I am trying to achieve maximum redundancy, and automatic failover/rejoin, with GitLab Omnibus Geo running 2 nodes on 2 hosts.
I have 2 Geo nodes sharing a floating IP (keepalived/VRRP).
When primary node goes down the secondary node automatically gets promoted to primary, by a small notify_master script.
At this point the previously-primary node is forgotten about, and once it comes up again, I need to automatically get added as a secondary node. I can then use a notify_backup script (and vice-versa) on this host to (re)set up the replication, but that does not add the node as a Geo node, which, as far as I am know, has to be done manually in the web-UI.
Is there a way to automatically have a node get added as a secondary node?
In short; I need a most redundant self-managed GitLab Omnibus running on exactly 2 hosts.