setboxes

setboxes does the following:

1. creates a backingStore qcow2 image that mimicks the arch cloud images
   • differences are that we target only EFI, our rootfs is ext4 and do not handle for cloud-init etc.
2. spins up a libVirt pool and network, creates volumes that overlay the base image, resizes these volumes to byte-perfectly fit the partitions in the baremetal targets, and spins up the domains.
3. these domains get configured once using both the init.yml, which performs one-off tasks for ansible to be ran proper such as the sshd configuration and the creation of the ansible_user, and by site.yml, which is the actual state of these baremetal targets that we aim for.
4. once the VMs get configured, they are shut off and with the help of partclone and qemu-nbd, these corresponding partitions for the targets are imaged.
5. targets get booted up using iPXE, the resulting images get written and inside a chroot the initramfs and efi bootloader entries are generated.
6. once rebooted, the site.yml playbook is called on them once again to finalize configuration.
7. from then on, the baremetal targets can be configured fully using ansible.

stages

# 1. provision the backingStore image and deploy the VMs (as root)
cd provision/
pacman -Syyuu # base img creation requires an arch host as it uses pacstrap
./mkbaseimg.sh
./setlibvirt.sh
cd ../

# 2. install the ansible requirements
ansible-galaxy install -r requirements.yml

# 3. init the configuration of the VMs
ansible-playbook --flush-cache init.yml site.yml

# 4. cut the qcow2's to compressed partclone images
cd provision/
./mkimg.sh

# 5. restore partclone images in baremetal with:
#    unzstd --stdout -k $image \
#       | partclone.$fstype -r -s - -o $part
#
#    then chroot to rootfs and call:
#    ./provision/setbootmgr.sh $machine

# 6. handle baremetal targets once the images are written
ansible-playbook --flush-cache site.yml