An attempt to create more than a single registry in the
same VM will fail. Specifically, if 

    createRegistry(++portNumber) ;

if called more than once, it will throw:

  java.rmi.server.ExportException: internal error:
      ObjID already in use

I discovered this while trying to implement 2 flavors of
the same rmi service on two different ports: one using
the default socket implementation, one using custom
sockets (ssl, in my case).  The  customer  was to service
both ssl-capable and non-ssl-capable clients from the
same vm, all without running an external rmiregistry.

The text of Bug Id 4188333, (a similar, but not identical,
problem) suggests whence the problem arises:
presumably, a single, well-known ObjId is used
for any and all registry instances, hence calling
createRegistry a second time uses this same ObjId for
the new registry instance.  The server's export mechanism
doesn't allow this,  and throws. 

The problem is easily reproducable, as the 
following code demonstrates:

---------------------8<-------------------------
import java.rmi.registry.* ;

public class Reg {

public static void main(String [] args)
{ // second createRegistry call fails
  try
  { registry1 = LocateRegistry.createRegistry(5555) ;
    registry2 = LocateRegistry.createRegistry(6666) ;
  }
  catch(Throwable t)
  { System.out.println(t) ;
  }
}
}
---------------------8<-------------------------


bash-2. customer $ java -version
java version "1.2.2"
HotSpot VM (1.0fcs, mixed mode, build E)
bash-2. customer $ java -fullversion
java full version "JDK-1.2.2-W"
bash-2. customer $
(Review ID: 93916) 
======================================================================




java version "1.3.0"
Java(TM) 2 Runtime Environment, Standard Edition (build 1.3.0-C)
Java HotSpot(TM) Client VM (build 1.3.0-C, mixed mode)

Creating multiple registries within the same VM causes a
java.rmi.server.ExportException.  For example, the following code:

import java.rmi.*;
import java.rmi.registry.*;

public class Foo {

    public static void main(String args[]) {
	try {
	    LocateRegistry.createRegistry(12345);
	} catch (Exception e) {
	    e.printStackTrace();
	    System.exit(1);
	}
	System.err.println("registry created on 12345");
	try {
	    LocateRegistry.createRegistry(23456);
	} catch (Exception e) {
	    e.printStackTrace();
	    System.exit(1);
	}
	System.err.println("registry created on 23456");
    }
}

when run produces the following output:

registry created on 12345
java.rmi.server.ExportException: internal error: ObjID already in use
        at sun.rmi.transport.ObjectTable.putTarget(ObjectTable.java:164)
        at sun.rmi.transport.Transport.exportObject(Transport.java:71)
        at sun.rmi.transport. customer .TCPTransport.exportObject(TCPTransport.java:184
)
        at sun.rmi.transport. customer .TCPEndpoint.exportObject(TCPEndpoint.java:319)
        at sun.rmi.transport.LiveRef.exportObject(LiveRef.java:119)
        at sun.rmi.server.UnicastServerRef.exportObject(UnicastServerRef.java:12
5)
        at sun.rmi.registry.RegistryImpl.setup(RegistryImpl.java:95)
        at sun.rmi.registry.RegistryImpl.<init>(RegistryImpl.java:81)
        at java.rmi.registry.LocateRegistry.createRegistry(LocateRegistry.java:1
67)
        at lib.Foo.main(Foo.java:17)

This bug was reported as bug id 4120329 previously, and labeled as CLOSED, when
it is clearly not.
(Review ID: 109097)
======================================================================

None, other than running registries in seperate
VMs.  This works "around" the bug, but incurs 
the extra cycles/management overhead of extra
processe(s).
======================================================================

This will be fixed when we have an object table per exported endpoint (probably
in the Merlin release).
  xxxxx@xxxxx   1999-09-02

It is possible with the new APIs in merlin to create multiple registries in
the same VM, so I am marking this RFE as fixed/integrated in merlin.
  xxxxx@xxxxx   2001-01-10

Because JSR-78 has not been approved:

	http://java.sun.com/aboutJava/communityprocess/jsr/jsr_078_rmicrr.html

this RFE will not be satisfied as described above for the Merlin release.
  xxxxx@xxxxx   2001-02-22

This bug is under consideration for Tiger.
  xxxxx@xxxxx   2003-07-10

The internal object table is now keyed on a combination of ObjID and Transport
so that multiple registries (each on a different port) can be exported.
  xxxxx@xxxxx   2003-09-17

Note: this problem appears to make it impossible to create an RMI registry on port 1099 from within a 
servlet running under Tomcat.

Still broken in jdk113!

the same in jdk1.3.1

The same behavior in jdk1.4.0.b92

Same in 1.4 jdk using servlet version 1.2.
Quite frustrating :)

the same in jdk1.4.1

Same in jdk 1.4.1_02-b06

I am still seeing this issue in 1.4.2_01!  Does anyone 
know when this might be fixed????!!!!!

I am still suffering this bug at 1.4.2

also having same problem with 1.4.2

This is fixed in JDK1.5. The real question is whether it wil be backported to the 1.4 series?

This certainly isn't fixed in 1.5! At least not for Mac OSX.

Works in 1.5 for me on Windows ...