[GitLab] Group runners: And why CI is slow in forks

[Carlos Soriano <csoriano gnome org>]

Hello all,

In case you missed my treasurer report talk at GUADEC, our annual costs for CI is around $22.000. This is okay since we are lucky to have GitLab Inc. as the main sponsor of that. However, we cannot offload all of our needs on them, even more in the long-term. So we need to 1) find sponsors, 2) make sure we spend money on just GNOME software.

In order to spend money just in GNOME software we restricted the fast runners to the groups GNOME/ and Infrastructure/. The runners in there are AWS and DO in a elastic set up, that means they get spawned dynamically as the load increases and they are quite beefy. Ideally this would be in a wiki page, with the set up and costs explained.

However there is an issue, forks and MR of forks are no part of this, so they get the local machine we have instead of the fast runners.

I was surprised this wasn't mentioned before, but fear not, it's the most upvoted issue upstream right now https://gitlab.com/gitlab-org/gitlab-ce/issues/23902.

Lately, if you think someone could sponsor part of that money or some machine for the rest of the software not in those groups, please reach to me!

Cheers