Adding vGPU using pyVmomi

[nachiketkarmark]

Hi Everyone

I have been trying to add a vGPU using pyVmomi. I found a way of doing it in powershell from a blog which is as follows:

Param ($VM, $vGPUProfile)

    $VM = Get-VM $VM

    $spec = New-Object VMware.Vim.VirtualMachineConfigSpec

    $spec.deviceChange = New-Object VMware.Vim.VirtualDeviceConfigSpec[] (1)

    $spec.deviceChange[0] = New-Object VMware.Vim.VirtualDeviceConfigSpec

    $spec.deviceChange[0].operation = 'add'

    $spec.deviceChange[0].device = New-Object VMware.Vim.VirtualPCIPassthrough

    $spec.deviceChange[0].device.deviceInfo = New-Object VMware.Vim.Description

    $spec.deviceChange[0].device.deviceInfo.summary = ''

    $spec.deviceChange[0].device.deviceInfo.label = 'New PCI device'

    $spec.deviceChange[0].device.backing = New-Object VMware.Vim.VirtualPCIPassthroughVmiopBackingInfo

    $spec.deviceChange[0].device.backing.vgpu = "$vGPUProfile"

    $vmobj = $VM | Get-View

    $reconfig = $vmobj.ReconfigVM_Task($spec)

    if ($reconfig) {

        $ChangedVM = Get-VM $VM

        $vGPUDevice = $ChangedVM.ExtensionData.Config.hardware.Device | Where { $_.backing.vgpu}

        $vGPUDevice | Select Key, ControllerKey, Unitnumber, @{Name="Device";Expression={$_.DeviceInfo.Label}}, @{Name="Summary";Expression={$_.DeviceInfo.Summary}}

However, I need a pyVmomi equivalent. The problem I am facing it is pyVmomi is not recognizing VirtualPCIPassthroughVmiopBackingInfo class.

Any ideas?

[nachiketkarmark]

The powershell way I was referring to is available in the blog here:

http://www.virtu-al.net/2015/10/26/adding-a-vgpu-for-a-vsphere-6-0-vm-via-powercli/‌

I am able to model removing a vGPU in a python function that uses pyVmomi and it works just fine. However, adding a vGPU results in an error saying:

Incompatible device backing specified for device '0'

Here is my python implementation for adding a vGPU.

spec = vim.vm.ConfigSpec()

spec.deviceChange = [vim.VirtualDeviceConfigSpec()]

spec.deviceChange[0].operation = 'add'

spec.deviceChange[0].device = vim.VirtualPCIPassthrough()

spec.deviceChange[0].device.deviceInfo = vim.Description()

spec.deviceChange[0].device.deviceInfo.summary = 'NVIDIA GRID vGPU '+vgpuProfile

spec.deviceChange[0].device.deviceInfo.label = 'New PCI device'

# spec.deviceChange[0].device.backing = vim.vm.device.VirtualPCIPassthrough.VmiopBackingInfo()

spec.deviceChange[0].device.backing = vim.VirtualPCIPassthroughVmiopBackingInfo()

spec.deviceChange[0].device.backing.vgpu = str(vgpuProfile)

vmMor.ReconfigVM_Task(spec)

I supply vmMor(managed object reference for virtual machine) and vgpuProfile (string indicating profile name such as 'grid_m60-8q')

What am I missing?

[nachiketkarmark · Accepted solution]

FINALLY!! The answer. I had to use

from pyVim.connect import SmartConnect

instead of the legacy way by importing module 'Connect'

from pyVim.connect import Connect

Use SmartConnect and the code below should work