On Tue, Nov 15, 2016 at 3:42 PM, Alvaro Herrera
<alvherre@2ndquadrant.com> wrote:
> FWIW I'm not sure "primary" is the right term here either. I think what
> we really want to know is whether the node can accept writes; maybe
> "pg_is_writable_node".
This made me think of another complication: what about cascading
replication where the middle node is both a master and a slave? You'd
probably not want to fallback to the middle node even though it is a
master for its downstream.