Postgresql 11 already changed the position of replication parameters, now all hosted in the main postgresql.conf file.
So on the standby node, instead of a recovery.conf file, just an empty STANDBY.SIGNAL file is needed in the $PGDATA directory to start Postgresql as a standby replica on that node.
However, as directed by the official documentation you've mentioned, section 26.2 https://www.postgresql.org/docs/12/warm-standby.html really advises that we set-up log-shipping to better support the stand-by recovery process, and make sure we can easily recover from any transient failure.
I'm a very happy user of wal-g for this; I'm currently running 4 different Postgresql clusters some pg10 ando some pg12, but wal-g is great for any of them; I'm running them in Kubernetes, so I'm configuring wal-g directly via the Environment of the Postgresql containers, making sure wal-g is installed in my custom Postgresql image, of course. Then it is configured just like this:
archive_mode = always
archive_command = 'wal-g wal-push %p'
archive_timeout = 60
restore_command = 'wal-g wal-fetch %f %p'
Adalberto
Il giorno mer 31 mar 2021 alle ore 17:39 Laurent FAILLIE <l_faillie@yahoo.com> ha scritto:
Replying to myself :)
It seems pg_basebackup did all the tricks, even restarting the slave. And it is in standby mode.
Do I have anything to do in addition ?
Thanks
Le mercredi 31 mars 2021 à 12:51:29 UTC+2, Laurent FAILLIE <l_faillie@yahoo.com> a écrit :
Hello,
We are running Postgresql 12 and I'm trying to put in place streaming wal replication.