Custom provider built with terraform-plugin-framework does not find resource/data source types with TF 1.6

[dhohengassner]

Terraform Version

1.6.0

Terraform Configuration Files

resource "custom_service" "my_service" {
  display_name  = "Custom Service"
  description   = "Test Custom Service"
}

Debug Output

Terraform has been successfully initialized!
You may now begin working with Terraform. Try running "terraform plan" to see
any changes that are required for your infrastructure. All Terraform commands
should now work.
If you ever set or change modules or backend configuration for Terraform,
rerun this command to reinitialize your working directory. If you forget, other
commands will detect it and remind you to do so if necessary.
##### Running 'terraform apply --auto-approve' ################################################################
2023-10-05T13:51:54.060Z [ERROR] provider.terraform-provider-customiam: Response contains error diagnostic: diagnostic_severity=ERROR @caller=github.com/hashicorp/terraform-plugin-go@v0.19.0/tfprotov6/internal/diag/diagnostics.go:58 tf_proto_version=6.4 tf_req_id=748e53e1-c5cf-fcc4-135f-563bb7db7de4 tf_rpc=ValidateResourceConfig @module=sdk.proto diagnostic_summary="Resource Type Not Found" tf_provider_addr=app.terraform.io/rbmh-mit/customiam tf_resource_type=example_service diagnostic_detail="No resource type named \"example_service\" was found in the provider." timestamp=2023-10-05T13:51:54.060Z
2023-10-05T13:51:54.060Z [ERROR] provider.terraform-provider-customiam: Response contains error diagnostic: tf_provider_addr=app.terraform.io/rbmh-mit/customiam tf_rpc=ValidateResourceConfig @caller=github.com/hashicorp/terraform-plugin-go@v0.19.0/tfprotov6/internal/diag/diagnostics.go:58 tf_req_id=c4075578-dba7-128a-6a2d-7637e983c6a2 @module=sdk.proto diagnostic_detail="No resource type named \"example_service\" was found in the provider." diagnostic_summary="Resource Type Not Found" tf_proto_version=6.4 diagnostic_severity=ERROR tf_resource_type=example_service timestamp=2023-10-05T13:51:54.060Z
2023-10-05T13:51:54.060Z [ERROR] vertex "example_service.example_service1" error: Resource Type Not Found
2023-10-05T13:51:54.060Z [ERROR] vertex "example_service.example_service2" error: Resource Type Not Found
╷
│ Error: Resource Type Not Found
│ 
│   with example_service.example_service1,
│   on resource-service.tf line 1, in resource "example_service" "example_service1":
│    1: resource "example_service" "example_service1" {
│ 
│ No resource type named "example_service" was found in the provider.

Expected Behavior

Provider works fine as it does with Terraform v1.5

Actual Behavior

The same provider binary crashes with errors:

No resource type named "custom_service" was found in the provider

Steps to Reproduce

terraform init
terraform apply

Additional Context

The provider was built this year on top of the latest framework.

https://github.com/hashicorp/terraform-plugin-framework

References

No response

[jbardin]

Hi @dhohengassner,

Thanks for filing the issue. I think we're going to need a more complete example to figure out what might be going on here. The provider is what is returning the error message, so it's not clear yet what Terraform could be doing differently to cause it. It looks like you have obfuscated the output a bit, but if the provider is called custom, it would normally have resources named custom_service, however the configuration in the error shows example_service which isn't going to match resources in the provider.

[dhohengassner]

Hey @jbardin,

Thanks for the quick response and sorry for not mentioning the adjustments earlier.

Yeah, you are right; I obfuscated the resource name as well as the output (and not even well 👎 ) since it is an internal provider not meant for public use.
The resources and data sources are found fine with 1.5 and older terraform versions; I have now pinned the version and it works with the same binary.

The resource/provider names are the only thing I adjusted; the rest of the output comes from an actual pipeline.
Is there anything else I can provide in addition to the logs that would help?

I'm also curious if other providers using the framework will run into the same issue.
Thanks!

[jbardin]

I haven't heard of anything similar with the framework, and haven't come up with any hypotheses yet about what could be happening either. Are you on the latest release of the framework?

If it's possible to put together a minimal config, and a full trace log from the plan, that may help pinpoint what is going on.

[andrewesweet]

I can replicated this with 1.6.0-beta2 and later by using the data source from https://github.com/hashicorp/terraform-provider-scaffolding-framework without change.

[drbild]

We're seeing the same issue with both our public provider and a private, internal-use-only, provider.

[jbardin]

Thanks @andrewesweet, the scaffolding example should work as a good starting point!

[apparentlymart]

Hi all,

Detailed investigation of this one is still ongoing, but based on internal discussion it seems like this is a bug in the plugin framework itself but it appears as new in Terraform v1.6.0 because this new Terraform Core release introduced a new provider protocol feature which the framework is attempting to use but is using it incorrectly, causing this error.

Based on the information in the framework's changelog it seems like building your providers using framework release v1.3.5 should avoid this problem for now, since the problematic use of the new protocol feature came in v1.4.0.

There are folks on the framework dev team digging deeper into this currently, and hopefully we'll have a fix coming soon. The fix is likely to be in the framework primarily, and so will probably require upgrading to a newer version of the framework once fixed, but I'll let the folks working on it share more information about what's required once the details are clearer.

Thanks again for reporting this!