Starting with Avi Load Balancer version 22.1.1, installing Linux KVM with DPDK Support for virtio-net data NICs and macvtap interface (in bridge mode) created from the physical interface available on the host, is supported. This macvtap interface is employed as the interface within the guest virtual machine.

For more information on macvtap interface, see MacVTap.

Prerequisites

  1. Hardware

    Check that your CPU supports hardware virtualization to run KVM. You need a processor that supports hardware virtualization. Intel and AMD have both developed extensions for their processors, deemed respectively Intel VT-x and AMD-V. To check if your processor supports one of these, you can review the output using this command: egrep -c '(vmx|svm)' /proc/cpuinfo

    1. If the value is zero, it means that your CPU does not support hardware virtualization.

    2. If the value is one or more, it means that your CPU supports hardware virtualization. However, you need to ensure that virtualization is enabled in the BIOS.

  2. Software Installation

    1. Ubuntu Distro

      For more information on KVM Installation and requisite packages installation, see KVM/Installation.

    2. RHEL/CentOS Distro

      For more information on Installing KVM packages on an existing Red Hat Enterprise Linux system and for requisite packages installation, see KVM with a new Red Hat Enterprise Linux installation.

  3. Software [Tested and Qualified]

    1. Ubuntu Distro

      Ubuntu ID

      Version

      DISTRIB_ID

      Ubuntu

      DISTRIB_RELEASE

      18.04

      DISTRIB_CODENAME

      Bionic

      DISTRIB_DESCRIPTION

      Ubuntu 18.04 LTS

      DISTRIB_ID

      Ubuntu

      OS kernel version

      4.15.0-20-generic

      libvirt-bin

      libvirtd (libvirt) 4.0.0

      qemu-kvm

      QEMU emulator version 2.11.1

      genisomage

      genisoimage 1.1.11 (Linux)

    2. RHEL/ CentOS Distro

      RHEL/CentOS ID

      Version

      CentOS Linux release

      7.9 (Maipo)

      OS kernel version

      5.4.17-2136.305.5.4.el7uek.x86_64

      libvirt-bin

      libvirtd (libvirt) 4.5.0

      qemu-kvm

      QEMU emulator version 2.0.0

      genisomage

      genisoimage 1.1.11 (Linux)

Installing Service Engine and Controller Virtual Machine

  1. Prerequisites: Ensure that you copy Avi Load Balancerse.qcow2 and controller.qcow2 images to /var/lib/libvirt/images/ directory of the host machine. se.qcow2 image can be fetched from the Controller UI once it is up, as explained in the Deploying Service Engine section below.

  2. Script Run: Run the install script mentioned in kvm-dpdk-virt-install.sh on the host machine for the Controller and the SE virtual machines installation.

    Note:

    You need to create the Controller virtual machine before creating the SE virtual machine.

Deploying the Controller

Run the above-mentioned install script on the host machine choosing Create AVI Controller VM option for deploying the Controller virtual machine.

These are some recommendations for the variables used to install script while creating the Controller virtual machine:

  1. Avi Load Balancer Controller management IP

  2. Avi Load Balancer Controller Management-IP Mask

  3. Avi Load Balancer Controller Default Gateway

The values for these variables are to be derived from the host management interface configuration.

Controller Initial Setup

The following are the steps to navigate to the Controller IP address to perform the initial setup:

  • Configure an administrator password

  • Set DNS information

  • Select No Orchestrator

Deploying Service Engine

Once the Controller virtual machine is up, the following are the steps to upload the SE image for installing Service Engine:

  1. On the Controller, navigate to Infrastructure > Clouds.

  2. Click the download icon on Default Cloud row and select Qcow2.

  3. Upload the se.qcow2 to the /var/lib/libvirt/images/ directory of the host machine.

Run the above-mentioned install script on the host machine choosing Create AVI SE VM option for deploying SE virtual machine.

