Deploying with libvirt

In order to deploy bifrost with libvirt, in order to support managing baremetal servers from with-in that libvirt VM, a special network configuration is required.

Two networks need to be created:

  • default network, that will be a standard virtual network, using NAT.

  • provisioning network, that will be used for PXE boot. As we need to setup a dhcp server on bifrost guest, creating a virtual network will give conflicts between guest and host. So to avoid it, we can define a network that uses macvtap interfaces, associated with the physical interface. Please note that you will need to have macvlan enabled on your kernel.

When creating the guest, a minimum of 8GB of memory is needed in order to build disk images along with run the services to support bifrost.

When defining the interfaces for the guest, the two networks that have been created need to be attached.

These sample commands will spin up a bifrost vm based on centos:

virsh net-define --file tools/virsh_dev_env/network/default.xml
virsh net-start default
virsh net-define --file tools/virsh_dev_env/network/br_direct.xml
virsh net-start br_direct
virsh define --file tools/virsh_dev_env/vm/baremetal.xml
virsh start baremetal
virsh console baremetal

When you login into baremetal, the interface for the provisioning network will be down. You may need to add an IP manually:

ip addr add <<provisioning_ip_address>>/<<mask>> dev <<interface>>
ip link set <<interface>> up

Where to get guest images

In order to create the guest VMs, you will need a cloud image for the distro you want to deploy. You will need to download the guest image on a directory on the host, and then in the template for the VM, you can specify it on the disk section, as shown in the example template.

Please see the OpenStack Image Guide for options and locations for obtaining guest images.

Add credentials to guest image

Normally guest images come without user and password, they rely on ssh to allow access. In this case, it can be useful to enable ssh access to some user from host to guest. A way to do that, is creating a config drive and reference it on the template for the guest VM.

A useful script to generate config drives can be found here.

Relying on this script, a config drive can be created with:

create-config-drive -k ~/.ssh/ config.iso

And then this ISO can be referenced on the guest VM template.