I am trying to deploy a model to a managed online endpoint in Azure Machine Learning.
(Along the lines of https://learn.microsoft.com/en-us/azure/machine-learning/how-to-deploy-managed-online-endpoints).
This works fine with a publically accessible AML workspace, but not with our workspace inside our private VNET.
With an AML workspace/storage account in a private VNET the deployment fails:
I found that this is directly related to the network setting of the storage account. This is the setting that fails. Note that allowing Azure Services does not mitigate the problem:
Is this is blind spot of managed endpoints, which is simply not yet supported or is this a bug?
The problem can be reproduced with the sample code at https://github.com/Azure/azureml-examples/tree/main/cli/endpoints/online/managed/sample i.e.
az ml online-deployment create --name blue -f endpoints/online/managed/sample/blue-deployment.yml
The 'troubleshooting' guide in the error message refers to the importance of accessibility of the storage account (and Azure Container Registry), but does not consider the usecase, where AML is inside a private VNET: https://learn.microsoft.com/en-us/azure/machine-learning/how-to-troubleshoot-online-endpoints?tabs=cli#authorization-error