[Bug]: equinix_metal_device loses track of existing devices when provisioning fails

[ctreatma]

Terraform Version

any

Equinix Provider Version

any

Effected Terraform Resources

equinix_metal_device

Terraform Config Files

No response

Debug Output

No response

Panic Output

No response

Expected Behavior

While provisioning some devices on hardware reservations using next-available, I ran into hard-to-resolve deployment failures:

1. The initial deployment timed out while the device was provisioning, causing the Equinix TF provider to explicitly remove the device from state
2. The second terraform run attempted to deploy a completely new device, but did not destroy the device from step 1 because it didn't know about that device
3. The second terraform run failed because I had no more reservations available

Terraform should keep a device in state when provisioning fails so that it can be cleaned up automatically on the next terraform run.

Actual Behavior

To resolve this, I had to go into the Metal console and manually delete the old servers to free up the reservations.

Steps to Reproduce

This is tricky to reproduce because it requires a deployment to run longer than the default 30-minute provisioning wait timer for equinix_metal_device. If you can figure out how to reliably trigger such a deployment, you can deploy a Metal device using next-available hardware reservation over and over until you run out of reservations.