Can not create compute instance in Azure ML studio

Question

Can not create compute instance in Azure ML studio

Aleh Yablonski 0

Can not create compute instance in Azure ML studio. The creation was failed after 6 hours
User's image

kothapally Snigdha 3,020 Reputation points Microsoft External Staff Moderator

2025-02-20T23:31:24.6333333+00:00

Hi Aleh Yablonski,

Can you please share me the REGION Details.
kothapally Snigdha 3,020 Reputation points Microsoft External Staff Moderator

2025-02-20T23:32:50.5966667+00:00

Hi Aleh Yablonski,

Can you please share the REGION Details.

Thank you!
Aleh Yablonski 0 Reputation points

2025-02-21T07:34:12.95+00:00

region WEST US
kothapally Snigdha 3,020 Reputation points Microsoft External Staff Moderator

2025-02-21T22:00:31.3133333+00:00
Hi Aleh Yablonski,

I understand that you're facing an issue with to create a compute instance. Here are some steps to follow.

Ensure your Azure subscription is active and has the necessary permissions to create compute instances.

Verify that you've created a resource group in the correct region where you intend to create the compute instance.

Sometimes, simply switching to a different browser can resolve the problem.

A common cause is a misconfigured storage account. Check that the storage account associated with your Azure ML workspace is correctly set up and linked. Confirm that the Azure ML service principal has the necessary permissions to access and manage the storage account.

This error often indicates an issue with the storage configuration that Azure ML creates for the workspace.

This suggests a problem with the access keys. Regenerate the keys in the Azure portal and update your application with the new keys5.

This indicates that you do not have enough cores for the selected virtual machine size in your region

I hope these helps you. Thank you!
Aleh Yablonski 0 Reputation points

2025-02-24T06:29:26.74+00:00

Thank you for the answer. But this is not the case.
I can create any other instance without problems. Only instance with GPU H100 could not be created
kothapally Snigdha 3,020 Reputation points Microsoft External Staff Moderator

2025-02-26T16:33:55.3166667+00:00

Hi Aleh Yablonski,

Please see if you have sufficient quota on GPU H100 SKU type. You can find quota details by below cli command. If you don't have sufficient quota, you can request for increase quota for H100 computes. If you have quota, check for other available regions by pass regional constraints. You find compute quota either through CLI command " az login --identity, az ml compute list-sizes -g <your RG name> -w <your AML workspace name> computelist.json" or Compute tab itself by clicking on "view my quota" as show in screenshot.

I hope these helps you Thank you!
kothapally Snigdha 3,020 Reputation points Microsoft External Staff Moderator

2025-02-28T15:45:39.1533333+00:00

Hi Aleh Yablonski,

Following up to see if the above response was helpful.

1 answer

Your answer

kothapally Snigdha 3,020 Reputation points Microsoft External Staff Moderator

2025-02-20T23:31:24.6333333+00:00

Hi Aleh Yablonski,

Can you please share me the REGION Details.
kothapally Snigdha 3,020 Reputation points Microsoft External Staff Moderator

2025-02-20T23:32:50.5966667+00:00

Hi Aleh Yablonski,

Can you please share the REGION Details.

Thank you!
Aleh Yablonski 0 Reputation points

2025-02-21T07:34:12.95+00:00

region WEST US
kothapally Snigdha 3,020 Reputation points Microsoft External Staff Moderator

2025-02-21T22:00:31.3133333+00:00

Hi Aleh Yablonski,

I understand that you're facing an issue with to create a compute instance. Here are some steps to follow.

Ensure your Azure subscription is active and has the necessary permissions to create compute instances.

Verify that you've created a resource group in the correct region where you intend to create the compute instance.

Sometimes, simply switching to a different browser can resolve the problem.

A common cause is a misconfigured storage account. Check that the storage account associated with your Azure ML workspace is correctly set up and linked. Confirm that the Azure ML service principal has the necessary permissions to access and manage the storage account.

This error often indicates an issue with the storage configuration that Azure ML creates for the workspace.

This suggests a problem with the access keys. Regenerate the keys in the Azure portal and update your application with the new keys5.

This indicates that you do not have enough cores for the selected virtual machine size in your region

I hope these helps you. Thank you!
Aleh Yablonski 0 Reputation points

2025-02-24T06:29:26.74+00:00

Thank you for the answer. But this is not the case.
I can create any other instance without problems. Only instance with GPU H100 could not be created
kothapally Snigdha 3,020 Reputation points Microsoft External Staff Moderator

2025-02-26T16:33:55.3166667+00:00

Hi Aleh Yablonski,

Please see if you have sufficient quota on GPU H100 SKU type. You can find quota details by below cli command. If you don't have sufficient quota, you can request for increase quota for H100 computes. If you have quota, check for other available regions by pass regional constraints. You find compute quota either through CLI command " az login --identity, az ml compute list-sizes -g <your RG name> -w <your AML workspace name> computelist.json" or Compute tab itself by clicking on "view my quota" as show in screenshot.

I hope these helps you Thank you!
kothapally Snigdha 3,020 Reputation points Microsoft External Staff Moderator

2025-02-28T15:45:39.1533333+00:00

Hi Aleh Yablonski,

Following up to see if the above response was helpful.

Answer 1

Hi Aleh Yablonski,

From the screenshot, it looks like your Azure ML compute instance creation has failed, with the error: "Provisioning error: An internal server error occurred. If the problem persists, contact support."

Here are a few possible causes and solutions to try:

1. Region Capacity Issues

The Standard_NC40ads_H100_v5 VM is a powerful GPU instance that may not be available in your selected region due to capacity constraints.

Solution: Try selecting a different region where the VM is available. You can check Azure VM quotas and availability via Azure Quota Portal

2. Subscription Quotas

You may not have the quota or permissions to create H100 instances.

Solution: Go to Subscriptions → Usage + quotas → Request Increase for NCas_H100 family. Make sure you’re using a subscription that supports GPU SKUs (some trial or student plans might not).

3. Internal Azure Error

Sometimes, Azure fails for internal reasons (especially for high-demand SKUs).

Solution: Try deleting the failed instance and recreating it after a short wait. Switch to a different compute type temporarily (e.g., Standard_NC6, Standard_NDv2, etc.) to continue your work.

4. Azure Support Ticket

If the issue persists even after trying different regions or SKUs:

Solution: Go to Help + support → New support request in the Azure Portal. Choose "Quota" or "Virtual Machines (compute)" and provide the error and region details.

Share via

Can not create compute instance in Azure ML studio

1 answer

Your answer