Deploy MLflow Model to Endpoint Without Egress Connectivity

Note: This question is part of a series of questions that present the same scenario. Each question in the series contains a unique solution that might meet the stated goals. Some question sets might have more than one correct solution, while others might not have a correct solution. After you answer a question in this section, you will NOT be able to return to it. As a result, these questions will not appear in the review screen. You have an Azure Machine Learning workspace named Workspace1. Workspace1 has a registered MLflow model named model1 with PyFunc flavor. You plan to deploy model1 to an online endpoint named endpoint1 without egress connectivity by using Azure Machine Learning Python SDK v2. You have the following code: You need to add a parameter to the ManagedOnlineDeployment object to ensure the model deploys successfully. Solution: Add the with_package parameter. Does the solution meet the goal? -

Yes Source Reference Answer
No

Community Votes

67%

33%

67% of anonymous learners picked answer A. Votes are pick records left by other test-takers — they are not the verified answer.

Community Insight

The exam tests knowledge of deploying MLflow models in air-gapped or restricted environments; the trap is assuming standard deployment works without explicit packaging configuration.

This question tests the configuration of Azure Machine Learning online endpoints for workspaces with restricted network access. The community consensus confirms that using the with_package parameter is the correct solution to enable model packaging for offline deployment.

Candidates often select 'No' because they are unaware of the `with_package` parameter or believe a scoring script is mandatory, leading them to reject the solution incorrectly.

Community Discussion (7 comments)

astone42 👍 1 Selected: A

"Workspaces without public network access: Before you can deploy MLflow models to online endpoints without egress connectivity, you have to package the models (preview). By using model packaging, you can avoid the need for an internet connection, which Azure Machine Learning would otherwise require to dynamically install necessary Python packages for the MLflow models."

Ben999 👍 1 Selected: A

without egress connectivity blue_deployment = ManagedOnlineDeployment( name="blue", endpoint_name=endpoint_name, model=model, instance_type="Standard_F4s_v2", instance_count=1, with_package=True, )

D0ktor 👍 1 Selected: B

No param called like that

f2a9aa5 👍 3

A. To deploy an MLflow model to an online endpoint in Azure Machine Learning without egress connectivity, you can use model packaging. Here’s how: First, ensure that your workspace has no public network access. Package your MLflow model using the --with-package flag: az ml online-deployment create --with-package --endpoint-name $ENDPOINT_NAME -f blue-deployment.yml --all-traffic Replace $ENDPOINT_NAME with your desired endpoint name. This approach allows you to avoid the need for an internet connection while deploying MLflow models. https://learn.microsoft.com/en-us/azure/machine-learning/how-to-deploy-mlflow-models-online-endpoints?view=azureml-api-2&tabs=cli

cryodax 👍 3

The ManagedOnlineDeployment class requires the following parameters: name: str: Name of the deployment resource. model: str | Model | None: Model entity for the endpoint deployment, defaults to None. code_configuration: CodeConfiguration | None: Code Configuration, defaults to None. environment: str | Environment | None: Environment entity for the endpoint deployment, defaults to None. These are the minimum required parameters to create an instance of the ManagedOnlineDeployment class. All other parameters are optional and have default values. Please note that while model, code_configuration, and environment are optional in the constructor, they are typically necessary for a successful deployment. If not provided in the constructor, they should be set before deployment.

cryodax 👍 1

Sorry, scoring_script is optional. Copilot response: When using the ManagedOnlineDeployment class with the pyfunc flavor, the parameters remain the same. However, the model and code_configuration parameters are particularly important: model: str | Model | None: This should be an instance of mlflow.pyfunc.PyFuncModel or the URI of the model saved with mlflow.pyfunc.save_model1. code_configuration: CodeConfiguration | None: This should be an instance of CodeConfiguration where the code_flavor is set to pyfunc.

cryodax 👍 2

B. No Scoring script is a requirement for managedonlinedeployment

Comments & Corrections

No comments yet — spotted an error or have a note? Share it below.

Expert Analysis

Why the Answer Is Correct

The with_package parameter in the ManagedOnlineDeployment object instructs Azure Machine Learning to bundle all necessary dependencies with the model artifact. This is essential when the workspace lacks egress connectivity, as it prevents the service from attempting to download packages from the internet during deployment.

Why the Other Options Are Wrong

Selecting 'No' would imply that another parameter is required or that this approach fails. While some candidates argue for a scoring script, documentation confirms it is optional for many model types if the environment handles inference correctly. Therefore, rejecting the with_package solution is incorrect.

Community Comment Notes

Comment [4] provides the official Microsoft documentation context, explaining that model packaging avoids the need for an internet connection. Comment [5] offers a code snippet demonstrating the correct usage of with_package=True. Comment [3] incorrectly claims a scoring script is required, which contradicts current SDK behavior for PyFunc models.

Official Reference

https://learn.microsoft.com/en-us/azure/machine-learning/how-to-deploy-mlflow-models-without-internet-access

Exam Strategy

Memorize specific parameters for edge-case scenarios like air-gapped deployments. When you see 'no egress connectivity' or 'no public network access', immediately look for options related to 'packaging' or 'bundling' dependencies.

Related Analysis

← Back to DP-100 Study Guide