Grissess
commented
3 weeks ago Yo, that's a good way to do it; :+1: especially on Ansible, would if I could have known that fleet management could be made a bit easier (even if we're using Chef here :P)
One note: while it might not be set up anymore, FOG ("Free Open-source Ghost") isn't a bad idea for deploying one base image to many places, and has first-class support for Linux deployment (except hostnames--for some weird reason, it only sets hostnames on Wandows, but it's not hard to write a postdownload script to write /etc/hostname to fix this--I did that at Canton for a while). The reason I recommend it is largely because it can do a "multicast deploy", which the ITL network is uniquely suited for--though I recommend you do about half (or even a third) of the room at a time, because multicast joins can sometimes be finicky :)
The release of Ubuntu 24.04 gives us an opportunity to reimage the ITL in a more efficient (read: automated) way.
Canonical's autoinstall can take care of partitioning, users, network configuration, etc.
Ansible can take care of installing software, routine maintenance, and updates/reboots. We can use it for pretty much anything we want to do to a bunch of machines at once.