Unable to use compute cluster in Azure ML

Lê Trịnh Thanh Thúy 5

I'm using Azure for students and I keep getting this error in compute cluster:

Provisioning error
The specified subscription has a total vCPU quota of 0 and cannot accomodate for at least 1 requested managed compute node which maps to 2 vCPUs. Talk to your Subscription Admin or refer to https://docs.microsoft.com/azure/machine-learning/how-to-manage-quotas#request-quota-increases to increase the total quota

I have also checked that there is no other resource group using compute resources and the usage in quota page of VM is 0 cores used.

Can you please help me to fix this error?

User's image

kobulloc-MSFT 25,961 Reputation points Microsoft Employee

2023-05-26T05:30:28.1866667+00:00

Hello, @Lê Trịnh Thanh Thúy !

If I remember correctly, Machine Learning uses a different compute quota than Virtual Machines which can make it a bit hard to track down. I'll look that up but the link provided in the error should give you the quota increase you are looking for (although the limit may be different for student accounts).

2 answers

Paul Kelly 5 Reputation points

2024-02-08T10:27:08.04+00:00

I had this issue with DP-100 training and it seemed to be related to the region. If you try and manually create a compute cluster and adjust the location it will show availability for Standard_DS2_v2 for some of the locations and not others. If you can find a location that allows that type thenyou can edit the setup.sh file in the Cloud shell using 'code .' and then change the list of regions to the one you want. The setup.sh file should now work properly
Please sign in to rate this answer.

1 person found this answer helpful.
kobulloc-MSFT 25,961 Reputation points Microsoft Employee

2024-02-08T16:00:44.5633333+00:00

Hello, @Paul Kelly ! Region availability is very important if you are trying to deploy a popular VM size. Thank you for adding this!
Sign in to comment
kobulloc-MSFT 25,961 Reputation points Microsoft Employee

2023-05-26T05:53:41.5466667+00:00

Hello, @Lê Trịnh Thanh Thúy !

Where can I find my Machine Learning Compute vCPU quota?

You can find your Machine Learning Compute vCPU quota in the portal by searching for Quotas, then clicking on Machine Learning:

In the portal, search for Quotas, then click on Machine Learning:

Is it the same as my Virtual Machine vCPU quota?

No, these are separate quotas for separate services even though they use similar resource types with similar names. Under quotas, you'll see that Compute is also available. This is where you'll find your Virtual Machine vCPU quota which is separate from your Machine Learning quota.

Below you can see Compute/VM vCPU quotas are different than Machine Learning quotas:

How do I request an increase?

The link from the error will direct you to the documentation for increasing your Machine Learning quota:

https://video2.skills-academy.com/en-us/azure/machine-learning/how-to-manage-quotas?view=azureml-api-2#request-quota-increases

You'll need to select "Machine Learning Service: Virtual Machine Quota" as the quota type you want to increase.

I hope this helps, @Lê Trịnh Thanh Thúy !

If you still need assistance, let us know in the comments.

If this has been useful, please take a moment to accept answers as this helps increase visibility of this question for other members of the Microsoft Q&A community. Thank you for helping to improve Microsoft Q&A!
Please sign in to rate this answer.
Lê Trịnh Thanh Thúy 5 Reputation points

2023-05-26T06:22:02.12+00:00

Thanks for your replying!

However, I checked for the Machine Learning Compute vCPU quota and the usage is 0%.

You can check the picture below.

Moreover, when I request a new quota increase, it said that "Your subscription isn't eligible for a quota increase. To request a quota increase, first upgrade to a Pay-As-You-Go subscription."

kobulloc-MSFT 25,961 Reputation points Microsoft Employee

2023-05-26T07:14:40.09+00:00

Thank you for the update, @Lê Trịnh Thanh Thúy!

You are correct, that shows 0% usage. Machine Learning is listed on the student account page so while I would expect Machine Learning resources to be supported, I know that there have been reports of new and promotional accounts which have been unable to use high demand resources (like AI targeted GPU VMs) during surges:

Before implementing any restrictions on customers, we first pull every lever available to expedite new capacity, and to preserve existing capacity – this includes shaping internal customers to non-constrained regions whenever possible. If these steps are not sufficient to avoid a capacity shortfall, we then implement “offer restrictions” which prevent newly-created subscriptions from being able to deploy to the constrained region. Should you be experiencing an allocation failure, there are many ways to troubleshoot. We also work closely with our impacted customers to offer alternate solutions, should a capacity constraint affect their deployment.

You still have the option of submitting a support ticket which would rule out any issues that may be unique to your account. Let me know if you need any help with this and I'd be happy to assist!

Lê Trịnh Thanh Thúy 5 Reputation points

2023-05-26T07:57:03.0066667+00:00

Can you please tell me if there are any alternative ways that I can send a request for my problem issues? Because I use Azure's student account so it's only available for basic support requests and unavailable for Technical issues.

kobulloc-MSFT 25,961 Reputation points Microsoft Employee

2023-05-27T00:31:47.43+00:00

Hello, @Lê Trịnh Thanh Thúy!

Requesting a quota increase is the right path to take however I recently tried to use Azure credits to experiment with AI on a GPU equipped VM and was given this response:

“You recently contacted us requesting access to the [N/GPU] Series virtual machines through your free/benefit/sponsorship subscription. Unfortunately, we are unable to approve this request. Due to high demand for these graphics-enabled VM types, availability is limited for customers using free/benefit subscriptions. Please consider alternative VM types that may be suitable for your requirements or consider creating a new pay-as-you-go subscription to explore expanded access through a separate paid subscription.”

If you received a similar response, then availability is currently limited for customers using free/benefit/sponsorship subscriptions due to high demand.

The good news is that new datacenters are constantly being added to address the surge in demand so hopefully this will be a temporary measure.
Sign in to comment

Share via

Unable to use compute cluster in Azure ML

2 answers