Jim Driscoll wrote:
> Jan Luehe wrote:
>>
>> This seems to be a known NIO bug. The J2SE folks have pointed me to:
>>
>> http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=6215050
>> ("SocketChannel created in CLOSE_WAIT and never cleaned up.. File
>> Descriptor leak")
>>
>> which was fixed in JDK 6 and JDK 5.0u7.
>
>
> Which means that if you run our app server on a version of JDK before
> jdk 5u7, it'll leak, right? This seems serious enough to document
> somewhere - let's wait for Jean-Francois to get back, and ask him.
I don't think GlassFish is suffering the problem. A socket is closed
after the keep-alive value is expired if HTTP 1.1 is used. So depending
on what the protocol your browser use, you might need to wait 30 seconds
before the socket is closed.
As for the leak, it happens on some linux kernel version, but not the
most up to date one (2.14.X and up). It only happens on Linux (sun.com
is running AS 8.2 without any problems) and as everyone knows, it is not
easy to submit a bug against the linux kernel and get a fix.
If I can get the exact linux kernel version we can document this
problem, but unfortunately with my current version I can't reproduce
such problem. The SQE team has started also writing tests about this issue.
-- Jeanfrancois
>
> Jim
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe_at_glassfish.dev.java.net
> For additional commands, e-mail: dev-help_at_glassfish.dev.java.net
>
>