Now, you can change the primary_conninfo on C so that it connects to
B. I guess that means a restart. Note that you only want to do that once
server B is the new master, so has been triggered as such. That means a
change of TimeLine, though, so it'll break streaming : that's when you
need to read the WALs from the archive, with the restore_command.
So does that mean we don't have to build C from scratch like we will be doing for A? Is that because B and C were both slaves and reading from the same archive directory and at same point in time when we did the switchover.
Then, to have the old master A a slave of the new master B, you need to
do a full slave setup here, with base backup and all.
This is what I expected but just wanted to confirm. Thanks a lot for the detailed explanation on all steps.
I hope we will be able to get cascading support in 9.1 or before so that
it's possible to simplify all this by having a "relay" server in the
mix, so that you only talk to this one when changing the master.
It will be just awesome to have such support as cascading at times is really helpful. But again comparing to other third party solutions it seems just very simple to setup for replication and sure does work nice. I cant wait for 9.1 already :)
Regards,
Shoaib Mir
http://shoaibmir.wordpress.com/
