[pgpool-general: 667] pgpool 3.1.2 - reusing named prepared statements

[Dan Biagini]

I am evaluating postgres connection poolers (pgbouncer and pgpool),
specifically I’m looking for a way to allow reusing prepared
statements on the backend connections across all frontend client
connections.

The use-case is for a handful of complex but performance critical
queries which get executed by short lived client processes.  So
ideally client app code would connect to pooler, and check if the
prepared statement already exists for this connection (via
pg_prepared_statement table).  If it doesn’t the client will create
the statement(s), and execute them, then disconnect, but leaving the
prepared statement allocated on the backend for use by a subsequent
client connection.

It doesn’t seem to work with pgpool (I found pgpool was calling
DEALLOCATE in the connection cleanup code, regardless of whether it
was configured in the reset_query_list configuration).

Is this a known behavior?

Thanks!