文件作为计数变量，terraform

Question

Im trying to pass a file with 3 names, to use with count in a terraform resource, the file is the following

userlist.json

["pepe", "pipo", "popo"]

what im trying to do with terraform is to create X quantity of policies based on the lenght of the json file, but im getting the following error

main.tf

resource "aws_iam_policy" "policy" {
  count       = length(file("userlist.json"))
  name        = file("userlist.json"[count.index])
  path        = var.path
  description = var.description
  policy      = file(var.policy_json_location[count.index])
}

Im getting the following error

 Error: Invalid index
│
│   on main.tf line 3, in resource "aws_iam_policy" "policy":
│   3:   name        = file("files/userlist.json"[count.index])
│     ├────────────────
│     │ count.index is a number, known only after apply
│
│ This value does not have any indices.

it seems to be that terraform doesnt like the way im passing the file in count, instead of a string, even if "file" is declared there

Answer 1

If I understood the question correctly, you may need to use jsondecode(file("test.json")) instead of file("test.json")

The following snippet illustrates the behavior

resource "null_resource" "test_res" {
    count = length(jsondecode(file("test.json")))
} 

output "file_content" {
  value = jsondecode(file("test.json"))
}

output "test_res" {
    value = null_resource.test_res
}

The value of the outputs was

file_cont = [
  "pipo",
  "pepo",
  "pepe",
]
test_res = [
  {
    "id" = "6672935723139812109"
    "triggers" = tomap(null) /* of string */
  },
  {
    "id" = "7815380621246912709"
    "triggers" = tomap(null) /* of string */
  },
  {
    "id" = "6843574097785729573"
    "triggers" = tomap(null) /* of string */
  },
]

Unsolicited suggestion: use also terraform console to debug these problems, eg:

$ terraform console
> file("test.json")
<<EOT
["pipo", "pepo", "pepe"]

EOT

> jsondecode(file("test.json"))
[
  "pipo",
  "pepo",
  "pepe",
]

Answer 2

Had to add jsondecode and move the count.index in order for it to work, this is the final resource working

resource "aws_iam_policy" "policy" {
  count       = length(jsondecode(file(var.policy_name)))
  name        = jsondecode(file(var.policy_name))[count.index]
  description = jsondecode(file(var.policy_name))[count.index]
  policy      = file(jsondecode(file(var.policy_json_location))[count.index])
  tags        = var.tags
  path        = var.path
}

Answer 3

Have you considered using a for_each instead of a count? This also removes the need for using the index. Something along these lines

resource "aws_iam_policy" "policy" {
  for_each    = to_set(jsondecode(file(var.policy_name)))
  name        = each.value
  description = each.value
  policy      =(jsondecode(file(var.policy_json_location))[each.key])
  tags        = var.tags
  path        = var.path
} 

This example most probably doesn't offer you a full solution, but mainly to demonstrate how for_each works in Terraform. Please also refer to Terraform for_each documentation

since there are two different json files, you will probably need to make sure that the keys between them match up, but this should make it more robust.