v3 - Managed Identity Service Principals not immediately available for use - Request_ResourceNotFound - 404.

[jamesw4]

Community Note

• Please vote on this issue by adding a 👍 reaction to the original issue to help the community and maintainers prioritise this request
• Please do not leave "+1" or "me too" comments, they generate extra noise for issue followers and do not help prioritise the request
• If you are interested in working on this issue or have submitted a pull request, please leave a comment

Terraform (and AzureAD Provider) Version

Terraform v1.9.8
on windows_amd64

• provider registry.terraform.io/hashicorp/azuread v3.0.2
• provider registry.terraform.io/hashicorp/azurerm v4.8.0

Affected Resource(s)

• azuread_group_member

Terraform Configuration Files

provider "azurerm" {
  features {

  }
  subscription_id = "##REDACTED##"
}

resource "azurerm_resource_group" "test" {
  name     = "test"
  location = "uksouth"
}

resource "azurerm_user_assigned_identity" "test" {
  name                = "test"
  resource_group_name = azurerm_resource_group.test.name
  location            = azurerm_resource_group.test.location
}

data "azuread_group" "test" {
  display_name = "jwtest"
}

resource "azuread_group_member" "test" {
  group_object_id  = data.azuread_group.test.object_id
  member_object_id = azurerm_user_assigned_identity.test.principal_id
}

Debug Output

https://gist.github.com/jamesw4/e163427ae58ebb24b2688a59759c255a

Panic Output

Expected Behavior

Managed identity to be created, and the underlying service principal added as a group member.

Actual Behavior

Most of the time, for initial deploy there seems to be a delay. The managed identity gets created, but AzureAD provider complains that the service principal does not exist to be added as a member to the group. This sounds like same situation as mentioned in ticket #1035. Re-running the plan and apply results in it succeeding.

Gist contains logs for provider 3.0.2 failing, then the re-run of it succeeding. Also provided logs from 2.53.1 where the issue does not seem to occur, seems to be a difference in the version of Graph API used.

Have rolled back to 2.53.1 of the provider for now.

Steps to Reproduce

1. Create an Entra group to be used as data source (jwtest in the example given)
2. terraform apply

Important Factoids

References

• azuread_directory_role_assignment dependent on azurerm_user_assigned_identity incurred unexpected status 404 with OData error: Request_ResourceNotFound: Resource 'xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxx' does not exist or one of its queried reference-property objects are not present. #1035

[bartoszkubiakpl]

I can add that we have the same issue. For time being setting up time sleep resolved problem but still waiting for proper solution. In our case below code generated error message:

resource "time_sleep" "wait_30_seconds_after_module.adf" {
  depends_on      = [module.adf]
  create_duration = "30s"

  triggers = {
    # Deplay the time at each switch of an module.adf.adf_id
    wait_for = module.adf.adf_id
  }
}


# add ADF MI as member to ACL writer group
resource "azuread_group_member" "adf_acl_sg_membership" {
  count            = var.create_adf == "True" ? 1 : 0
  group_object_id  = data.azuread_group.sg_acl_writer.object_id
  member_object_id = module.adf[0].adf.identity[0].principal_id
  depends_on       = [time_sleep.wait_30_seconds_after_module.adf]
}

# add ADF MI as member to cicd group
resource "azuread_group_member" "adf_cicd_sg_membership" {
  count            = var.create_adf == "True" ? 1 : 0
  group_object_id  = data.azuread_group.sg_principal.object_id
  member_object_id = module.adf[0].adf.identity[0].principal_id
  depends_on       = [time_sleep.wait_30_seconds_after_module.adf]
}