well, it's just my guess.
I've already seen ConcurrentLinkedQueue to leak Nodes, that's why in
Grizzly we changed the type of the working queue.
As for the LinkedBlockingQueue Nodes - I still hope it's not a leak,
but just "many Nodes created", which will be cleaned by GC.
So, if it's possible, can you investigate behavior of the patched GF a
little bit longer to make sure Nodes are really leaking and really
inside Selector class?
Thanks.
WBR,
Alexey.
On Mar 30, 2010, at 16:14 , glassfish_at_javadesktop.org wrote:
> That's what I was wondering about...
>
> I thought the type changed with the grizzly update. but you say that
> is not the case? Weird.
>
> The stacktrace from the heapdump before the update is the same, btw.
> Except one or two line numbers.
>
> We did not change anything else yet, do you have an idea why a
> different node type is used?
>
> Kim
> [Message sent by forum member 'kimschneider']
>
> http://forums.java.net/jive/thread.jspa?messageID=394486
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: users-unsubscribe_at_glassfish.dev.java.net
> For additional commands, e-mail: users-help_at_glassfish.dev.java.net
>