Document manual disaster recovery process for systems with multiple secondaries
Related to https://gitlab.com/gitlab-org/gitlab-ee/issues/4284
Although a single secondary geo replica is sufficient for disaster recovery, geo replication also provides performance improvements for distributed teams by running geo replicas close to each team.
The current documentation is insufficient to perform a failover when using multiple secondaries. We should update the documentation to cover this situation.
Keep in mind that the new command sudo gitlab-ctl promote-to-primary-node
will not work in this case, so we probably need to use an old one sudo gitlab-rake geo:set_secondary_as_primary
which only fixes the node status in a database.
Proposal
- Update documentation
- Test the documentation in an environment with multiple secondaries
Links
Edited by Valery Sizov