State migration for old attribute "body" is missing.

[pascal-hofmann]

Terraform CLI and Provider Versions

terraform 1.2.5
provider 3.0.0

Terraform Configuration

irrelevant

Expected Behavior

Proper migration of old state.

Actual Behavior

… could not be decoded from the state: unsupported attribute "body". :(

Steps to Reproduce

Create state with old module, update to new module and try to apply again.

How much impact is this issue causing?

High

Logs

No response

Additional Information

No response

Code of Conduct

• I agree to follow this project's Code of Conduct

[pascal-hofmann]

See #140 / #142

[pascal-hofmann]

I'm looking into this. Will try to create a PR today.

[bendbennett]

Hi @pascal-hofmann sorry you ran into trouble here.

Could you possibly post a complete example of the terraform code you are using along with the version of the http provider that you are/were using prior to upgrading to v3.0.0?

Are you trying to use the body attribute somewhere in your configuration?
Reason I ask is that the body attribute has been removed in v3.0.0 and replaced by response_body so I'm wondering if this is part of the issue?

[pascal-hofmann]

This is the issue. data resources are also persisted to state. When the schema changes, a state upgrader is required. I'm currently working on a PR. I'm just getting used to the new framework instead of sdk.

[pascal-hofmann]

@bendbennett I think currently the framework does not support state upgraders for data sources. :(

@bflad Do you have an idea, how to handle this case?

In the meantime I think I'll pin the provider version to 2. (I don't want to fix the state manually, because we have a lot of these data sources in a lot of states.)

[pascal-hofmann]

I was able to create a minimal test case for this.

It only happens if the deprecated body attribute has still been used in the last apply with terraform-provider-http 2.x.

If you then upgrade to 3.0.0, it will fail with:

╷
│ Error: Unsupported attribute
│
│   on main.tf line 15, in resource "local_file" "this":
│   15:   content  = data.http.this.body
│
│ This object has no argument, nested block, or exported attribute named "body".

Workaround: Switch to using response_body prior to upgrading to terraform-provider-http 3.

Test case:

terraform {
  required_providers {
    http = {
      source  = "hashicorp/http"
      version = "<3" # comment after first apply, run terrform init -upgrade, change body to response_body and apply again
    }
  }
}

data "http" "this" {
  url = "https://www.hashicorp.com"
}

resource "local_file" "this" {
  content  = data.http.this.body
  filename = "${path.module}/hashicorp.com.html"
}

I'll close this issue.

[bflad]

This feels like a redux of this recent core issue hashicorp/terraform#30823 and pull request hashicorp/terraform#30830 -- it seems valuable to raise an issue upstream to see if Terraform core can actually ignore data source prior state in this situation, if its being consulted for plan generation.

[bflad]

By the way, there is no concept of state upgrades for data sources. There is only the UpgradeResourceState RPC that is used with managed resources. The prior data source state is only supposed to be "informative" for plan generation.

[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.