Hi all, I hope some kind soul here can help me. I am trying to create one Azure machine learning compute instance (the docs are here https://www.pulumi.com/registry/packages/azure-native/api-docs/machinelearningservices/machinelearningcompute/#create-an-computeinstance-compute) Here is the code (every resource needed for this compute instance has been successfully created beforehand): compute = azure_native.machinelearningservices.MachineLearningCompute( _resource_name_="kompjuter", _compute_name_="komputer", _resource_group_name_=rg.name, location=rg.location, _workspace_name_=workspace.name, properties=azure_native.machinelearningservices.ComputeInstanceArgs( _compute_type_="ComputeInstance", # Policy for sharing applications on this compute instance among users of parent workspace. If Personal, only the creator can access applications on this compute instance. When Shared, any workspace user can access applications on this instance depending on his/her assigned role. properties=azure_native.machinelearningservices.ComputeInstancePropertiesArgs( _application_sharing_policy_="Personal", _compute_instance_authorization_type_="personal", # setup_scripts= subnet=vmsubnet.id, _vm_size_="Standard_D8_v4", _personal_compute_instance_settings_=azure_native.machinelearningservices.PersonalComputeInstanceSettingsArgs( _assigned_user_=azure_native.machinelearningservices.AssignedUserArgs( _object_id_="", _tenant_id_="", ) ), _ssh_settings_=azure_native.machinelearningservices.ComputeInstanceSshSettingsArgs( _ssh_public_access_="Disabled", ) ) ) ) The error I am getting is this: azure-nativemachinelearningservicesMachineLearningCompute (kompjuter): error: Code="UserError" Message="The request format was invalid. Details: {0} (Parameter 'Error converting value \"/subscriptions/SUBSCRIPTION_ID/resourceGroups/ml-test/providers/Microsoft.Network/virtualNetworks/mltestnet/subnets/vmsubnet\" to type 'Microsoft.MachineLearning.Common.BatchAIClient.ResourceId'. Path 'properties.subnet', line 1, position 520.')" Details=[] As far as I can tell I am providing the right ID for the subnet, the docs (what little there is) use the same id, and that crappy Pulumi "AI" is of no help as it's either giving the same answer as the docs or unusable junk.

Hi Slobodan! Since Pulumi directly interfaces with the Azure API's, often the Azure documentation itself is of great help, also with regards to this resource: If we go to https://learn.microsoft.com/en-us/azure/templates/microsoft.machinelearningservices/workspaces/computes?pivots=deployment-language-bicep#comp[…]ects You see that the ComputeInstance object requires a subnet ID in the following way: subnet: { id: 'string' } This is also shown if you look at the ComputeInstancePropertiesArgs object:

class ComputeInstancePropertiesArgs:
    def __init__(__self__, *,
                 application_sharing_policy: Optional[pulumi.Input[Union[str, 'ApplicationSharingPolicy']]] = None,
                 compute_instance_authorization_type: Optional[pulumi.Input[Union[str, 'ComputeInstanceAuthorizationType']]] = None,
                 custom_services: Optional[pulumi.Input[Sequence[pulumi.Input['CustomServiceArgs']]]] = None,
                 enable_node_public_ip: Optional[pulumi.Input[bool]] = None,
                 personal_compute_instance_settings: Optional[pulumi.Input['PersonalComputeInstanceSettingsArgs']] = None,
                 schedules: Optional[pulumi.Input['ComputeSchedulesArgs']] = None,
                 setup_scripts: Optional[pulumi.Input['SetupScriptsArgs']] = None,
                 ssh_settings: Optional[pulumi.Input['ComputeInstanceSshSettingsArgs']] = None,
                 subnet: Optional[pulumi.Input['ResourceIdArgs']] = None,
                 vm_size: Optional[pulumi.Input[str]] = None):

tl;dr: set the subnet as subnet=ResourceIdArgs(id=vmsubnet.id) and you're good to go

Thanks for the detailed answer.

What questions do you have about an App Gateway? Note that there’s a gazillion examples of Bicep templates for that which are ultimately the same as Pulumi