Integrating Kube-OVN with OVS‑DPDK for KubeVirt VM DPDK Network Interfaces

This document describes the process of using Kube‑OVN together with OVS‑DPDK to give KubeVirt virtual machines DPDK‑type network interfaces.

Prerequisites : each node must have a NIC dedicated to DPDK drivers and hugepages must be enabled.

Configure the NIC for DPDK using driverctl set-override 0000:00:0b.0 uio_pci_generic . Refer to the DPDK documentation for other drivers.

Label OVS‑DPDK nodes so Kube‑OVN can recognize them:

kubectl label nodes
ovn.kubernetes.io/ovs_dp_type="userspace"

Create the OVS‑DPDK configuration file /opt/ovs-config/ovs-dpdk-config with the following variables:

ENCAP_IP=192.168.122.193/24

DPDK_DEV=0000:00:0b.0

These variables define the tunnel endpoint address and the PCI ID of the DPDK device.

Install Kube‑OVN with DPDK support :

wget https://raw.githubusercontent.com/kubeovn/kube-ovn/release-1.10/dist/images/install.sh

bash install.sh --with-hybrid-dpdk

Deploy the KVM Device Plugin to create VMs:

kubectl apply -f https://raw.githubusercontent.com/kubevirt/kubernetes-device-plugins/master/manifests/kvm-ds.yml

Create a NetworkAttachmentDefinition for the DPDK network:

apiVersion: k8s.cni.cncf.io/v1
kind: NetworkAttachmentDefinition
metadata:
  name: ovn-dpdk
  namespace: default
spec:
  config: >-
    {
        "cniVersion": "0.3.0",
        "type": "kube-ovn",
        "server_socket": "/run/openvswitch/kube-ovn-daemon.sock",
        "provider": "ovn-dpdk.default.ovn",
        "vhost_user_socket_volume_name": "vhostuser-sockets",
        "vhost_user_socket_name": "sock"
    }

Build a VM image using the following Dockerfile:

FROM quay.io/kubevirt/virt-launcher:v0.46.1
# wget http://cloud.centos.org/centos/7/images/CentOS-7-x86_64-GenericCloud.qcow2
COPY CentOS-7-x86_64-GenericCloud.qcow2 /var/lib/libvirt/images/CentOS-7-x86_64-GenericCloud.qcow2

Define the VM and deployment resources (excerpt):

apiVersion: v1
kind: ConfigMap
metadata:
  name: vm-config
data:
  start.sh: |
    chmod u+w /etc/libvirt/qemu.conf
    echo "hugetlbfs_mount = \"/dev/hugepages\"" >> /etc/libvirt/qemu.conf
    virtlogd &
    libvirtd &
    mkdir /var/lock
    sleep 5
    virsh define /root/vm/vm.xml
    virsh start vm
    tail -f /dev/null
  vm.xml: |
... (VM definition with vhostuser interface, hugepages, and resources) ...
---
apiVersion: apps/v1
kind: Deployment
metadata:
  name: vm-deployment
spec:
  replicas: 1
  selector:
    matchLabels:
      app: vm
  template:
    metadata:
      labels:
        app: vm
      annotations:
        k8s.v1.cni.cncf.io/networks: default/ovn-dpdk
    spec:
      nodeSelector:
        ovn.kubernetes.io/ovs_dp_type: userspace
      containers:
      - name: vm
        image: vm-vhostuser:latest
        command: ["bash", "/root/vm/start.sh"]
        resources:
          limits:
            cpu: "2"
            memory: "8784969729"
            hugepages-2Mi: 2Gi
          requests:
            cpu: 666m
            memory: "4490002433"
        volumeMounts:
        - name: vhostuser-sockets
          mountPath: /var/run/vm
        - name: xml
          mountPath: /root/vm/
        - name: hugepage
          mountPath: /dev/hugepages
        - name: libvirt-runtime
          mountPath: /var/run/libvirt

After the VM is created, set its password and access the console:

# virsh set-user-password vm root 12345
# virsh console vm

Configure the VM network inside the guest:

ip link set eth0 mtu 1400
ip addr add 10.16.0.96/16 dev eth0
ip route add default via 10.16.0.1
ping 114.114.114.114

For more details, refer to the official Kube‑OVN documentation at https://kubeovn.github.io/docs/v1.10.x/ .