2.4 Verify Azure Quota

This solution contains an Azure Developer CLI azd-template that provisions the required resources in Azure and deploys the starter app to Azure Container Apps (ACA). The template allows for the infrastructure to be deployed with a single azd up command.

• Verified your Azure ML CPU quota
• Authenticated with Azure
• Provisioned Azure resources and deployed the starter solution

Verify Azure ML CPU Quota

This solution accelerator contains a section dedicated to setting up and using a Semantic Ranking model directly from your PostgreSQL database. The deployment of this component of the architecture requires sufficient CPU quota (32 cores) in Azure Machine Learning to accommodate the Hugging Face BGE reranker model deployment. In this task, you must verify you have available quota for the target virtual machine (VM) instance type (STANDARD_D16AS_V4), and if not, request additional quota.

1. To view your available quota, you first need to retrieve your Microsoft Entra ID Tenant ID from the Azure portal.

2. In the Azure portal, enter "Microsoft Entra ID" into the search bar, then select Microsoft Entra ID from the Services list in the results.

   

3. On the Overview page of your Microsoft Entra ID tenant, select the Copy to clipboard button for your Tenant ID.

   

4. Open a new browser window or tab and navigate to the following URL, replacing the <your-tenant-id> token with the Tenant ID you copied from the Entra ID overview page in the Azure portal.

   Azure ML Quota page

   https://ml.azure.com/quota?tid=<your-tenant-id>
   

5. On the Azure ML Quota page, select the subscription you are using for this workshop.

   

6. On the quota page for your selected subscription, select the Azure region you plan to use for this workshop. This should be the region you chose in previous task that supports the required Azure OpenAI models.

7. You should now see a list of CPUs and their quotas within your subscription. Locate Standard DASv4 Family Cluster Dedicated vCPUs in the list and inspect the Quota available.

   

8. If you have 32 cores or more available, then you do not need to request a quota increase. Otherwise, select the Standard DASv4 Family Cluster Dedicated vCPUs by checking the box to the left of the name, then scroll up to the top of the page and locate the Request quota button.

   

9. In the Request quota dialog, increase your New cores limit value by 32 and then select Submit.

   

   Quota increase example

   Your new cores limit should be increased to ensure 32 cores are available for a new deployment. For example, if you have zero cores available, your new cores limit should be set to 32. If your core limit is 100 and you are currently using 90, your new cores limit should be set to 122.

10. Quota increase requests typically take a few minutes to complete. You will receive notifications in the Azure portal as the request is processed and when it completes.

11. If your request is denied, you don't have permissions to issue the request, or you prefer not to request additional quota, you have the option to exclude the Semantic Ranking model deployment when running the azd up command by setting the deployAMLModel flag to false when prompted.

Verify Azure Resource Providers

There are a few Azure Resource Providers that will need to be registered on the Azure Subscription for the solution accelerator to successfully deploy the Azure Machine Learning resources.

The required Azure Resource Providers are:

• Microsoft.Cdn
• Microsoft.PolicyInsights
• Microsoft.MachineLearningServices
• Microsoft.ApiManagement

Follow these steps to check if the Resource Providers are registered, and if not then you'll register them:

1. Run the following command to check whether the Resource Providers are registered on your Azure Subscription:

   az provider list --query "[?namespace=='Microsoft.Cdn' || namespace=='Microsoft.PolicyInsights' || namespace == 'Microsoft.MachineLearningServices' || namespace == 'Microsoft.ApiManagement'].{Namespace: namespace, RegistrationState: registrationState}" -o table
   

   Alternatively, you could also navigate to the Subscription within the Azure Portal, then navigate to Resource providers under Settings. This will also allow you to view the registered Resource Providers for the Subscription, as well as register them.

   The console output will look similar to the following:

   Namespace                          RegistrationState    
   ---------------------------------  -------------------  
   Microsoft.MachineLearningServices  NotRegistered        
   Microsoft.Cdn                      Registered           
   Microsoft.PolicyInsights           NotRegistered        
   Microsoft.ApiManagement            NotRegistered        
   

   The console output will show the Resource Provider and the RegistrationState. If the RegistrationState shows a value of Registered then the Resource Provider is registered on the Azure Subscription.

2. To register all the Resource Providers on the Azure Subscription, run the following commands:

   az provider register --namespace Microsoft.MachineLearningServices
   az provider register --namespace Microsoft.Cdn
   az provider register --namespace Microsoft.PolicyInsights
   az provider register --namespace Microsoft.ApiManagement
   

   If one or more of the Resource Providers are already registered, then only run the command for the Resource Providers that are not registered.