[Bug]: `linode_object_storage_bucket` saves different region to tfstate than declared

[pkawa-dev]

Terraform Version

v.1.13.0

Linode Provider Version

version = "3.4.0"

Effected Terraform Resources

linode_object_storage_bucket

Terraform Config Files

terraform {                                                                         
  required_providers {                                                              
    linode = {                                                                      
      source  = "linode/linode"                                                     
      version = "3.4.0"                                                             
    }                                                                               
  }                                                                                 
}                                                                                   
                                                                                    
provider "linode" {                                                                 
  token = "X"        
}                                                                                   
                                                                                    
resource "linode_object_storage_bucket" "foobar" {                                  
  region = "us-ord-1"                                                               
  label  = "testpkawamzuber"                                                        
}

Debug Output

No response

Panic Output

No response

Expected Behavior

As region is not valid (it's a correct cluster parameter, but not region), bucket should not be created.

Actual Behavior

In the first run, bucket creates, and then, in 2nd run, bucket gets recreated because the terraform detects region change.

  # linode_object_storage_bucket.foobar must be replaced
-/+ resource "linode_object_storage_bucket" "foobar" {
      ~ cluster       = "us-ord-1" -> (known after apply)
      ~ cors_enabled  = true -> (known after apply)
      ~ endpoint      = "us-ord-1.linodeobjects.com" -> (known after apply)
      ~ endpoint_type = "E1" -> (known after apply)
      ~ hostname      = "testpkawamzuber.us-ord-1.linodeobjects.com" -> (known after apply)
      ~ id            = "us-ord:testpkawamzuber" -> (known after apply)
      ~ region        = "us-ord" -> "us-ord-1" # forces replacement
      ~ s3_endpoint   = "us-ord-1.linodeobjects.com" -> (known after apply)
      + versioning    = (known after apply)

Steps to Reproduce

Run the pasted terraform 2 times in a row.

[jifwin]

+1

[zliang-akamai]

Hi @pkawa-dev @jifwin, thanks for the feedback! us-ord-1 is generally not a valid region for Terraform provider, even though it's accepted by the Linode API v4 for backward compatibility considerations during bucket creation. You can set the region to be "us-ord" in this case to resolve the issue. Let me know if you have any other question or comment. Thank you again!

[fenio-akamai]

Seriously? For whom are you preserving this backward compatibility? For people clicking in WebUI? You can hide this incompatibility for them. Please focus on terraform users.

[zliang-akamai]

@fenio-akamai We pass the input values from TF config/plan as is into Linode API v4 so the TF execution is unexpectedly impacted by the API behavior of accepting cluster names as aliases of the corresponding regions. We plan to add a validation logic in our TF provider in the future to prevent clusters being used as regions in TF config for solving this issue.

[pkawa-dev]

@zliang-akamai validation logic would be great. I was aware of the fix (us-ord instead of us-ord-1), but that generally should either fail or save us-ord-1 in the state. Otherwise, terraform will be stuck. Having validation setup will be awesome!