[pgpool-general: 5291] Re: Connections to backends

Lazaro Garcia lazaro3487 at gmail.com
Fri Feb 3 17:35:08 JST 2017

The weights property is used for load balancer functionality, with it you
can distribute your read queries over your servers for instance 30% to
master and 70% to slaves.  

Pgpool always tries to connect to servers because failover and swithover
functionalities. If the primary server fails, it promotes a new master.
Because of these it needs to know the current state of the servers
configured in the cluster.    

As a connection pooler, it maintains open the connections to backends for
future use, suppose an app connects to pgpool, send some queries and then
disconnect, but this connection is saved in the pool for reutilization.     


-----Mensaje original-----
De: pgpool-general-bounces at pgpool.net
[mailto:pgpool-general-bounces at pgpool.net] En nombre de Tema Zelikin
Enviado el: viernes, 3 de febrero de 2017 09:50 a. m.
Para: pgpool-general at pgpool.net
Asunto: [pgpool-general: 5290] Connections to backends


I have a following configuration:

3 backends with replication in master-slave mode.
2 pgpools with delegate_IP and watchdog, with weights of backends as
master = 1
both slaves = 0

My understanding is that pgpool shouldnt make requests to backend nodes with
weight of 0.

The question is why does pgpool establishes a connections to those nodes?

Second question, something i couldn't understand from reading a
documentation: why does number of connections to backends is much more than
number of connections to frontend?

pgpool-general mailing list
pgpool-general at pgpool.net

More information about the pgpool-general mailing list