Are there existing examples when there's a configuration that's available for Self-managed instances to override (admin setting) but is hidden(?) on Dedicated and controlled by the Dedicated tooling?

@dmeshcharakou Yes this is where it gets tricky. We have settings we control but right now the practice is simply to tell customers not to change these. There's a larger discussion I don't want to drag into this blueprint on whether we should be doing any kind of "isDedicated ? hide_setting : show_setting" type of thing for Application settings such as this. What we generally tell customers is that if you change these settings you will break functionality and even so they will be reset on the next maintenance.

Examples where we currently control the setting but it IS available to customer admins:

• Enable/disable and various settings on Advanced search
• User and IP rate limits
• Runner creation and deletion (for customers using Dedicated Hosted Runners if they are using their own runners those runners would not be under the "don't touch this setting" rule).
• Initial license application
• There are likely others I'm forgetting right now but this is the general idea

Some of these are set in GET and some of them are set by our automation in instrumentor.

Summary

Add additional CODEOWNERS to reduce bus-factor risk.

Proposed reviewers: @reprazent @ayeung @fforster @e_forbes @pguinoiseau @amknight

Related: mstaff#529

For 4. Check this too., we don't currently have an easy answer: the changes made in runbooks to the get hybrid need to be picked up in Instrumentor through a dependency update, and then instrumentor deploys the changes in to dedicated tenants. You don't have access to a dedicated tenant unless you set one up yourself: https://gitlab-com.gitlab.io/gl-infra/gitlab-dedicated/team/engineering/sandbox-setup.html. Which I think will be out of scope for this. You might want to ask for help from groupenvironment automation for that.

Indeed. I can probably help with a hybrid setup that avoids having to set up a full sandbox but it will likely still need either an AR to Switchboard or assistance from EA to push buttons during testing but overall plan looks really good.

This runbooks update is in v21.40.0 which is rolling out this week. I also see updated and new tasks in https://gitlab.com/gitlab-com/gl-infra/capacity-planning-trackers/gitlab-dedicated/-/work_items so I'm going to mark this as done. Thank you @reprazent @ktchernov and @troblot 🦸 the lot of ya.