Open iamahern opened 2 years ago
Need this capability for azurerm_machine_learning_compute_cluster
as well, please.
While waiting for the feature to be integrated into azurerm_machine_learning_compute_instance you can leverage azure/azapi provider to target the last ARM schema.
Example:
resource "azapi_resource" "nopip_compute_instance" {
name = "${random_string.ci_prefix.result}instance"
parent_id = azurerm_machine_learning_workspace.default.id
type = "Microsoft.MachineLearningServices/workspaces/computes@2022-06-01-preview"
location = "westeurope"
body = jsonencode({
properties = {
computeType = "ComputeInstance"
disableLocalAuth = true
properties = {
enableNodePublicIp = false
vmSize = "STANDARD_DS2_V2"
subnet = {
id = "${azurerm_subnet.snet-training.id}"
}
}
}
})
depends_on = [
azurerm_private_endpoint.mlw_ple
]
}
Thanks @chboudry for the response. This is in fact what we are doing -- I should have put that here as a workaround. Seems to be working just fine, but certainly not as pretty as an azurerm...
block would be in our template :).
Community Note
Description
In order to secure azure ML resources, teams will seek to use a private endpoint configuration. Currently, virtual machine compute resources created by the AzureRM provider have no means to disable the public IP. https://registry.terraform.io/providers/hashicorp/azurerm/latest/docs/resources/machine_learning_compute_instance#arguments-reference
The feature is currently in preview (through the GUI), so I am unsure if the setting is available via the Azure GO API.
New or Affected Resource(s)
azurerm
machine_learning_compute_instance
Potential Terraform Configuration
N/A
References
Not at this time.