How to register Managed SQL Instance as a datastore in Azure Machine Learning workspace?

Question

I am trying to connect a database within the Managed SQL Instance as a datastore. Managed Instance is not among the datastore type options so I tried to select the Azure SQL database and "enter manually" option for account selection. When I enter the full hostname of the managed instance which is something like "xxx-xxxx-xxx-xxxx-sql.45bsa4569.database.windows.net" it complaints due to dot "." that comes after 'sql'. Since I do not have control over that part of the hostname (Azure adds something like '.45bsa4569' on the managed instance name itself) I do not know how to fix it.

Questions:

1. Is it possible to register a managed instance as a datastore by using the method above? (Azure SQL database and 'enter manually' options).
2. If possible, how to handle the '.' in the host name so that it does not complain?

FYI: The managed instance is behind a vnet so I am using service principal to authenticate.

Answer 1

@azure-machine-learning

Answer 2

[azure-machine-learning]

Answer 3

@Engin Kapti Thanks for the question. Please follow the repo and snapshot.

Supported data storage types:https://learn.microsoft.com/en-us/azure/machine-learning/how-to-access-data#supported-data-storage-service-types

Azure SQL MI is not officially supported per doc .
It's possible to create a datastore on Azure SQL MI, by providing connection string to API register_azure_sql_database, but unexpected behavior may occur.

We have forwarded to the product team and AzureML data team is looking into this issue, and will update if we find any work around.