Microsoft Azure Batch AI Management Client Library

7.0.0 · deprecated · verified Thu Apr 09

The `azure-mgmt-batchai` Python client library provides functionality to manage Azure Batch AI resources. However, the Azure Batch AI service itself was retired on March 31, 2019. The `azure-mgmt-batchai` package is deprecated and received only security fixes until October 31, 2024. Users are strongly advised to migrate to the Azure Machine Learning service and its corresponding SDK, `azureml-core`, for at-scale training and machine learning capabilities. This library is not actively maintained for new features or non-security bug fixes.

Warnings

breaking The underlying Azure Batch AI service has been retired as of March 31, 2019. New subscriptions cannot register for or create Batch AI resources. Existing deployments were subject to a grace period, but the service is no longer supported.
Fix: Migrate your workloads to the Azure Machine Learning service (using the `azureml-core` SDK) or other appropriate Azure compute services.
deprecated The `azure-mgmt-batchai` package is deprecated and will no longer be maintained for new features or non-security bug fixes after October 31, 2024. Only security fixes were provided until that date.
Fix: Avoid using this package for new development. Migrate existing applications to `azureml-core` for machine learning workloads.
breaking Version 7.x introduced a complete revamp of the credential system. Old authentication methods (e.g., `azure.common.credentials`, `msrestazure.azure_active_directory`) are no longer supported.
Fix: Use classes from the `azure-identity` package, such as `DefaultAzureCredential`, for authentication. The `credentials` parameter was also renamed to `credential`.
breaking Significant API specification changes in versions 2.0.0 and 7.x. For example, `Clusters` and `FileServers` must be created under a `Workspace`, `Jobs` under an `Experiment`. Resources no longer accept `location` or `tags` during creation. Method names like `BatchAIManagementClient.usage` were renamed to `BatchAIManagementClient.usages`.
Fix: Refer to the specific version's changelog and documentation for the updated resource hierarchy and method signatures. However, given the service retirement, migration to Azure Machine Learning is the primary recommendation.
gotcha Long-Running Operations (LROs) in version 7.x now return `azure.core.polling.LROPoller` and are typically prefixed with `begin_` (e.g., `begin_create`). This replaces `msrest.polling.LROPoller` from earlier versions.
Fix: Update calls to LRO methods to use the `begin_` prefix and handle the `azure.core.polling.LROPoller` object.

Install

pip install azure-mgmt-batchai Install `azure-mgmt-batchai`
pip install azure-identity Install `azure-identity` (for authentication)

Imports

BatchAIManagementClient

from azure.mgmt.batchai import BatchAIManagementClient

DefaultAzureCredential
```
from azure.identity import DefaultAzureCredential
```
Old credential systems (e.g., `azure.common.credentials`, `msrestazure.azure_active_directory`) are no longer supported in `azure-mgmt-batchai` version 7.x. Use `azure-identity` for unified authentication.

Quickstart

This quickstart demonstrates how to instantiate the `BatchAIManagementClient`. Due to the service's retirement and the SDK's deprecation, actual resource management operations will fail. It uses `DefaultAzureCredential` for authentication, which attempts to authenticate through various methods (environment variables, managed identity, etc.). Set the `AZURE_SUBSCRIPTION_ID` environment variable before running.

import os
from azure.identity import DefaultAzureCredential
from azure.mgmt.batchai import BatchAIManagementClient

# WARNING: The Azure Batch AI service has been retired as of March 31, 2019.
# This SDK package is deprecated and is no longer maintained for new features.
# New subscriptions cannot create Batch AI resources. This code is for illustration only.
# Users should migrate to Azure Machine Learning service (azureml-core).

subscription_id = os.environ.get("AZURE_SUBSCRIPTION_ID", "") # Your Azure Subscription ID
if not subscription_id:
    raise ValueError("AZURE_SUBSCRIPTION_ID environment variable not set.")

# Authenticate with Azure (using recommended DefaultAzureCredential)
credential = DefaultAzureCredential()

# Create a Batch AI Management client
client = BatchAIManagementClient(credential, subscription_id)

print("Azure Batch AI Management Client instantiated.")
print("Note: The Azure Batch AI service has been retired and this SDK is deprecated.")
print("Attempting operations will likely result in service-level errors.")
print("Consider migrating to the Azure Machine Learning service.")

view raw JSON →