These are some recommendations for the variables used to install script while creating the Service Engine virtual machine:

  • Avi Load Balancer SE Management IP

  • Avi Load Balancer Controller Management-IP Mask

  • Avi Load Balancer SE Default Gateway

The values for these variables are to be derived from the host management interface configuration.

Note:

For the SE virtual creation, it is recommended a 4-queue configuration per macvtap interface and virtual CPUs configuration of greater than or equal to four for best throughput performance.

The bond interfaces sequence can be specified as follows:

Bond-ifs sequence

1,2 3,4. This implies interface 1,2 are in bond and interface 3,4 are in bond (Note the space between 1,2 and 3,4).

Bond-if sequence

1,2,3,4. This implies interface 1,2,3,4 are in bond.

You can verify the SE to connect to the Controller by navigating to Infrastructure > Dashboard on the Controller UI (this might take a few minutes).

Sample SE Virtual Machine Installation Output

  1. On the Host

    1. Host management interface (from which management macvtap interface is created).

        eno1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc mq state UP group default qlen 1000
  link/ether 0c:c4:7a:b4:15:b4 brd ff:ff:ff:ff:ff:ff
  inet 10.217.144.19/22 brd 10.217.147.255 scope global eno1

    2. Host physical interfaces (from which data macvtap interfaces are created).

      ens1f0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc mq state UP group default qlen 1000
link/ether 0c:c4:7a:bb:96:ea brd ff:ff:ff:ff:ff:ff
inet 100.64.50.56/24 brd 100.64.50.255 scope global dynamic ens1f0
ens1f1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc mq state UP group default qlen 1000
link/ether 0c:c4:7a:bb:96:eb brd ff:ff:ff:ff:ff:ff
inet 100.64.67.55/24 brd 100.64.67.255 scope global dynamic ens1f1

    3. Macvtap interfaces as seen on the host (post install script run).

        macvtap0@eno1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc fq_codel state UP group default qlen 500
  link/ether 52:54:00:f4:ed:9f brd ff:ff:ff:ff:ff:ff
  macvtap1@ens1f0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc fq_codel state UP group default qlen 500
  link/ether 52:54:00:4f:3a:2d brd ff:ff:ff:ff:ff:ff
  macvtap2@ens1f1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc fq_codel state UP group default qlen 500
  link/ether 52:54:00:68:38:ef brd ff:ff:ff:ff:ff:ff

  2. On the SE virtual machine post install script run

    1. SE management interface

        eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc mq state UP group default qlen 1000
  link/ether 52:54:00:f4:ed:9f brd ff:ff:ff:ff:ff:ff
  inet 10.217.144.251/22 brd 10.217.147.255 scope global eth0

    2. SE data-nic interfaces (as seen in namespace post ip-address assignment)

      root@Avi-Service-Engine:/home/admin# ip netns exec avi_ns1 bash
avi_eth2: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc fq_codel state UP group default qlen 1000
link/ether 52:54:00:68:38:ef brd ff:ff:ff:ff:ff:ff
inet 100.64.67.22/24 brd 100.64.67.255 scope global dynamic avi_eth2
avi_eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc fq_codel state UP group default qlen 1000
link/ether 52:54:00:4f:3a:2d brd ff:ff:ff:ff:ff:ff
inet 100.64.50.41/24 brd 100.64.50.255 scope global dynamic avi_eth0

  3. Post Host Reboot

    The following enlists steps to be executed post every reboot.

    1. Post reboot, all virtual machines will automatically be in the STOP state.

    2. All the virtual machine names can be checked using virsh list output.

    3. Bring up all the virtual machines using virsh start <VM-name> command.

  4. Destroying the Controller and Service Engines

    The virtual machines and their corresponding images can be cleared using the above-mentioned script.

    Note:

    1. If the disk space of the host is getting exhausted it might be due to the N number of qcow2s being used in /var/lib/libvirt/images/ as part of the virtual machine’s creation. Clear up any unused ones by deleting the respective virtual machine as per above section.

    2. Ensure to manually cleanup (force-delete) the stale Service Engine entries present in the Controller GUI post destroying SE virtual machines.