I've run quite a few tests and it seems that if I deploy vRA in distributed/redundant configuration, using a primary/replica external postgres cluster, that installation of IaaS Model Manager Data always fails at the building phase. When I deploy only 1 postgres node, or promote and then shutdown the replica to break the cluster, then IaaS model manager data installation succeeds. The load balancer is always ever referencing the primary/active postgres node, the replica is only ever a replica.
Has anyone seen this? Is this known? I've ran tests across 4 distributed deployments in numerous configurations and it always boils down to model manager data fails to install if the external postgres are clustered and the replica is online.
