The prepare.sh script generates a lokocfg file and runs lokoctl to create the cluster.
By default it uses the QEMU mode where it creates two bridges (one external and one internal) and QEMU VMs that PXE-boot.
For real hardware it requires an interface to be specified where the internal L2 network is, and the MAC addresses of the servers and the BMCs in this internal L2 network.
qemu-script: Use a second bridge for an internal network
In the final setup we can use an internal L2 network and don't need to
rely on ipvlan. This makes the QEMU setup resemble the final setup but
still keeps the PXE/DHCP server on the host system, not in a VM.
The required lokoctl binary comes from
kinvolk/lokomotive#1317
qemu-script: Use a default subnet and no ipvlan interface
Instead of picking a subnet we can define one and make it configurable
in case of conflict. Since we know the subnet we can directly assign it
for the QEMU bridge and also remove the ipvlan interface for matchbox.
The prepare.sh script generates a lokocfg file and runs lokoctl to create the cluster. By default it uses the QEMU mode where it creates two bridges (one external and one internal) and QEMU VMs that PXE-boot. For real hardware it requires an interface to be specified where the internal L2 network is, and the MAC addresses of the servers and the BMCs in this internal L2 network.
It uses a lokoctl binary from: https://github.com/kinvolk/lokomotive/pull/1333