help with magma production application

[Chris Muller]

Hi Esteban, please excuse me but... I cannot possibly determine from
this description what the problem could be.  Could you post at least
the stack trace?

Also, I am curious, why did you change the ReadStrategy depth to 0?
Someone else did that and then complained of performance problems, and
I explained that doing that causes only one object at a time to be
retrieved, which I hope is obvious.  I guess I am just confused why
people are changing the default ReadStrategy for worse?

 - Chris

On Tue, Nov 30, 2010 at 3:40 PM, Esteban Lorenzano <estebanlm at gmail.com> wrote:
> Hi,
> I have a production application, which needs to have a very few amount of objects (less than 2k, in a very short graph of two or tree levels, never more than four, and that is very rare). For that, I configured a Pharo 1.1.1 (dev), with latest magma, and it is running in a very strange linux (I think is suse, but I can't know), with latest unix cog vm.
> I have two problems, but while I can live with one of them, the other one is killing me :)
>
> First, the most important one: time to time, magma use is crashing the application. I think this is because the memory consumption scales up to 300M very fast, and it shouldn't be like that (at start, the image is around 60M)... so, I try to commit a graph of no more than 10 objects (and not really big), and the image crashes. This behavior was not present before, when I was using a "image persistence" strategy.
>
> Second: performance. Yes, everybody complains about this... and it seems to be some guidelines, that I shouldn't be following... but I'll explain a little my structure, maybe you know what I'm doing wrong.
> This is what I have:
>
> Root is a MaDictionary, which have a serie of keypairs String, MagmaCollection
> I have no more than 5 keypairs, and just one of the magma collections have more than 500 objects.
> Some of the internal objects have a tree form, always having magmacollections for the children, never more than 4 levels.
>
> The read strategy is 0
>
> Any ideas?
>
> Cheers,
> Esteban_______________________________________________
> Magma mailing list
> Magma at lists.squeakfoundation.org
> http://lists.squeakfoundation.org/mailman/listinfo/magma
>