Extensibility and flexible deployment with PostgreSQL
Various ways of extending the documented functionality of the warm
standby system is the essence of our discussion. Existing documentation
is focused on migrating to a standby server during failure events only.
Not much is known or at least documented on ways of "reshaping" such
systems during planned maintenance; site migrations etc.
Recently we speculated that it should be possible to re-point a standby
server at a peer standby that was just promoted to master. This is
convenient to do and is not documented as possible. We have been using
this technique here ever since.
We are engaged in some R&D here to see if EDB/Postgres master/standby
systems can experience role reversals without need to re-initialize the
master; as seems might be required based on existing documentation.
From what we've seen so far in testing, this is possible and it's also
been given the nod by some of your engineers.
Also related to warm standby operations; we have developed log shipping
logic to allow the standby to hold back some configurable amount of time
behind the master. This is perhaps the least interesting item of these
but may be desired by some users of EDB/Postgres and is yet another
Oracle feature that we can put a"check mark" next to.