[ clif-Bugs-301039 ] in CLIFv1.0a1, embedded SocketException in console at deployment time with Windows

[<noreply@xxxxxxxxxxxxxxxxxxx>]

Bugs item #301039, was opened at 25/11/2004 17:35
You can respond by visiting: 
http://forge.objectweb.org/tracker/?func=detail&atid=100057&aid=301039&group_id=57

Category: None
Group: None
Status: Open
Resolution: None
Priority: 8
Submitted By: Bruno Dillenseger (dillense)
Assigned to: Bruno Dillenseger (dillense)
Summary: in CLIFv1.0a1, embedded SocketException in console at deployment 
time with Windows

Initial Comment:
When deploying a test plan with CLIFv1.0a1, with the console running on
Windows, the console crashes with the following stack trace:

     [java] org.objectweb.fractal.rmi.RemoteException: error during
marshalling/unmarshalling by stub: java.io.IOException: Cannot export object:
ava.net.SocketException: No buffer space available (maximum connections
eached?): listen failed
     [java]     at
org.objectweb.fractal.rmi.stub.Stub.handleException(Stub.java:292)
     [java]     at
org.objectweb.clif.server.api.ClifServerControl_Stub.addBlade (Unknown Source)
     [java]     at 
org.objectweb.clif.deploy.BladeDeploy.run(BladeDeploy.java:74)



----------------------------------------------------------------------

>Comment By: Bruno Dillenseger (dillense)
Date: 25/11/2004 17:45

Message:
connecté 
user_id=291

This problem is related to a socket proliferation due to a
bad socket management by the RMIClassLoader, which is (too
much) used at deployment time (and possibly at
initialization time).
This results in hundreds (and possibly thousands) of open
sockets in the FIN_WAIT_2 state (see 'netstat -a' command).
It happens exactly the same way with Windows, Linux and
Solaris, except that the maximum number of sockets is
reached much quicker with Windows.
The transparent class and resource loading feature must be
redesigned in CLIF, without relying on the RMIClassLoader.

----------------------------------------------------------------------

You can respond by visiting: 
http://forge.objectweb.org/tracker/?func=detail&atid=100057&aid=301039&group_id=57