Re: [Fwd: postgres 8.4.1 number of connections] - Mailing list pgsql-performance

From Magnus Hagander
Subject Re: [Fwd: postgres 8.4.1 number of connections]
Date
Msg-id AANLkTimpzjUFx24ecMs_exu8EhMUs_N4VoTagWwS2QBR@mail.gmail.com
Whole thread Raw
In response to Re: [Fwd: postgres 8.4.1 number of connections]  (Scott Marlowe <scott.marlowe@gmail.com>)
List pgsql-performance
On Thu, Aug 26, 2010 at 23:13, Scott Marlowe <scott.marlowe@gmail.com> wrote:
> On Thu, Aug 26, 2010 at 2:29 PM, Maria L. Wilson
> <Maria.L.Wilson-1@nasa.gov> wrote:
>>
>>
>> we have this application (using jboss/java/hibernate) on linux accessing
>> data on 3 postgres database servers using 8.4.1.
>>
>> One of our many concerns has been the way we handle connections to the
>> database.  java/hibernate handle their own pooling so I understand that
>> using anything else is out of the question.
>
> It's not impossible to put another connection pooler between hibernate
> and your database server.  May not be useful or better, but it's
> possible.

I've been using pgbouncer in combination with the jboss connection
pooler several times, mainly for the reason that reconfiguring the
jboss connection poolers (particularly if you have lots of them) can
cause quite a bit of downtime. Just sticking a 1-1 mapping pgbouncer
in between with support for SUSPEND makes a lot of difference if you
switch master/slave on your replication /ha. It'll still break the
connections for jboss, but it'll recover from that a *lot* faster than
a reconfig.

--
 Magnus Hagander
 Me: http://www.hagander.net/
 Work: http://www.redpill-linpro.com/

pgsql-performance by date:

Previous
From: Scott Carey
Date:
Subject: Re: Performance on new 64bit server compared to my 32bit desktop
Next
From: Willy-Bas Loos
Date:
Subject: Re: turn off caching for performance test