400 Client Error when calling resources.check_existence/check_existence_by_id on virtual machine

[dharmab]

When calling resources.check_existence or resource.check_existence_by_id on a resource of type Microsoft.Compute/virtualMachines, the following exception is raised:

msrestazure.azure_exceptions.CloudError: 400 Client Error: Bad Request for url: https://management.azure.com/subscriptions/REDACTED_GUID/resourceGroups/REDACTED/providers/Microsoft.Compute/virtualMachines/REDACTED?api-version=2017-05-10

I've tested with both (azure-mgmt-resource==1.2.1, api_version=='2017-05-10') and (azure-mgmt-resource==2.0.0rc2, api_version='2018-05-10').

[lmazuel]

Hi @dharmab
I know it's confusing, but the api-version to use when using any methods in the "resources" operation group is the one of the provide, not the generic one of ARM.
If you case, this should be (random example) 2017-12-01

client.resources.check_existence_by_id(
    '/subscriptions/REDACTED_GUID/resourceGroups/REDACTED/providers/Microsoft.Compute/virtualMachines/REDACTED',
    '2017-12-01'
)

Could you confirm?

[dharmab]

Ah, that makes sense! I retried using the same API version that the VM was provisioned with, and don't get the 400 error. However, I now get a 405:

  File "/usr/local/lib/python3.6/site-packages/azure/mgmt/resource/resources/v2017_05_10/operations/resources_operations.py", line 793, in check_existence_by_id
    raise exp
msrestazure.azure_exceptions.CloudError: 405 Client Error: Method Not Allowed for url: https://management.azure.com/subscriptions/REDACTED_GUID/resourceGroups/REDACTED/providers/Microsoft.Compute/virtualMachines/REDACTED?api-version=2017-12-01

A code snippet to demonstrate my usage:

    vm_resource_args = {
        'resource_id': '/subscriptions/{}/resourceGroups/{}/providers/Microsoft.Compute/virtualMachines/{}'.format(config['Subscription'], config['ResourceGroup'], vm_name),
        'api_version': azutil.resource.ApiVersion.VMSS.value
    }

    try:
        if resource_client.resources.check_existence_by_id(**vm_resource_args):
            logger.info(f'Deleting VM: {vm_name}')
            poller = resource_client.resources.delete_by_id(**vm_resource_args)
            __wait_for_long_azure_operation('Deletion', poller=poller)
        else:
            logger.info(f'Could not find etcd VM: {vm_name} (previously deleted?)')
    except CloudError as e:
        __log_inner_azure_errors_and_rethrow(e)

[lmazuel]

"check existence" is implemented as HTTP HEAD, which is basically a HTTP GET without body (for performance). Not all RP supports this.
The approach I would suggest is to try "check existence", and if not allowed try a "get_by_id". Something like

def resource_exists(**vm_resource_args):
    try:
        return resource_client.resources.check_existence_by_id(**vm_resource_args)
    except CloudError as e:
        if e.status_code == 405: # HEAD not supported
            try:
                resource_client.resources.get_by_id(**vm_resource_args)
                return True
            except CloudError:
                return False  # Likely 404, might want to test it explicitly
        raise # If not 405, not expected

This should still raise for unrelated problems (connection, credentials, etc.), and return True/False if applicable.
Note that it's not an official sample, it's just an example of the possible approach if you want to try the approach HEAD/GET.

[dharmab]

Adapting the get_by_id fallback worked great! Is it possible for check_existence_by_id to fallback?

In any case, the workaround is usable and I can accomplish what I initially wanted to do. Thanks!