Virtual GPUs

Kayobe contains playbooks to configure virtualised GPUs on supported NVIDIA hardware. This allows you to statically create mdev devices that can be used by Nova to present a virtualised GPU to guest VMs. Known working GPUs are:

  • NVIDIA A100

BIOS configuration

Intel

  • Enable VT-x in the BIOS for virtualisation support.

  • Enable VT-d in the BIOS for IOMMU support.

AMD

  • Enable AMD-V in the BIOS for virtualisation support.

  • Enable AMD-Vi in the BIOS for IOMMU support.

Example: Dell

Enabling SR-IOV with racadm:

/opt/dell/srvadmin/bin/idracadm7 set BIOS.IntegratedDevices.SriovGlobalEnable Enabled
/opt/dell/srvadmin/bin/idracadm7 jobqueue create BIOS.Setup.1-1
<reboot>

Enabling CPU Virtualization with racadm:

/opt/dell/srvadmin/bin/idracadm7 set BIOS.ProcSettings.ProcVirtualization Enabled
/opt/dell/srvadmin/bin/idracadm7 jobqueue create BIOS.Setup.1-1
<reboot>

Obtain driver from NVIDIA licensing portal

Download NVIDIA GRID driver from here (This requires a login).

Configuration

See also

For further context, please see:

Add hosts with supported GPUs to the compute-vgpu group. If using bifrost and the kayobe overcloud inventory discover mechanism, this can be achieved with:

$KAYOBE_CONFIG_PATH/overcloud.yml
overcloud_group_hosts_map:
  compute-vgpu:
    - "computegpu000"

Configure the location of the NVIDIA driver:

$KAYOBE_CONFIG_PATH/vgpu.yml
 ---
 vgpu_driver_url: "https://example.com/NVIDIA-GRID-Linux-KVM-525.105.14-525.105.17-528.89.zip"

If you don’t know which vGPU types your card supports, these can be determined by following NVIDIA vGPU types.

You can then define group_vars describing the vGPU configuration:

$KAYOBE_CONFIG_PATH/inventory/group_vars/compute-vgpu/vgpu
#nvidia-692 GRID A100D-4C
#nvidia-693 GRID A100D-8C
#nvidia-694 GRID A100D-10C
#nvidia-695 GRID A100D-16C
#nvidia-696 GRID A100D-20C
#nvidia-697 GRID A100D-40C
#nvidia-698 GRID A100D-80C
#nvidia-699 GRID A100D-1-10C
#nvidia-700 GRID A100D-2-20C
#nvidia-701 GRID A100D-3-40C
#nvidia-702 GRID A100D-4-40C
#nvidia-703 GRID A100D-7-80C
#nvidia-707 GRID A100D-1-10CME
vgpu_definitions:
  # Configuring a MIG backed VGPU
  - pci_address: "0000:17:00.0"
    mig_devices:
      # This section describes how to partition the card using MIG. The key
      # in the dictionary represents a MIG profile supported by your card and
      # the value is the number of MIG devices of that type that you want
      # to create. The vGPUS are then created on top of these MIG devices.
      # The available profiles can be found in the NVIDIA documentation:
      # https://docs.nvidia.com/grid/15.0/grid-vgpu-user-guide/index.html#virtual-gpu-types-grid-reference
      "1g.10gb": 1
      "2g.20gb": 3
    virtual_functions:
      # The mdev type is the NVIDIA identifier for a particular vGPU. When using
      # MIG backed vGPUs these must match up with your MIG devices. See the NVIDIA
      # vGPU types section in this document.
      - mdev_type: nvidia-700
        index: 0
      - mdev_type: nvidia-700
        index: 1
      - mdev_type: nvidia-700
        index: 2
      - mdev_type: nvidia-699
        index: 3
  # Configuring a card in a time-sliced configuration (non-MIG backed)
  - pci_address: "0000:65:00.0"
    virtual_functions:
      - mdev_type: nvidia-697
        index: 0
      - mdev_type: nvidia-697
        index: 1

To apply this configuration, use:

(kayobe) $ kayobe overcloud host configure -t vgpu

NVIDIA vGPU types

See also

For further context, please see:

The NVIDIA vGPU drivers must be installed to be able to query for the available vGPU types. This can be achieved by not defining any virtual functions in the vGPU definition:

$KAYOBE_CONFIG_PATH/inventory/group_vars/compute-vgpu/vgpu
vgpu_definitions:
  - pci_address: "0000:17:00.0"
    virtual_functions: []

See Configuration. You can then use mdevctl to query for the available vGPU types.

mdevctl types

Kolla Ansible configuration

To use the mdev devices that were created, modify nova.conf to add a list of mdev devices that can be passed through to guests:

$KAYOBE_CONFIG_PATH/kolla/config/nova/nova-compute.conf
{% raw %}
{% if inventory_hostname in groups['compute-vgpu'] %}
[devices]
enabled_mdev_types = nvidia-700, nvidia-699, nvidia-697
[mdev_nvidia-700]
device_addresses = 0000:17:00.4,0000:17:00.5,0000:17:00.6
mdev_class = CUSTOM_NVIDIA_700
[mdev_nvidia-699]
device_addresses = 0000:17:00.7
mdev_class = CUSTOM_NVIDIA_699
[mdev_nvidia-697]
device_addresses = 0000:65:00.4,0000:65:00.5
mdev_class = CUSTOM_NVIDIA_697
{% endif %}
{% endraw %}

