On Thu, Aug 29, 2013 at 11:06 PM, Bocap <kakalot4_9@yahoo.com> wrote:
> 1. Instead of using pg_basebackup for standby DB, i initdb for both standby
> and primary.
> 2. Create recovery.conf in standby DB, and start both nodes.
> Now it work fine for me, but is it sure that no problem?
A standby node *needs* to be based on a base backup of its primary or
the node it connects to (it can be as well a slave in a cascading
configuration). One of the reasons being that they need to share the
same system identifier (Database system identifier output in
pg_controldata). Using initdb on multiple nodes to set up a cluster
will simply not satisfy this condition, and your cluster setup will
fail. More details here:
http://www.postgresql.org/docs/9.2/static/continuous-archiving.html
--
Michael