Ordering

The order of objects (resource, data, variable, output, locals or any other object) is technically not relevant. You can put any object in any .tf file.

There is an order that is nice for humans. Making a decision help to make a repository more maintainable.

Dependency oriented

In this order, you describe the object that’s reused later, first:

resource "x" "foo" {
  name = "bar"
}

resource y "bar" {
  name = "foo"
  relation = x.foo.id
}

This means that the top of the file contains resources without a relation, where the bottom likely has resources that depend on above resources.

Architecture oriented

In this order, you keep in mind how the architecture is used. For example:

resource "dns_record" "default" {
  name  = "www"
  value = loadbalancer.default.ip
}

resource "loadbalancer" "default" {
  name = "bar"
}

Bottom-up oriented

In this order, you describe the resource that provides a service first, and introduce depending resources later. The Terraform documentation for GCP uses this ordering frequently.

resource "instance" "default" {
  name    = "x"
  network = network.default.id
  volume  = volume.default.id
}

resource "network" "default" {
  name = "y"
}

resource "volume" "default" {
  name = "z"
  size = "23G"
}