2008/1/16, Joshua D. Drake <jd@commandprompt.com>:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> On Wed, 16 Jan 2008 00:29:16 -0500
> Tom Lane <tgl@sss.pgh.pa.us> wrote:
>
> > "Joshua D. Drake" <jd@commandprompt.com> writes:
> > > If we are sure that this issue is apparent actual row insertion it
> > > should be easy to duplicate.
> >
> > I think you missed my point entirely: I believe it's specific to
>
> Wouldn't be the first time.
>
> > Clodoaldo's installation. Certainly I didn't have any luck
> > reproducing a huge 8.2-to-8.3 slowdown here.
>
> Well I would agree that all the testing that I have done doesn't show
> 8.3 being slower. I do see this though in the very first post:
>
> http://archives.postgresql.org/pgsql-general/2008-01/msg00143.php
>
> Production: Fedora Core 6, postgresql 8.2.5, AMD XP2600, 2 GB mem, two
> 7200 ide disks with pg_xlog alone in the second disk.
> New: Fedora 8, postgresql 8.3-beta4, Core Duo 2.33 MHz, 4 GB mem, two
> 7200 sata disks in Raid 1.
>
> On the old machine pg_xlog is on its own disk but on the new machine he
> is getting random churn on the I/O because he only has the RAID 1 for
> DATA and pg_xlog.
>
> He tested 8.3 on the old box and had similar results.
No, I tested 8.3 only on the new box where I also tested 8.2. If it is
a hardware problem, what I don't know how to investigate, it is only
triggered by 8.3. When I tested 8.3 and 8.2 in the new box the raid
problem was apparently fixed as you noted.
> Although, I wonder
> if he tested 8.3b4 on the old box in the exact disk layout (with xlog
> on a separate disk).
I didn't test 8.3 on the old box. Now the new box is already the
production box running 8.2.6 and I can't test on it anymore, but I can
test everything again in the old box. That would rule out the raid
factor. I will have to wait for the weekend.
Regards, Clodoaldo Pinto Neto