Azure ML Error:

Question

I keep getting the following error when trying to provision Compute in Azure ML.

The specified Azure ML Compute Instance ABC setup failed with error "Failed to mount storage. Details - Hint:Failed to mount due to permissions issue. Make sure the user/identity has the right permissions like 'Storage File Privileged Contributor' on the storage account. If you have recently updated the role, please try after sometime ". Please delete and try to recreate. If the problem persists, please follow up with Azure Support.

I have granted the suggested permissions on the Azure Storage Account that is in the same Resource Group for both the Managed Instance and the Users identity?

But the issue still occurs?

Answer 1

I remember this issue from an old post where I left an answer. You can check it here !

Azure ML uses a service principal to interact with Azure resources so you need to verify that this principal has the necessary permissions on your Azure Storage Account:

1. Go to your Azure Storage Account in the Azure portal.
2. Go to Access control (IAM).
3. Click on Add role assignment.
4. Select Storage File Data Contributor.
5. Assign this role to the Azure ML service principal.

If your Azure ML workspace or Storage Account is within a Virtual Network (VNet), so you need to check the network configurations and if your VNet allows outbound connections to ports 443 and 445 for the storage service tag.

If your Storage Account has a firewall enabled, confirm that it permits access from your Azure ML workspace:

• In the Azure portal, navigate to your Storage Account.
• Go to Firewalls and virtual networks.
• Ensure that access is allowed from your Azure ML workspace or the specific VNet.

If you are using private endpoints, verifythat DNS settings are correctly configured to resolve the Storage Account's private endpoint.

Misconfigured DNS can lead to mounting failures. For instance, a misconfigured DNS entry for a private endpoint can cause the compute to attempt mounting the storage at an incorrect FQDN/IP, resulting in timeouts.

Links to help you :

https://learn.microsoft.com/en-us/answers/questions/639949/azure-ml-compute-%28instance-or-cluster%29-times-out-m

https://github.com/Azure-Samples/ai-studio-in-a-box/issues/7

https://github.com/Azure/azure-storage-fuse/issues/1606

https://learn.microsoft.com/en-us/answers/questions/1855156/completely-stuck-in-azure-ml-for-creating-compute