Provisioning CodeBuild - "Error: cache location is required when cache type is "S3""

[Dzhuneyt]

Community Note

• Please vote on this issue by adding a 👍 reaction to the original issue to help the community and maintainers prioritize this request
• Please do not leave "+1" or "me too" comments, they generate extra noise for issue followers and do not help prioritize the request
• If you are interested in working on this issue or have submitted a pull request, please leave a comment

Terraform Version

Terraform v0.12.7

Affected Resource(s)

• aws_codebuild_project

Terraform Configuration Files

Downloadable here: https://drive.google.com/file/d/1QM47TzsCFfpNB0KtspWwJoiOZ6uWYbqV/view?usp=sharing

Expected Behavior

CodeBuild is provisioned

Actual Behavior

Error: cache location is required when cache type is "S3"

  on codebuild/codebuild-tests.tf line 1, in resource "aws_codebuild_project" "codebuild_tests":
   1: resource "aws_codebuild_project" "codebuild_tests" {

References

• CodeBuild cache location doesn't honor aws resource dependency correctly #4639

[Dzhuneyt]

Exact line in the Terraform code where this is triggered: https://github.com/terraform-providers/terraform-provider-aws/blob/master/aws/resource_aws_codebuild_project.go#L579

I have no idea why this happens though. Maybe client side validation should be skipped when dynamic resource tokens are present?

[Dzhuneyt]

I figured out a workaround:

The S3 bucket resource that I was referencing, had the following structure:

resource "aws_s3_bucket" "ci_bucket" {
  bucket_prefix = "${lower(var.tag)}-${var.branch_name}-cb"
}

Solution was to replace it with:

resource "aws_s3_bucket" "ci_bucket" {
  bucket = "${lower(var.tag)}-${var.branch_name}-cb"
}

[nathancarmona]

Want to add my experience, the issue seems to be intermittent when trying to plan/apply AND when trying to delete. A template applied and was unchanged, but can't proceed past refreshing state on a destroy due to identical error:

Error: cache location is required when cache type is "S3"

  on pipelines.tf line 7, in resource "aws_codebuild_project" "build":
   7: resource "aws_codebuild_project" "build" {

example of the cache settings on the resource:

resource "aws_codebuild_project" "build" {
...
    cache {
        location = "${module.s3_bucket.bucket_id}/build_cache"
        modes    = ["LOCAL_DOCKER_LAYER_CACHE", "LOCAL_SOURCE_CACHE"]
        type     = "S3"
    }
...

[sakisv]

I ran into the same thing.

The following does not work, and it also doesn't work if the aws_s3_bucket resource is in a different module and it exposes the name as an output.

resource "aws_s3_bucket" "codebuild" {
  bucket = var.codebuild_bucket_name
  acl    = "private"
}

resource "aws_codebuild_project" "project" {
  name = var.name

  cache {
    type     = "S3"
    location = "${aws_s3_bucket.codebuild.id}/${var.name}"
  }
# ...more stuff
}

My workaround was to pass the name directly to the aws_codebuild_project, with the downside that it doesn't create the dependency between the two resources.

resource "aws_s3_bucket" "codebuild" {
  bucket = var.codebuild_bucket_name
  acl    = "private"
}

resource "aws_codebuild_project" "project" {
  name = var.name

  cache {
    type     = "S3"
    location = "${var.codebuild_bucket_name}/${var.name}"
  }
# ...more stuff
}

[sunilkumarmohanty]

However, using aws_s3_bucket.codebuild.bucket instead of aws_s3_bucket.codebuild.id works.

[T00mm]

We have the same error, we use aws_s3_bucket.codebuild.bucket. I've tried switching from BUCKET to ID and back but always the same error.

[fruuf]

ran into the same issue, seems to work fine when the terraform already has the bucket state. got it to work with commenting out the aws_codebuild_project, deploying everything else including the cache bucket and then uncommenting aws_codebuild_project. (edit: aws version "3.23.0", terraform 0.14.4)

[mazzespazze]

@fruuf I have same aws version and same terraform version. Though, if I comment out the aws_codebuild_project still fails to to outputs.tf modules referring to it.

I'm following this tutorial: https://github.com/ashwin9798/node-react-nginx-docker-boilerplate

Any insights?

###########

My answer was to have inside the aws_codebuild_project a reference to the cache as:

cache {
    type  = "LOCAL"
    modes = ["LOCAL_DOCKER_LAYER_CACHE", "LOCAL_SOURCE_CACHE"]
  }

[jbg]

@mazzespazze that is not really related to this issue, since this issue is about using S3 cache, and you're using local cache.

I get the same issue if using bucket_prefix, but using bucket works fine. It seem like the cache location cannot be dependent on something that is not known until the plan applies.

[udayreddym]

Earlier my cache block used to look like below:

cache { type = "S3" location = aws_s3_bucket.artifacts.bucket }

my code used to work but all of sudden it stopped working. As of now, I have a workaround, by replacing the above block with the below code block, it started working.

cache {
   type     = "S3"
   # location = aws_s3_bucket.artifacts.bucket
   location = var.s3_bucket_artifacts
 }

Please let me know if this works for you.