Azure Labs VMs in existing lab will not start

VinceO 26 Reputation points
2023-12-29T14:53:59.8266667+00:00

We are unable to start Azure Labs Services VMs of a particular size in the eastus region. The size is "Small GPU (Compute)|6 cores | 56GB RAM". We are able to start other sizes which we use.

We first noticed this issue yesterday afternoon and it is continuing to happen this morning.

The Activity Log includes the following message: "Allocation failed. We do not have sufficient capacity for the requested VM size in this region."

How long does it normally take Microsoft to deal with this type of issue? Is this size being phased out?

Azure Lab Services
Azure Lab Services
An Azure service that is used to set up labs for classrooms, trials, development and testing, and other scenarios.
279 questions
{count} votes

3 answers

Sort by: Most helpful
  1. Luke Murray 10,526 Reputation points MVP
    2024-01-04T07:59:58.7266667+00:00

    Hi, Vince

    There are a few things to consider when talking about capacity, ie are you using a pay-as-you-go subscription (this takes priority) vs a dev-test or Visual Studio subscription, is the region you are in support availability zones, etc?

    If the issue isn't self-resolved by now, your best bet is to:

    Raise a Support Case[​](https://luke.geek.nz/2022/07/27/microsoft-azure-zonalallocationfailed/#raise-a-support-case"Direct link to Raise a Support Case")

    • Take a screenshot of the error
    • Copy the Activity/Deployment ID
    • Take note of the Region
    • Take note of the Availability Zone.

    Let Azure Support know; that Microsoft may already be aware, but raising a support request helps identify potentially impacted customers. If you know of other SKUs you need to deploy, you can let them know.

    Purchase On-demand Capacity Reservation[​](https://luke.geek.nz/2022/07/27/microsoft-azure-zonalallocationfailed/#purchase-on-demand-capacity-reservation"Direct link to Purchase On-demand Capacity Reservation")

    [On-demand Capacity Reservation](https://learn.microsoft.com/en-us/azure/virtual-machines/capacity-reservation-overview?WT.mc_id=AZ-MVP-5004796"On-demand Capacity Reservation") enables you to reserve Compute capacity in an Azure region or an Availability Zone for any duration of time.

    Unlike [Reserved Instances](https://azure.microsoft.com/en-us/pricing/reserved-vm-instances/?WT.mc_id=AZ-MVP-5004796"Reserved Instances"), you do not have to sign up for a 1-year or a 3-year term commitment.

    Once the Capacity Reservation is created, the capacity is available immediately and is exclusively reserved for your use until the reservation is deleted.

    Capacity Reservations are priced at the same rate as the underlying VM size.

    For example, if you create a reservation for the D2s_v3 VMs, you will start getting billed for the D2s_v3 VMs, even if the reservation is not being used.

    So why would you purchase On-demand Capacity reservations?

    • You are operating Azure workloads that scale out and run off a fresh image, like a Citrix farm and want to ensure the capacity is available for the minimum workloads you need.
    • You have a project coming up where you need the capacity to be available.

    Redeploy to another Availability Zone[​](https://luke.geek.nz/2022/07/27/microsoft-azure-zonalallocationfailed/#redeploy-to-another-availability-zone"Direct link to Redeploy to another Availability Zone")

    The server cluster that ARM (Azure Resource Manager) attempted to deploy your workload may not have the necessary capacity, but another Availability Zone (datacenter) might.

    Make sure your Virtual Machine is not in a Proximtry or Avalibility Group and do the following.

    1. Take note of the Availability Zone that your deployment failed (i.e. Availability Zone 1)
    2. Remove any resources that may have been created as part of the original failed deployment.
    3. Redeploy your workload and select another Availability Zone, such as (2 - if your failed deployment was in Zone 1)

    Change the Virtual Machine version[​](https://luke.geek.nz/2022/07/27/microsoft-azure-zonalallocationfailed/#change-the-virtual-machine-version"Direct link to Change the Virtual Machine version")

    By version, I don't mean [Generation 1 and Generation 2](https://learn.microsoft.com/en-us/azure/virtual-machines/generation-2?WT.mc_id=AZ-MVP-5004796#features-and-capabilities"Generation 1 vs. generation 2 features") Virtual Machines; I mean the version of underlying Compute; when you look at a VM SKU size, you will see:

    Standard_DC24s_v3 [Family] + [Sub-family]* + [# of vCPUs] + [Constrained vCPUs]* + [Additive Features] + [Accelerator Type]* + [Version]

    You can read more about Virtual Machine Naming conversions "[here](https://learn.microsoft.com/en-us/azure/virtual-machines/vm-naming-conventions?WT.mc_id=AZ-MVP-5004796"Azure virtual machine sizes naming conventions")".

    The version of the VM series links to the underlying hardware associated with the Virtual Machine series; with most new hardware releases, the version changes; an example is: from v3 to v4.

    #Tip: Microsoft may run a promotion on the pricing for early adopters from time to time, to move to the new version; they can be seen from the Azure Portal with "Promo" in the name.

    1. You can change the version of the SKU by looking in the Azure Portal, Sizing, and you should be select different versions of the same SKU; if you are at v5, try resizing to v4 - or the other way around.

    Remember that changing the VM SKU will force the Virtual Machine to deallocate (stop), as it triggers ARM to stand up the Virtual Machine on different server clusters/hardware.

    I have found that there are no noticeable decreases in performance for most workloads, but keep in mind you may be returning on older hardware - but it should get you going, and then you can update the SKU to the latest version later.

    Increase regional vCPU quotas[​](https://luke.geek.nz/2022/07/27/microsoft-azure-zonalallocationfailed/#increase-regional-vcpu-quotas"Direct link to Increase regional vCPU quotas")

    Azure Resource Manager enforces two types of vCPU quotas for virtual machines:

    standard vCPU quotas

    spot vCPU quotas Standard vCPU quotas apply to pay-as-you-go VMs and reserved VM instances. They are enforced at two tiers, for each subscription, in each region:

    The first tier is the total regional vCPU quota.

    The second tier is the VM-family vCPU quota such as D-series vCPUs.

    Check your subscription quotas and if necessary, raise a request to increase them by following the guide here: [Increase regional vCPU quotas]

    Reference: https://luke.geek.nz/azure/microsoft-azure-zonalallocationfailed/

    0 comments No comments

  2. VinceO 26 Reputation points
    2024-01-04T21:58:42.4866667+00:00

  3. VinceO 26 Reputation points
    2024-01-05T14:45:22.4666667+00:00

    Looks like Microsoft deprecated some of the machine series used by GPU sized Lab Services VMs. My older labs using "Small GPU (Compute)|6 cores | 56GB RAM" VMs probably used the deprecated series. I created new "Small GPU (Compute)|6 cores | 56GB RAM" labs and things were back on track.

    https://techcommunity.microsoft.com/t5/azure-lab-services-blog/outage-gpu-labs-that-use-lab-accounts/ba-p/4020895

    This gives a little insight into which machine series are used by Lab Services

    https://learn.microsoft.com/en-us/azure/lab-services/administrator-guide-1#vm-sizing

    0 comments No comments