This is a followup task of T285867.
Currently GitLab replica is only a passive instance. We should think about what is needed to promote the replica to an active instance in case of emergency (DC loss, networking issues, hardware issues in ganeti, ...).
So we would need some kind of cookbook/documentation what steps are needed. Some topics which come to my mind:
- restore backup (either from bacula or from production GitLab, if reachable)
- make sure SSH host keys match so users don't get an error
- switch DNS entries (here we might need a CNAME instead of A/AAA which we can switch easily)
- re-assign Runners(?)
- change CAS/SSO settings
Related Wikitech page: https://wikitech.wikimedia.org/wiki/GitLab/Failover