Jump to content

Talk:GitLab/Gitlab Runner/Shared Runners

Rendered with Parsoid
From Wikitech

Why configure each instance individually?

In the docs how to setup a new runner we say to configure each instance individually with the puppet role class and Hiera data. Wouldn't it be simpler if we just used project-wide puppet config (prefix puppet) once and told it to apply the role to anything called gitlab-runner* and then nobody has to ever do it again? And for the Hiera data, we could just add it to the repo and nothing has to be done in Horizon either. Thoughts? dzahn (talk) 19:44, 6 December 2022 (UTC)Reply

The project wide config for role application seems reasonable to me. +1
Another factor in easing setup would be to figure out how to secure set/store the registration token. The way it was done for the shared runners was to set up a standalone puppetmaster and set the token value in /etc/puppet/secret/hieradata/common.yaml, but IMO the burden of setting up a standalone puppetmaster is much higher than setting roles for example. Is there some better way to set up tokens as well? Dduvall (talk) 16:55, 7 December 2022 (UTC)Reply