Edit

Share via

How to troubleshoot your deployments and monitors in Azure AI Foundry portal

Important

Items marked (preview) in this article are currently in public preview. This preview is provided without a service-level agreement, and we don't recommend it for production workloads. Certain features might not be supported or might have constrained capabilities. For more information, see Supplemental Terms of Use for Microsoft Azure Previews.

This article provides instructions on how to troubleshoot your deployments and monitors in Azure AI Foundry portal.

Deployment issues

For the general deployment error code reference, see Troubleshooting online endpoints deployment and scoring in the Azure Machine Learning documentation. Much of the information there also apply to Azure AI Foundry deployments.

Error: Use of Azure OpenAI models in Azure Machine Learning requires Azure OpenAI in Azure AI Foundry Models resources

The full error message states: "Use of Azure OpenAI models in Azure Machine Learning requires Azure OpenAI in Azure AI Foundry Models resources. This subscription or region doesn't have access to this model."

This error means that you might not have access to the particular Azure OpenAI model. For example, your subscription might not have access to the latest GPT model yet or this model isn't offered in the region you want to deploy to. You can learn more about it on Azure OpenAI in Azure AI Foundry Models.

Error: Out of quota

For more information about managing quota, see:

Error: ToolLoadError

After you deployed a prompt flow, you got the error message: "Tool load failed in 'search_question_from_indexed_docs': (ToolLoadError) Failed to load package tool 'Vector Index Lookup': (HttpResponseError) (AuthorizationFailed)."

To fix this error, take the following steps to manually assign the ML Data scientist role to your endpoint. It might take several minutes for the new role to take effect.

Note

You must use a hub based project for this feature. A Foundry project isn't supported. See How do I know which type of project I have? and Create a hub based project.

  1. Go to your project in Azure AI Foundry and select Management center from the left pane to open the settings page.
  2. Under the Project heading, select Overview.
  3. Under Quick reference, select the link to your resource group to open it in the Azure portal.
  4. Select Access control (IAM) from the left pane in the Azure portal.
  5. Select Add role assignment.
  6. Select Azure ML Data Scientist, and select Next.
  7. Select Managed Identity.
  8. Select + Select members.
  9. Select Machine Learning Online Endpoints in the Managed Identity dropdown field.
  10. Select your endpoint's name.
  11. Select Select.
  12. Select Review + Assign.
  13. Return to your project in Azure AI Foundry portal and select Deployments from the left pane.
  14. Select your deployment.
  15. Test the prompt flow deployment.

Error: Deployment failure

The full error message is as follows:

"ResourceNotFound: Deployment failed due to timeout while waiting for Environment Image to become available. Check Environment Build Log in ML Studio Workspace or Workspace storage for potential failures. Image build summary: [N/A]. Environment info: Name: CliV2AnonymousEnvironment, Version: 'Ver', you might be able to find the build log under the storage account 'NAME' in the container 'CONTAINER_NAME' at the Path 'PATH/PATH/image_build_aggregate_log.txt'."

You might have come across an ImageBuildFailure error: This error happens when the environment (docker image) is being built. For more information about the error, you can check the build log for your <CONTAINER NAME> environment.

This error message refers to a situation where the deployment build failed. You want to read the build log to troubleshoot further. There are two ways to access the build log.

Option 1: Find the build log for the Azure default blob storage.

Note

You must use a hub based project for this feature. A Foundry project isn't supported. See How do I know which type of project I have? and Create a hub based project.

  1. Go to your project in Azure AI Foundry and select Management center from the left pane to open the settings page.
  2. Under the Hub heading, select Overview.
  3. In the section for Connected resources, select the link to your storage account name. This name should be the name of the storage account listed in the error message you received. You'll be taken to the storage account page in the Azure portal.
  4. On the storage account page, select Data Storage > Containers from the left pane.
  5. Select the container name that's listed in the error message you received.
  6. Select through folders to find the build logs.

Option 2: Find the build log within Azure Machine Learning studio.

Note

This option to access the build log uses Azure Machine Learning studio, which is a different portal than Azure AI Foundry.

  1. Go to Azure Machine Learning studio.
  2. Select Endpoints from the left pane.
  3. Select your endpoint name. It might be identical to your deployment name.
  4. Select the link to Environment from the deployment section.
  5. Select Build log at the top of the environment details page.

Error: UserErrorFromQuotaService

The full error message is: "UserErrorFromQuotaService: Simultaneous count exceeded for subscription."

This error message means that the shared quota pool has reached the maximum number of requests it can handle. Try again at a later time when the shared quota is freed up for use.

Question: I deployed a web app but I don't see a way to launch it or find it

We're working on improving the user experience of web app deployment at this time. In the meantime, here's a tip: if your web app launch button doesn't become active after a while, try to deploy it again, using the update an existing app option. If the web app was properly deployed, it should show up on the dropdown list of your existing web apps.

Question: I deployed a model but I don't see it in the playground

Playground only supports select models, such as Azure OpenAI models and Llama-2. If playground support is available, you see the Open in playground button on the model deployment's Details page.

Related content