how to provide secrets to an azure machine learning inference server

user_6401281 0
Viewed 159

Within our azure machine learning inference server we want to be able to make calls to a password protected database. What is the best way to provide a secret to the running inference server?

I have tried following this tutorial: https://docs.microsoft.com/en-us/azure/machine-learning/how-to-use-secrets-in-runs. However the request to run = Run.get_context() always returns an offline instance. Presumably that is only used for passing secrets to models when training them?

I was also thinking I could pass my secrets in as env variables but I'm not sure if that is safe?

env = Environment.from_pip_requirements(env_name, "requirements-azure.txt")
env.environment_variables.update({"mysecret": "password"})

Any suggestions are greatly appreciated!

azure-machine-learning-service
1 Answers

Here is link to Access Azure resources from a managed online endpoint (preview) with a managed identity and Azure Machine Learning inference HTTP server.

Currently the following feature is in the roadmap to Provide deployed web service some kind of easy access path to the associated Key Vault, similar to how an experimental run of a model can use Run.get_context().get_secret().

user_112974060
Related