I'm putting together a test now. Should be ready in about an hour
> On Dec 2, 2013, at 11:28 PM, Mark Proctor wrote:
>
> At this point we are probably stuck, without a unit test that reproduces the
> issue. Or better still a patch fixing it.
>
> Mark
>> On 2 Dec 2013, at 19:41, brachi wrote:
http://www.commonj.com/blogs/2010/11/13/drools-tutorial-by-examples/
> On Dec 2, 2013, at 7:41 PM, brachi wrote:
>
> thanks, I used CMSClassUnloadingEnabled, but this can't help because the GC
> can't garbage this classes because they are in use somewhere, that is
> actually the leak.
>
>
>
No, set the dialect in the drl itself.
Dialect "mvel"
I think. I'll check tomorrow.
> On Dec 2, 2013, at 7:10 PM, brachi wrote:
>
> how did you changed to mvel?
> this property?drools.dialect.mvel.strict
> false or true?
> i think i tried it with false...
>
>
>
>
>
> --
> View this mess
This is exactly what happened to me. I was getting > 500mb of perm gen that
would eventually blow the heap when refreshing the rules. I switch to Mvel and
changed that setting and now I just have a steady 50m perm gen that doesn't grow
> On Dec 2, 2013, at 6:49 PM, brachi wrote:
>
> i did, not
and set
drools.permgenThreshold=0 in drools.rulebase.conf
On 2 December 2013 16:05, brachi wrote:
> i tried it. wasn't help.
> i used MAT to analyze the leak, and it points on mvel2.optimizer
>
>
>
> --
> View this message in context:
> http://drools.46999.n3.nabble.com/permgen-leak-tp4027038p
I found that changing dialect to mvel stopped jitting classes and the
performance hit was not noticeable.
On 2 December 2013 15:51, brachi wrote:
> i'm using drools 5.5.0, after the rules deployment, (associate DRL's with
> the
> KnowladBase and compile them)
> my permgen grows to 500m, and als
