I think that the words used in the bug description can cause confusion. The leader/follower concept is tied to RAFT (active/active) as per my understanding.
However, the goal of removing the writes for metadata agent I think it's more aligned with the active/backup configuration where clients can't perform writes on the active instance. This way, metadata agents can connect to active and backups indistinctly and put some stress off the active OVSDB server instance where ovn-controller and neutron-server will have to connect to.
I think that the words used in the bug description can cause confusion. The leader/follower concept is tied to RAFT (active/active) as per my understanding.
However, the goal of removing the writes for metadata agent I think it's more aligned with the active/backup configuration where clients can't perform writes on the active instance. This way, metadata agents can connect to active and backups indistinctly and put some stress off the active OVSDB server instance where ovn-controller and neutron-server will have to connect to.