You will need to adjust the PCI addresses to match the virtual function addresses. These can be obtained by checking the mdevctl configuration after applying the configuration:

# mdevctl list
73269d0f-b2c9-438d-8f28-f9e4bc6c6995 0000:17:00.4 nvidia-700 manual (defined)
dc352ef3-efeb-4a5d-a48e-912eb230bc76 0000:17:00.5 nvidia-700 manual (defined)
a464fbae-1f89-419a-a7bd-3a79c7b2eef4 0000:17:00.6 nvidia-700 manual (defined)
f3b823d3-97c8-4e0a-ae1b-1f102dcb3bce 0000:17:00.7 nvidia-699 manual (defined)
330be289-ba3f-4416-8c8a-b46ba7e51284 0000:65:00.4 nvidia-700 manual (defined)
1ba5392c-c61f-4f48-8fb1-4c6b2bbb0673 0000:65:00.5 nvidia-700 manual (defined)
f6868020-eb3a-49c6-9701-6c93e4e3fa9c 0000:65:00.6 nvidia-700 manual (defined)
00501f37-c468-5ba4-8be2-8d653c4604ed 0000:65:00.7 nvidia-699 manual (defined)

The mdev_class maps to a resource class that you can set in your flavor definition. Note that if you only define a single mdev type on a given hypervisor, then the mdev_class configuration option is silently ignored and it will use the VGPU resource class (See bug 1943934).

To apply the configuration to Nova:

(kayobe) $ kayobe overcloud service deploy -kt nova

OpenStack flavors

Define some flavors that request the resource class that was configured in nova.conf. An example definition, that can be used with openstack.cloud.compute_flavor Ansible module, is shown below:

openstack.cloud.compute_flavor:
  name: "vgpu.a100.2g.20gb"
  ram: 65536
  disk: 30
  vcpus: 8
  is_public: false
  extra_specs:
    hw:cpu_policy: "dedicated"
    hw:cpu_thread_policy: "prefer"
    hw:mem_page_size: "1GB"
    hw:cpu_sockets: 2
    hw:numa_nodes: 8
    hw_rng:allowed: "True"
    resources:CUSTOM_NVIDIA_700: "1"

Changing VGPU device types

Converting the second card to an NVIDIA-698 (whole card). The hypervisor should be empty so we can freely delete mdevs. If this is not the case you will need to check which mdevs are in use and use extreme caution. First clean up the mdev definitions to make room for the new device:

[stack@computegpu000 ~]$ sudo mdevctl list
5c630867-a673-5d75-aa31-a499e6c7cb19 0000:21:00.4 nvidia-697 manual (defined)
eaa6e018-308e-58e2-b351-aadbcf01f5a8 0000:21:00.5 nvidia-697 manual (defined)
72291b01-689b-5b7a-9171-6b3480deabf4 0000:81:00.4 nvidia-697 manual (defined)
0a47ffd1-392e-5373-8428-707a4e0ce31a 0000:81:00.5 nvidia-697 manual (defined)

[stack@computegpu000 ~]$ sudo mdevctl stop --uuid 72291b01-689b-5b7a-9171-6b3480deabf4
[stack@computegpu000 ~]$ sudo mdevctl stop --uuid 0a47ffd1-392e-5373-8428-707a4e0ce31a

[stack@computegpu000 ~]$ sudo mdevctl undefine --uuid 0a47ffd1-392e-5373-8428-707a4e0ce31a
[stack@computegpu000 ~]$ sudo mdevctl undefine --uuid 72291b01-689b-5b7a-9171-6b3480deabf4

[stack@computegpu000 ~]$ sudo mdevctl list --defined
5c630867-a673-5d75-aa31-a499e6c7cb19 0000:21:00.4 nvidia-697 manual (active)
eaa6e018-308e-58e2-b351-aadbcf01f5a8 0000:21:00.5 nvidia-697 manual (active)

# We can re-use the first virtual function

Secondly remove the systemd unit that starts the mdev device:

[stack@computegpu000 ~]$ sudo rm /etc/systemd/system/multi-user.target.wants/nvidia-mdev@0a47ffd1-392e-5373-8428-707a4e0ce31a.service
[stack@computegpu000 ~]$ sudo rm /etc/systemd/system/multi-user.target.wants/nvidia-mdev@72291b01-689b-5b7a-9171-6b3480deabf4.service

Adapt your Kayobe and Kolla Ansible configuration to match the desired state and then re-run host configure:

(kayobe) $ kayobe overcloud host configure --tags vgpu --limit computegpu000

Check the result:

[stack@computegpu000 ~]$ mdevctl list
5c630867-a673-5d75-aa31-a499e6c7cb19 0000:21:00.4 nvidia-697 manual
eaa6e018-308e-58e2-b351-aadbcf01f5a8 0000:21:00.5 nvidia-697 manual
72291b01-689b-5b7a-9171-6b3480deabf4 0000:81:00.4 nvidia-698 manual

Reconfigure nova to match the change:

(kayobe) $ kayobe overcloud service reconfigure -kt nova --kolla-limit computegpu000 --skip-prechecks