Salut,
glassfish_at_javadesktop.org wrote:
> Salut,
>
> just a short reply from home: I thought I had posted it from my work, but somehow it did not here...
>
> Using the Linux /proc/<PID>/task/<THREAD-ID> interface I found that the culprit for the 100% CPU usage was actually the SelectorThread as originally posted. More details, e.g. stack trace tomorrow.
>
> We are using JDK 1.6.0_03 if I remember correctly. Is there any issue with NIO on Linux 2.6.20?
Yes, they are 2:
6403933 arises with channels that are registered with an events ops of 0
and the connection is terminated by the peer. (Guess who filled that one
:-) That one should be fixed in ur4)
6693490 has two failure modes, one of which is a spinning Selector.
For 6693490, I can get you a JDK patch to try. Would you be able to test
it?
Thanks!!
-- Jeanfrancois
>
> I downloaded the V2.1 build, but had no time to try this. Will do so tomorrow...
>
> Looking forward to your findings and cheers, Jörg
> [Message sent by forum member 'jthoennes' (jthoennes)]
>
> http://forums.java.net/jive/thread.jspa?messageID=279790
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: users-unsubscribe_at_glassfish.dev.java.net
> For additional commands, e-mail: users-help_at_glassfish.dev.java.net
>