[Pgpool-general] pg_pool-I load balancing

[Mike Rylander]

On 3/11/07, Tatsuo Ishii <ishii at sraoss.co.jp> wrote:
> > So ... I thought I understood the documentation regarding the
> > weight_master and weight_secondary settings but things are acting
> > backwards.  I currently have
> >
> > weight_master = 1
> > weight_secondary = 2
> >
> > and I expected the secondary box to get twice as much traffic as the
> > master.
>
> Yes, your expectation is correct.

Right, and I've confirmed this by simplifying my tests.  There were
transaction blocks that were conflating my measurements, along with
some very unfortunate query rotation -- every third query, on average,
was more expensive than the intervening two.  By adjusting the ratio
(it's now 2:5) I am seeing the load on each that I expect, including
the in-transaction queries.

>
> > It seems, though, that the opposite is true, and the
> > secondary is getting the majority of the traffic.  Am I completely
> > misunderstanding the setting here?
>
> So the secondary gets more traffic than master. Doesn't it what you
> expect?
>

I'm sorry for the confusion.  I meant to say that the /master/ was
getting the majority of the load.

> Also please let me know the pgpool version.

I'm using version 3.1, and after looking at the release notes for the
later releases I don't see any particular reason to switch at the
moment.  Everything is in fact working as it should, I just happen to
hit a particularly bad situation with query rotation that was
compounded by normal transactional selects.

Thanks for the response ... I think everything is under control now!

-miker

> --
> Tatsuo Ishii
> SRA OSS, Inc. Japan
>


-- 
Mike Rylander
mrylander at gmail.com
GPLS -- PINES Development
Database Developer
http://open-ils.org