As for the non-Weld aspects of this, there are some small changes with
what we hope will be major impact we're testing. They are not yet in
the builds but should be soon if the testing goes well.
- Tim
On Jun 3, 2011, at 7:50 AM, Richard Kolb wrote:
> Hi Harald
>
> Great to see you still working with the WELD issue.
>
> I have another issue with weld in Glassfish 3.1 , maybe related ?
> Perhaps you can put in some of your input.
> http://java.net/jira/browse/GLASSFISH-16655
>
> thanks
> Richard.
>
> On 3 June 2011 13:37, Harald Wellmann <harald.wellmann_at_gmx.de> wrote:
> Glassfish pre-3.1 used to have severe memory issues, mainly due to
> Weld which I'm glad to say are gone since integrating Weld
> 1.1.1.Final.
>
> On the other hand, Glassfish still uses excessive amounts of memory,
> now due to its own internal structures.
>
> Deploying the same WAR to two different servers, triggering GC and
> taking a heap dump with Eclipse Memory Analyzer, I get the following
> measurements:
>
> Glassfish 3.1.1-b06 ~250M
> Resin 4.0.pre19 ~ 60M
>
> A slightly different version of the same app running on Tomcat 6 and
> Spring 3 has a footprint of ~ 50M.
>
> As long as these figures don't change, it will be next to impossible
> to convince people to move from custom Tomcat-based solutions to
> Glassfish.
>
> The Memory Analyzer clearly identifies the culprit, please see http://java.net/jira/browse/GLASSFISH-16747
> for details.
>
> It seems the root cause is still the same as reported in
> GLASSFISH-15266 during the 3.1 stabilization phase.
>
> I do hope someone can look into this issue and fix it before the
> 3.1.1 release.
>
> Best regards,
> Harald
>