Re: pgpool - Mailing list pgsql-general

From John Cunningham
Subject Re: pgpool
Date
Msg-id 6f782a240501170805c63811@mail.gmail.com
Whole thread Raw
In response to Re: pgpool  (Tatsuo Ishii <t-ishii@sra.co.jp>)
Responses Re: pgpool  (Tatsuo Ishii <t-ishii@sra.co.jp>)
List pgsql-general
So - would it then be worth doing pgpool?


On Sat, 15 Jan 2005 11:12:04 +0900 (JST), Tatsuo Ishii
<t-ishii@sra.co.jp> wrote:
> > John Cunningham wrote:
> > > concerned that if I drop the number of connections to less than the
> > > number of databases I have, that pgpool would open the limit of
> > > connections, hold them open and not allow any connections to the
> > > remaining databases.  Is this a concern?  If I set up pgpool will I
> > > have to have the same number of connections as I have databases?
> >
> > That depends on how you configure pgpool.  pgpool is not aware of the
> > connection limit count in the PostgreSQL server, so it will happily
> > open connections until there are no more slots available.
> >
> > pgpool will require max_pool * num_init_children connection slots.
> >
> > max_pool should be the number of database/user combinations you use
> > (300 in your case, assuming only one database user account),
>
> Not really. If a user connects to pgpool and all onnection slots are
> already full, then pgpool will release the oldest connection slot and
> reuse it for the new connection. So even if there are 300
> database/user combinations, it's ok to set max_pool as low as, for
> example, 4. Of course this will have unwanted side effect in that
> connection caches are not very well kept, though.
> --
> Tatsuo Ishii
>
> >  and
> > num_init_children should be on the order of how many concurrent
> > connections you expect to each combination ("several" in your case).
> > So you should have at least 300 * several PostgreSQL connection slots,
> > which is probably more than the 1000 or so that is the default.
> >
> > --
> > Peter Eisentraut
> > http://developer.postgresql.org/~petere/
> >
> > ---------------------------(end of broadcast)---------------------------
> > TIP 4: Don't 'kill -9' the postmaster
> >
>

pgsql-general by date:

Previous
From: Tom Lane
Date:
Subject: Re: PQexecParams and CURSOR
Next
From: Tom Lane
Date:
Subject: Re: Index optimization ?