec): Graceful leave complete
null_resource.worker["da/0"]: Destruction complete after 5s

Steps to Reproduce

1. Create and apply Terraform similar to the above, where a remote-exec provision is also using the lifecycle hook.

Additional Context

Running in Terraform Cloud

[jbardin]

Thanks for filing the issue @arbourd. We are tracking this issue in #13549.

[arbourd]

Oh sorry, I searched but didn't find anything. Appreciate it @jbardin 🙏

[github-actions]

I'm going to lock this issue because it has been closed for 30 days ⏳. This helps our maintainers find and focus on the active issues.
If you have found a problem that seems similar to this, please open a new issue and complete the issue template so we can capture all the details necessary to investigate further.