azure-sdk-for-python: AzureML: User managed identity attached to compute cluster does not work

• Package Name: azure-identity
• Package Version: 1.5.0
• Operating System: Linux
• Python Version: 3.7

Describe the bug

A custom docker container running on a compute cluster with a user managed identity does not seem to authenticate with Azure identity when running in AzureML

This is the logs I when I run my python code:

WARNING 2020-12-21 18:04:30,166 azure.identity._internal.get_token_mixin AppServiceCredential.get_token failed: No token received.
WARNING 2020-12-21 18:04:30,166 azure.identity._internal.decorators ManagedIdentityCredential.get_token failed: No token received.
WARNING 2020-12-21 18:04:30,166 azure.identity._credentials.chained ChainedTokenCredential.get_token failed: ManagedIdentityCredential raised unexpected error "No token received."
WARNING 2020-12-21 18:04:30,166 azure.identity._credentials.chained ChainedTokenCredential failed to retrieve a token from the included credentials.
Attempted credentials:
	ManagedIdentityCredential: No token received.

To Reproduce Steps to reproduce the behavior:

1. Create a compute cluster with a managed identiy attached to read data from a blob storage
2. Create a custom docker image with the azure sdk to fetch data from blob storage
3. Run the aforementioned docker image in the AzureML cluster created above.

Expected behavior Python SDK can authenticate using a manged identity.