Yes, it was not clear from the discussion that the problem is a design 
issue within the JVM and that removing PermGen really just pushes the 
problem to another generation. I'd love a deep dive into this issue by the 
The Posse and/or guests. It's a little pathetic/scary that if you redeploy 
a Java app one too many times to a container, it may blow up (typically 
stop responding at all and needing a kill -9). It makes me question the 
whole container aspect and instead deploy with 
embedded light-weight containers (Jettt/Grizzly).

/Casper

On Friday, September 7, 2012 12:01:25 PM UTC+2, kirk wrote:
>
> Hi all, 
>
> Just finished listening to 393 while on my way back home from Zurich. I've 
> been watching the removal of PermGen and I have to say that I'm not all the 
> excited about it leaving us. Dick mentioned that reloading applications 
> into TomCat could fill up perm space. Indeed it does but the old app you've 
> replaced should be GC'ed.. but it can't be because of dreaded classloader 
> leaks. And that is only one scenario where a developer, doing nothing 
> wrong, ends up leaking classes into perm space. I should point out that 
> IBM's JVMs and JRockit, neither of which have permgen also suffer from this 
> classloader leak problem. The problem is; those classes leak into general 
> heap space and this degrades GC throughput while it fills up heap until 
> eventually you degrade into Full GCs and the OOME. WIth Perm Gen you run 
> out of memory (sooner) but in a way that doesn't degrade performance while 
> it's happening. Removing Perm Gen is fixing the symptom, not the 
> problem.ac 
>
> Not exactly what you're looking for but there is a Thread constructor that 
> allows you to set Java stack size. The problem with native stack is that 
> it's sandwiched between text and C heap. You'd have to move C heap to 
> dynamically change the size of a stack frame and although we can do that 
> for stuff in Java heap... I don't know how you'd do it generically in C 
> heap. But again, this problem would be mute if proper support for recursion 
> was put into the JVM. 
>
> -- Kirk 
>
>

-- 
You received this message because you are subscribed to the Google Groups "Java 
Posse" group.
To view this discussion on the web visit 
https://groups.google.com/d/msg/javaposse/-/DOHi1gvq-pMJ.
To post to this group, send email to [email protected].
To unsubscribe from this group, send email to 
[email protected].
For more options, visit this group at 
http://groups.google.com/group/javaposse?hl=en.

Reply via email to