Setting up Virtualization on penguin

Contents

Installing the virtualization RPMs on penguin was straightforward:

for PACKAGE in qemu-kvm libvirt virt-install virt-viewer
do
    yum install -y $PACKAGE 2>&1 | jtime | tee ~/packages/install.$PACKAGE.text
done

Installing the FreePBX distribution

A network bridge is needed for the VM, but when I checked penguin one had already been created by the installation of qemu-kvm:

# nmcli
virbr0: connected to virbr0
  "virbr0"
  bridge, 52:54:00:C1:42:3D, sw, mtu 1500
  inet4 192.168.122.1/24
  route4 192.168.122.0/24

Note, however, the bridge has ip address 192.168.122.1/24, meaning it has its own sub-net. That’s not what I want: I want the guest system to be able to talk to the local network as a normal system. See Setting up a bridge for transparent networking on libvirt guests for details.

Install the VM:

virt-install --name freepbx --cpu host --memory 2048 --disk size=8 \
  --cdrom /var/local/iso/misc/FreePBX-64bit-1904-2.iso \
  --network bridge=virbr0 --os-variant rhel7

Typically I run commands on penguin by making an ssh connection to the system, but this time I had to run the command in a terminal from the graphical desktop on the penguin itself. This is because by default virt-install starts virt-viewer, and details on running virt-viewer remotely are hard to come by.

But there’s another way to get the console on sparrow, by connecting to penguin using ssh -Y penguin. That forwards my X11 connection to the server, letting sparrow’s X11 handle the window.

Setting up a bridge for transparent networking on libvirt guests

It took a ridiculous amount of effort to come up with a networking configuration that allowed libvirt domains to participate in my LAN as a normal system as opposed to being isolated on penguin. The official libvirt documentation is practically silent on this issue. Searching on the web turned up a couple of variations on binding a bridge device to a tun/tap interface, but my attempts to use these were not successful.

In the end I had to do the following:

• Remove all IP configuration from the network card eno1
• Create a bridge br0 using the same the IP configuration eno1 had used
• Add eno1 as an interface on br0
• Also assign 192.168.1.3 (used for WAN domain resolution) to br0

In theory that’s what the tap/tun stuff is supposed to be doing without having to muck around with IP addresses on the primary NIC. I’m at a loss to explain why those approaches didn’t work.

Connecting to a VM’s display on a remote system

I ran the following command on my laptop:

virt-viewer --debug --direct --connect qemu+ssh://penguin/system

The above command brought up a list of the running VMs. I selected the only one listed, and the viewer gave an error:

(virt-viewer:3402): virt-viewer-DEBUG: Guest CentOS-7-A has a spice display
(virt-viewer:3402): virt-viewer-DEBUG: Guest graphics address is 127.0.0.1:5900
(virt-viewer:3402): virt-viewer-DEBUG: graphics listen '127.0.0.1' is not reachable from this machine

Note that penguin told my laptop to connect to 127.0.0.1 and not its IP address of 192.168.1.1.