REP-1293: Allocation group resource

[roman-castai]

Example usage from the current schema:

resource "castai_allocation_group" "ag_example" {
  name = "ag_example"
  cluster_ids = [
    "1a58d6b4-bc0e-4417-b9c7-31d15c313f3f",
    "d204b988-5db5-472e-a258-bf763a0f4a93"
  ]

  namespaces = [
    "namespace-a",
    "namespace-b"
  ]

  labels = {
    environment              = "production",
    team                     = "my-team",
    "app.kubernetes.io/name" = "app-name"
  }

  labels_operator = "AND"
}

[roman-castai]

Conflict or similar status would be useful if someone would like to move their existing allocation groups into terraform so that we don't create duplicate allocation groups.

[zolinas]

At the moment we do not require Unique names for Allocations Groups in the console. We should either add this check to the API as well, or remove it from here.

[roman-castai]

I will check to see if there is an organization that already uses same name for different labeled allocation group and therefore the change would affect someone, and if not I would add this check to the API.

If we don't enforce this it would be possible to have a terraform configuration with the completely equal two entries, e.g.:

resource "cast_ai_allocation_group" "my_allocation_group" {
  clusterIds = ["uuid1", "uuid2"]
  namespaces = ["namespace1"]
}

resource "cast_ai_allocation_group" "my_allocation_group" {
  clusterIds = ["uuid1", "uuid2"]
  namespaces = ["namespace1"]
}

We could probably check the whole allocation group definition instead of just the name and would guard from most of the unintended use cases like:

• Given that the id would be implicit it wouldn't be noticeable in the tf config that those are really two different allocation groups.

• If migration of previously created allocation groups is not done through importing but just by copying everything in the tf files, the client would recreate all their allocation groups which I would say we should guard for.

But I would argue checking the name uniqueness would make sense from the start as I presume it would be confusing to have two allocation groups with the same name showing same or even different statistics.

tl;dr: I will first check if anyone uses non-unique allocation group names, if not I would enforce it.