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

[roffe]

module "ecs-codepipeline" {
  source             = "cloudposse/ecs-codepipeline/aws"
  version            = "0.23.1"
  name               = "hello-world"
  namespace          = "eg"
  stage              = "staging"
  github_oauth_token = var.github_oauth_token
  region             = "eu-***-1"
  image_repo_name    = "hello-world"
  //github_webhooks_token = "xxxxxxxxxxxxxx"
  repo_owner       = "someowner"
  repo_name        = "ecs-hello-world"
  branch           = "master"
  service_name     = "hello-world"
  ecs_cluster_name = aws_ecs_cluster.lab-1.name
  privileged_mode  = "true"
}

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

  on .terraform/modules/ecs-codepipeline.codebuild/main.tf line 206, in resource "aws_codebuild_project" "default":
 206: resource "aws_codebuild_project" "default" {
 *

[inhumantsar]

#80 does not fix this. This error should not have occurred even without the passthrough since the the bucket name should have defaulted to something valid.

The passthrough enabled a workaround but does not solve this issue.

[nitrocode]

When you set closes and then the issue number in the PR, when the PR is merged, github will auto close the issue linked. I've reopened it for now.

[nitrocode]

Seems like this may be related hashicorp/terraform-provider-aws#10195

[inhumantsar]

I'm not sure that issue is related. With the testing I did on cloudposse/terraform-aws-codebuild#91 the issue went away when I used a string, whether interpolated or directly. It was only when random_string or random_id got involved that the error would occur.

It's really strange. I would expect that if random_string failed, the result would be simply missing the suffix, like bucket-name-. In this situation though, it's like the entire interpolated string ends up converted into a null or something.

[nitrocode]

What terraform versions are being used? The test uses 0.13 and I'm using 1.0 locally. Both seem to work as expected.

How can I reproduce the issue?

[nitrocode]

Perhaps you could PR a terratest and if we can get a failing result then we can figure out how to update the module so the tests pass.

[inhumantsar]

what version of hashicorp/random are you using?

❯ terraform --version
Terraform v1.0.1
on darwin_amd64
+ provider registry.terraform.io/hashicorp/aws v3.52.0
+ provider registry.terraform.io/hashicorp/random v3.1.0

i'm not sure where to even start with getting terratest to repro the issue. my best guess atm is that it's a platform thing. mac vs linux. but being that random and terraform are Go apps, i'm not sure it could be a platform thing.

@roffe @techies20 @pgbce @putarpuar @nilreml if you're also experiencing this issue, can you post your terraform --version output as well?

[nilreml]

terraform --version
Terraform v0.14.11
on linux_amd64
+ provider registry.terraform.io/hashicorp/aws 3.54.0
+ provider registry.terraform.io/hashicorp/random v3.1.0

[nk9]

I'm seeing this too.

$ terraform --version
Terraform v1.1.4
on darwin_amd64
+ provider registry.terraform.io/hashicorp/aws v3.59.0
+ provider registry.terraform.io/hashicorp/random v3.1.0
+ provider registry.terraform.io/hashicorp/template v2.2.0

│ Error: cache location is required when cache type is "S3"
│ 
│   with module.project_module.module.ecs_codepipeline.module.codebuild.aws_codebuild_project.default[0],
│   on .terraform/modules/project_module.ecs_codepipeline.codebuild/main.tf line 292, in resource "aws_codebuild_project" "default":
│  292: resource "aws_codebuild_project" "default" {

module "ecs_codepipeline" {
  source                = "cloudposse/ecs-codepipeline/aws"
  version               = "0.28.5"
  enabled               = true

  region                = var.region
  name                  = var.project_name_new
  namespace             = "eg"
  stage                 = "staging"
  github_oauth_token    = "/Prod/GITHUB_OAUTH_TOKEN"
  repo_owner            = var.repository_owner
  repo_name             = var.repository_name
  branch                = var.repository_branch
  service_name          = "api"
  image_repo_name       = var.ecr_repository_name
  ecs_cluster_name      = "${var.project_name}_cluster"
  privileged_mode       = true
}

[nk9]

And as noted in #80, an effective workaround is to set cache_bucket_suffix_enabled = false.