Quick and dirty PXEing of Alpine Linux

I already have some iPXE infra to boot/deploy VMs and bare metal hosts. But it is more Debian/CentOS centric and I decided to throw Alpine Linux into the mix.

Integration was suprisingly easy. boot.alpinelinux.org has good set of resources. Especially nice is that UEFI images are signed.

The simplest solution was just to chainload single boot.ipxe.

chain -ar http://boot.alpinelinux.org/boot.ipxe ||

More robust version could be something like this:

#!ipxe

set alpine-repo http://dl-cdn.alpinelinux.org/alpine/
set version v3.16
set arch x86_64
set flavor virt
set netboot ${alpine-repo}${version}/releases/${arch}/netboot

kernel ${netboot}/vmlinuz-${flavor}
initrd ${netboot}/initramfs-${flavor}

boot vmlinuz-${flavor} \
    debug_init \
    ip=dhcp \
    hostname=${hostname} \
    modloop=${netboot}/modloop-virt \
    alpine_repo=${alpine-repo}${version}/main/ \
    ssh_key=${base-url}/ssh/authorized_keys

There is a page with options you can pass to initramfs, but for example hostname is not supported and the only source of truth is source code.

I still don’t know how I want to proceed from here.

Option one: mirror repo and rely on network boot every time while treating local volume as a persistent storage. Definitely not for mission-critical stuff, but having framework/knowledge for this is definitely nice to have. I have Atomic PI cluster that could be a perfect fit for this approach.
Option two: research/find or craft init script that would do automated install based on /proc/cmdline. Seems like not a rocket science.

Bonus materials#

Bonus materials