Does logical replication slot itself would be physically replicated to slaves?
Hi All,
My topology is as shown below:
slave1 <------physical streaming protocol-----master1 <----logical
streaming protocol-----> master2 ------ physical streaming protocol
---> slave2
The master1 and slave1 are within DC1, while the others are within DC2.
The logical streaming replication is used to replicate changes between
DC. I need to confirm that whether the logical replication slot would
be saved in wal and replicated to slave via physical replication? So
that in case of failover, the new master still holds the slot needed
by peer.
On Sat, Sep 08, 2018 at 05:37:43PM +0800, Jinhua Luo wrote:
The logical streaming replication is used to replicate changes between
DC. I need to confirm that whether the logical replication slot would
be saved in wal and replicated to slave via physical replication? So
that in case of failover, the new master still holds the slot needed
by peer.
Replication slot creation is not an operation WAL-logged. Please note
that slots are not kept in base backups when taking them using the
replication protocol (see basebackup.c for the list of things filtered),
so as the standby or the new instance deployed with this base backup
does not bloat pg_wal with data which may not be around anymore.
--
Michael