CODEWITHSUNDEEP
terraform
GreenyMcDuff
GreenyMcDuff

Reputation: 3622

Running python script from Terraform local-exec provisioner

Versions

Setup

I have the following folder structure

.
├── main.tf
└── scripts
    └── my_script.py

main.tf:

locals {
  python = (substr(pathexpand("~"), 0, 1) == "/") ? "python3" : "python.exe"
}

resource "null_resource" "custom_objects" {
  for_each = local.custom_objects

  triggers = {
    name        = each.key
  }

  provisioner "local-exec" {
    command = <<-EOT
      python3 -m my_script '{"key_1": "value_1"}'
    EOT
    interpreter = [
      local.python
    ]
    working_dir = "${path.module}/scripts"
  }
}

./scripts/my_script.py:

if __name__ == '__main__':
    print('my_script executed.')

Problem When I run terraform plan/apply I get the following error:

Error running command 'python3 -m my_script '{"key_1": "value_1"}' │ ': exit status 2. Output: /home_path/.pyenv/versions/3.9.1/bin/python3: can't open file '/path_to_repo/scripts/python3 -m my_script '{"key_1": "value_1"}' │ ': [Errno 2] No such file or directory

Qeustions

Upvotes: 3

Views: 11432

Answers (1)

Marko E
Marko E

Reputation: 18203

It seems when the interpreter argument is provided it is appended to the path defined in the working_dir, which explains this error:

'/path_to_repo/scripts/python3 -m my_script '{"key_1": "value_1"}'

As you can see, the path ends with python3 which probably is not there.

I managed to get the result using the following syntax:

resource "null_resource" "custom_objects" {
  for_each = local.custom_objects

  triggers = {
     name = each.key
  }

  provisioner "local-exec" {
    command = <<-EOT
      ${local.python} -m my_script '{"key_1": "value_1"}'
    EOT
    working_dir = "${path.module}/scripts" # works with ${path.root} as well
  }
}

You might also find terraform_data as a good replacement for the null_resource.

Upvotes: 8

Related Questions