Simon Josefsson <si...@josefsson.org> writes:
> Ludovic Courtès <l...@gnu.org> writes:
>
>>> How about adding another ppc64el build machine on a fresh host (I have
>>> another Talos II around and finally got stable network connectivity for
>>> it) -- while developing the documentation/instructions -- and then once
>>> completed and operational, tear down the current sjd-p9 machine and
>>> re-install it using the new instructions?
>>
>> We could do that yes. And yes, documentation on how to set up
>> individual build machines specially for ci.guix (in this case) probably
>> needs to be written.
>
> Who could write this documentation? I can volunteer to be a guinnea pig
> to try to read the documentation and follow instructions so that a new
> ppc64el VM gets setup, so we later can recycle the current one with a
> more modern install. However I have a very vague understanding of how
> the Guix server-side build system works so I can't help write this.
If we’re talking about connecting a build machine to ci.guix.gnu.org
(remember that the process is different for bordeaux.guix.gnu.org), the
steps go along these lines:
1. Set up WireGuard to connect the machine on the ci.guix VPN (there
are details under hydra/.)
2. Add the machine’s /etc/guix/signing-key.pub under
hydra/keys/guix/berlin in maintenance.git.
3. Register hydra/keys/guix/berlin.guixsd.org-export.pub with ‘guix
archive --authorize’ on the build machine.
4. Set up and start ‘cuirass remote-worker’ on the build machine.
Details defer depending on whether the new build machine runs Guix
System (in that case it’s easy) or another distro (in that case one
first needs to install Guix, then install the ‘cuirass’ package, drop
the .service files in the right place, etc.).
HTH!
Ludo’.
