Add MFE config plumbing and instance kill switch
What does this MR do and why?
Configuration foundation for registry-baked micro-frontend (MFE) delivery: the optional gitlab.yml mfe section (static kill switch plus registry source url/group/token), request-log filtering for the registry token, and Gitlab::Mfe.enabled? gated by the new mfe_enabled feature flag (:instance actor).
Nothing reads this configuration yet — catalog, bake and delivery follow in the next parts of the stack.
How to verify
bin/rspec spec/lib/gitlab/mfe_spec.rbWith no mfe: section in gitlab.yml, Gitlab::Mfe.enabled? is false; with mfe: { enabled: true } and the flag enabled it is true.
Stack
This is part of a stacked series extracted from a single 1700+ line MR for reviewability. Each part is independently verifiable and leaves master inert: nothing is user-reachable until an application is pinned.
| # | Scope | MR |
|---|---|---|
| 1 | Config plumbing + instance kill switch | !245043 (closed) |
| 2 | Sidecar manifest validation + pin-file catalog | !245044 (closed) |
| 3 | Verified bake task + baked manifest | !245045 (closed) |
| 4 | Same-origin delivery endpoint | !245046 (closed) |
| 5 | Dev-mode proxy for local development | !245301 (closed) |
Merging bottom-up; when a part merges, GitLab retargets the next one to master automatically.
A separate demonstration MR shows the whole stack working end to end with the real Duo Chat MFE (production bake + dev-mode HMR): !245014.
Related
- Phase 1 issue: #605798
- Epic: gitlab-org#22777
Related to #605798.