I'm now testing with pg_pool installed on each apache frontend with 260
pg_pool preforked clients in each machine.
The database seems to work better. At least when it goes to swap it
doesn't stop working...
I also reduced the shared buffers and moved the pg_xlog folder to
another disk on another raid container.
I'll let go the test for all the night and tomorrow I'll let you know
the results...
Thank you for your help,
Denis
Jeff Trout wrote:
>
> On Jul 27, 2005, at 4:22 AM, denis@edistar.com wrote:
>
>> Already checked.
>> We set a machine with only pg_pool installed as a fronted to the
>> real db.
>> If I disable persistent connections and I use pg_pool with 4096
>> preforked clients,
>
>
> no no no.
> you don't want 4096 preforked clients.
>
> What you want to do is leave it at the default 32 preforks and run a
> copy of pgpool on each frontend. Then point your app to connect ot
> localhost (or the unix socket). Be sure you change things to use a
> "normal" connect not a persistant connect. This should be fine for
> you. I used to run a site doing hundreds of connects & queries /
> second across 4 frontends each of which ran a pg_pool connected to a
> db box.
>
> If you fine 32 clients isn't enough, then bump it up. The idea is to
> keep as few connections as possible.
>
> --
> Jeff Trout <jeff@jefftrout.com>
> http://www.jefftrout.com/
> http://www.stuarthamm.net/
>
>
>
> ---------------------------(end of broadcast)---------------------------
> TIP 6: explain analyze is your friend
>