Attaching or Detaching an fGPU

You can attach a flexible GPU (fGPU) to a virtual machine (VM), and detach it at any time.

To be able to attach an fGPU, you must first allocate it to your account. For more information, see Allocating or Releasing an fGPU.

Linking or Unlinking an fGPU Using Cockpit v2

Linking an fGPU

  1. Click inside the Flexible GPUs dashboard to make checkboxes appear.

  2. Check the box of the fGPU you want to link to a VM.
    The fGPU is selected and an action menu appears.

  3. Click IconLink Link to a VM.
    The LINK TO A VM dialog box appears.

  4. In the VM list, select the VM you want to link the fGPU to.

  5. Click Link.
    The fGPU is scheduled to be linked to the VM.

    To complete the linking of the fGPU, you need to stop the VM, unless it is already stopped.

Unlinking a fGPU

  1. Click inside the Flexible GPUs dashboard to make checkboxes appear.

  2. Check the box of the fGPU you want to unlink from a VM.
    The fGPU is selected and an action menu appears.

  3. Click IconUnlink Unlink VM.
    A confirmation dialog box appears.

  4. Click Unlink.
    The fGPU is scheduled to be unlinked from its VM.

    To complete the unlinking of the fGPU, you need to stop the VM, unless it is already stopped.

Attaching or Detaching an fGPU Using Cockpit v1

Cockpit v1 is no longer fixed nor supported. For more information, see End-of-Life Policy.

Attaching an fGPU

  1. Click Compute > Flexible GPU.

  2. Click the fGPU you want to attach to an instance.

  3. Click Attach .
    The ATTACH INSTANCE dialog box appears.

  4. In the Instance list, select the instance you want to attach the fGPU to.

  5. Click Attach to validate.
    The fGPU is scheduled for attachment to the instance.

    To complete the attachment of the fGPU, you need to stop the instance, unless it is already stopped.

Detaching an fGPU

  1. Click Compute > Flexible GPU.

  2. Click the fGPU you want to detach from an instance.

  3. Click Detach .
    The DETACH FLEXIBLE GPU dialog box appears.

  4. Click Detach to validate.
    The fGPU is scheduled for detachment from its instance.

    To complete the detachment of the fGPU, you need to stop the instance, unless it is already stopped.

Linking or Unlinking an fGPU Using OSC CLI

Linking an fGPU

The LinkFlexibleGpu command attaches one of your allocated flexible GPUs (fGPUs) to one of your virtual machines (VMs).
The fGPU is in the attaching state until the VM is stopped, after which it becomes attached.

Request sample
$ osc-cli api LinkFlexibleGpu --profile "default" \
    --FlexibleGpuId "fgpu-12345678" \
    --VmId "i-12345678"

This command contains the following attributes that you need to specify:

  • DryRun: (optional) If true, checks whether you have the required permissions to perform the action.

  • FlexibleGpuId: The ID of the fGPU you want to attach.

  • VmId: The ID of the VM you want to attach the fGPU to.

The LinkFlexibleGpu command returns the following elements:

  • ResponseContext: Information about the context of the response.

    • RequestId: The ID of the request.

Result sample
{
  "ResponseContext": {
    "RequestId": "0475ca1e-d0c5-441d-712a-da55a4175157"
  }
}

Unlinking an fGPU

The UnlinkFlexibleGpu command detaches a flexible GPU (fGPU) from a virtual machine (VM).
The fGPU is in the detaching state until the VM is stopped, after which it becomes available for allocation again.

Request sample
$ osc-cli api UnlinkFlexibleGpu --profile "default" \
    --FlexibleGpuId "fgpu-12345678"

This command contains the following attributes that you need to specify:

  • DryRun: (optional) If true, checks whether you have the required permissions to perform the action.

  • FlexibleGpuId: The ID of the fGPU you want to detach from your VM.

The UnlinkFlexibleGpu command returns the following elements:

  • ResponseContext: Information about the context of the response.

    • RequestId: The ID of the request.

Result sample
{
  "ResponseContext": {
    "RequestId": "0475ca1e-d0c5-441d-712a-da55a4175157"
  }
}

Related Pages

Corresponding API Methods

Export to PDF