Geo HA + Geo documentation lacks clarity on machine vs geo node vs external_url / load balancer

Talking about this document: https://docs.gitlab.com/ee/administration/geo/replication/high_availability.html

In discussion with @abuerer we encountered confusion over the following points:

  • This document doesn't talk about adding the secondaries to /admin/geo_nodes at all
  • It's unclear that every machine in a secondary node should share the same external_url (that of the load-balancer), and that that exact external_url is the value that needs to be used for the secondary when adding it to the admin panel
  • It's unclear that a "a secondary node" refers to N machines forming one Geo node by dint of sharing database, redis and file configuration. It can be taken to mean "one server" instead, meaning that people talk about a load balancer as if it points to multiple secondary nodes.

We can clarify and fix all these issues through documentation. It would be nice to rename a geo node to something else, but I suspect that's too much effort.

Assignee Loading
Time tracking Loading