Starting with VMware Integrated OpenStack 3.1, you can create OpenStack instances that use GPU physical functions (enabled using directpath I/O) or virtual functions (SR-IOV) from vSphere.
About this task
Consumption of GPU and passthrough features is achieved by using the appropriate flavor. Modify the metadata parameters of the flavor to create the instance.
Prerequisites
Make sure that you perform the following settings in your environment before you can configure GPU passthrough devices:
Enable DirectPath I/O in vSphere. See the DirectPath I/O chapter in the VMware vSphere 6.5 Documentation.
Enable SR-IOV on GPU devices on your ESXi hosts. See Configuring AMD Multiuser GPU Using vDGA in the VMware Horizon documentation.
Procedure
- Log in to the OpenStack Management Server.
- Create custom.yml file if it does not exist.
sudo mkdir -p /opt/vmware/vio/custom
sudo cp /var/lib/vio/ansible/custom/custom.yml.sample
/opt/vmware/vio/custom/custom.yml
- Open the /opt/vmware/vio/custom/custom.yml file in a text editor.
- Create PCI alias using VIO customization by editing the custom.yml file according to your configuration.
- Edit
nova_pci_alias
value to create PCI alias based on device_type
, vendor_id
, and product_id
and name the alias, for example:
nova_pci_alias: ["product_id": "692f", "vendor_id": "1002", “device_type:” “type-VF”, "name": "gpu-vf"}]
- Save the custom.yml file.
- Push the new configuration to your VMware Integrated OpenStack deployment.
Refresh of the configuration briefly interrupts the OpenStack services.
viocli deployment configure --tags nova_api_config