What I ended up doing was: 1. Stopping the master. 2. Triggering the promotion of slave A. 3. After slave A was promoted I turned it off and modified the postgresql.conf file to make it a permanent master and setup WAL shipping to slave B. 4. Rsync the pg_xlog directory from slave A to slave B and set recovery_target_timeline to latest 5. Start slave B, then start slave A now master.
So far everything seems to work. -Joseph Kregloh On Tue, Jul 21, 2015 at 2:37 PM, William Dunn <dunn...@gmail.com> wrote: > As I am aware, you would have two options depending on your configuration: > > 1. Change the primary_conninfo value on the second standby's > recovery.conf to point to the standby that has been promoted to master. > However, I think this would require that instance to be rebooted for the > configuration changes to take effect > 2. Modify the DNS entries so that the host value in primary_conninfo points > to the new master rather than the old master > > Does anyone know of another way? > > *Will J. Dunn* > *willjdunn.com <http://willjdunn.com>* > > On Fri, Jul 17, 2015 at 9:43 AM, Joseph Kregloh <jkreg...@sproutloud.com> > wrote: > >> Hello, >> >> I am getting ready to take my master server out of service today and >> promote the slave to take it's place. In my current setup I have one master >> server which ships WAL files to two separate slave servers. >> >> My plan is to promote one of those slaves. How do I get the slave that I >> promoted to start replicating to the other slave the moment I trigger it's >> promotion? >> >> Thanks, >> -Joseph Kregloh >> > >
