Panic creating/reading remote state

[agonzalezro]

Terraform Version

Terraform v0.7.0-dev

Affected Resource(s)

• terraform_remote_state using s3 bucket.

Terraform Configuration Files

resource "aws_s3_bucket" "state_bucket" {
  bucket = "terraform-state-abcdefg123"
  acl = "private"
}

resource "terraform_remote_state" "on_s3_by_environment_and_prefix" {
  backend = "s3"

  config {
    bucket = "${aws_s3_bucket.state_bucket.bucket}"
    key    = "terraform.tfstate" # "${var.prefix}something.tfstate"
  }
}

Debug Output

I can privately share it or anonymized it if needed.

Panic Output

I can privately share it or anonymized it if needed.

Expected Behavior

• Create a bucket and upload the sate to it.

Actual Behavior

• The bucket is created but the state file isn't.

Steps to Reproduce

1. terraform apply

Important Factoids

We are using roles for the accounts.

[stack72]

Hi @agonzalezro

My apologies this has happened - please can you tell me post the piece of the debug output that has the panic in it? We don't need anything that has any config in it, just the actual panic message and the few lines after it that with give us an idea

Paul

[agonzalezro]

Hi @stack72, thanks for the quick response.

Could this help? https://gist.github.com/agonzalezro/69097ba7c3488c8ab7a802c18129ae81#file-error_console

[apparentlymart]

@agonzalezro was the remote state you're reading also generated by Terraform 0.7.0, or was that produced by an earlier version of Terraform?

There were some changes to the state format for 0.7.0 so I'm wondering if you've hit a bug in the state upgrade code.

[jbardin]

Hi @agonzalezro,

Thanks for the extra info. I can reproduce this locally, and we will look into it.

[agonzalezro]

@apparentlymart it doesn't seem like it. I tried in a completely clean state (no previous tfstate).

@jbardin thanks! Good luck & let me know if there is something else I can help with.

[jbardin]

The following config will produce a crash on terraform apply with no previous state required (and without the warning message):

resource "aws_s3_bucket" "state_bucket" {
  bucket = "terraform-state-bucket-1234567"
}

data "terraform_remote_state" "remote_state" {
  backend = "s3"

  config {
    bucket = "${aws_s3_bucket.state_bucket.bucket}"
    key    = "terraform.tfstate"
  }
}

[jbardin]

Hi @agonzalezro,

After reading more carefully, I think the part of the problem here is that you're not starting with any remote state. This config won't create a remote state, because the terraform_remote_state resource is read-only. This is a bit confusing, but the recent change to a data source makes this more apparent. You first have to run terraform remote config with the required options. https://www.terraform.io/docs/state/remote/s3.html to sync the state.

The crash when the remote state is missing is still a bug, which we can fix for the next release.

[agonzalezro]

Do you mean that I require a manual step before being able to use the remote state? If I need to run

terraform remote config x y z

knowing the values o x. y & z. I don't really see the advantage of having the remote configuration in a file.

We are trying to create different states per environment based on ${environment} and ${prefix} so we will end up with N different states. I don't see if this is going to be manageable if we can't trust that the state is there os it's going to be created.

Is it possible perhaps to work around this adding an empty file to the s3 bucket with an empty state?

Do you perhaps know how other people ended up managing this?

Thanks!

[rmenn]

I have build this from source Terraform v0.7.0-dev (bf63dcc172c2edcd09ca17e43ca13148d01b8c33)

I am hitting this as well. Crash on creating the remote state.

aws_s3_bucket.ops-bucket: Creation complete
terraform_remote_state.state-storage: Creating...
  backend:       "" => "s3"
  config.%:      "" => "3"
  config.bucket: "" => "xxxxxxxx"
  config.key:    "" => "ap-south-1/terraform.tfstate"
  config.region: "" => "ap-south-1"
Error applying plan:

1 error(s) occurred:

* terraform_remote_state.state-storage: unexpected EOF

Terraform does not automatically rollback in the face of errors.
Instead, your Terraform state file has been partially updated with
any resources that successfully completed. Please address the error
above and apply again to incrementally change your infrastructure.

Debug message available here : https://gist.github.com/rmenn/88fe8528ab4396138a7539da2bd81deb

[jbardin]

Hi @rmenn,

The above PR wasn't merged yet. The latest version on master should no longer crash in this case.

@agonzalezro

The terraform_remote_state resource, which is now referred to as a data source, is read only. You can use the terraform_remote_state to produce outputs for other parts of your config based on a remote state file, but having terraform_remote_state in your config will not change that remote state file whether it exists or not. Syncing with a remote state file is done through a different mechanism setup by the terraform remote config command. (I agree there could be improvement to the documentation around this, and we still need to add more documentation for the new concept of data sources).

[ghost]

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.