GitLab pages accesses API through internet instead of using internal URL(possibly resulting in timeouts)
From slack thread, @cmiskell:
Ref this web-pages thing timing out on the API calls to gitlab.com: that request is going out to the internet and back in again. Any thoughts on whether it might be better to point Pages at int.gprd.gitlab.net (the internal load balancer) for these API calls? Looks like the command-line arg -gitlab-server=https://gitlab.com is relevant (maybe -artifacts-server also, don't know what that's for though). It'd certainly remove the vagaries of the internet + cloudflare from getting between an internal system and the API.
- OAuth should be fine since we have
auth-redirect-uri
param set separately fromgitlab-server
- Let's Encrypt integration will fail since it redirects LE validation requests to
gitlab-server
Proposed solution
Add another param internal-gitlab-server
(draft naming), defaulting to gitlab-server
and used in API config source(and possibly also use it for internal auth requests/proxying artifacts)