On Thu, 2020-03-05 at 17:06 +0000, Virendra Kumar wrote: > Failover is easy but failback is little bit tricky. > I have implemented failback by doing following steps: > > 1. Start original primary which will be doing crash recovery. It should be designed in such a way that once it is up application should not start connecting to it otherwise there will be split brain > and data-mistach between two instances. I implemented it by using a virtual IP mounting on server which is actual primary using keepalived. > 2. Shutdown original primary and do a pg_rewind to make that as slave for new primary. > 3. Once slave (original primary) is caught up with primary do failback > 4. Repeat steps #1-#3 to make failed over instance slave again. Or you use Patroni and simply say patronictl.py switchover --candidate db2 --scheduled '2020-03-06 12:00:00' mycluster Yours, Laurenz Albe -- Cybertec | https://www.cybertec-postgresql.com
