Is there any unique identifier for PostgreSQL? How can the master identify its slave unqiuely?

Question

Is there any way to have more than one PostgreSQL slave servers in the same system.

If yes, how come a master detect each server as different slaves from the below query

select * from pg_stat_replication;

will give one row for each slave which are connected to it, with PID, IP-address/client_address, client_port, etc..

If two servers are running in the same machine gives the same IP-address so it is not a unique identifier.

If slave got restarted then both client_port and PID will change so these are also cant be a unique identifier.

So, what is the unique identifier used by Postgres master to identify each slave server uniquely? Or how it is finding its slave servers differently?

jjanes · Accepted Answer · 2019-04-16 22:43:10Z

1

The master does not "find" the replicas. The replicas find the master.

The master identifies the replicas by either their slot_name, or their application_name, or not at all, depending on the context.

What are you trying to do?

answered Apr 16, 2019 at 22:43

jjanes

41.3k3 gold badges40 silver badges54 bronze badges

In that case how come master is generating pg_stat_repication view with different slaves uniquely?
– manvith
Commented Apr 23, 2019 at 8:50
Active WAL senders are identified by their process id. This isn't much of an identifier, as it is not stable across system (or just WAL sender) restarts.
– jjanes
Commented Apr 23, 2019 at 16:06

Add a comment |

1 Answer 1