[Pgpool-general] INSERT slow on one system not on another

[Diego Gutierrez]

Hi

There's a thing you can do without disrupting your live cluster: try to make 
your development environment behave as closely as possible to the live one: 
PXE booting, 2.6.12 kernel, the likes. That would allow you to pinpoint the 
cause of the delay.

As a not-that-unrelated note: I understand that the current behaviour does not 
make much sense, so trying seemingly unrelated changes may lead you to a 
solution.

Regards,

DiegoG

On Wednesday 04 March 2009 13:53:08 Derek Jones wrote:
> Thanks again Lukasz.
>
> I could possibly investigate a FS change. I may be able to do that on a
> limited basis.
>
> Problem is the production cluster is in use. I only have a certain clump
> of grace periods to try this in - and then it has to be back the way it
> was for multiM$ run processing - which being mostly SELECTS works well.
>
> That'd be true whatever I was changing...
>
> It's my development cluster that's working properly :-/
>
> (!)
>
> I appreciate your suggestions tho' - always good to brainstorm thru'
> things like this. Thankful.
>
> Kind regards
>
> Derek.
>
> Łukasz Jagiełło wrote:
> > 2009/3/4 Derek Jones <scunacc at yahoo.com>:
> >> Dear Lukasz,
> >>
> >> Thanks for the feedback.
> >>
> >> The FS is ext3 on both - with default block sizes.
> >
> > Strongly recommend check reiserfs.
> >
> >> I can't easily change the kernel on a production system - but I quite
> >> appreciate the need to consider doing that if I had to - would be a
> >> *big* change to make and I'd have to significantly plan to do that over
> >> a collection of nodes.
> >
> > You said you got 2 clusters with 4 nodes. Change kernel at one node
> > and see stats (cacti whatever), change FS and see stats. You can
> > always sync that node with others.
>
> _______________________________________________
> Pgpool-general mailing list
> Pgpool-general at pgfoundry.org
> http://pgfoundry.org/mailman/listinfo/pgpool-general