Cache remote interface stub in clustered env

Hi,
Can we put in cache the interface remote stub of a ejb session stateless? Which can be the impact when a serveur crashes ? Can we use the reference in cache in an multithreaded environment (a reference used simultaneously by several threads)?
Thanks.

Similar Messages

When do remote interface get invaliated? Can I cache them?

Hello,
I am designing an application, and considering using the "Service Locator" pattern, acting as a singleton that will cache jndi context, and bean home and remote interfaces. However I am unsure whether caching of remote interface is a problem or not:
1) Under what circumstances will the cached remote interface for a stateless session bean be invalidated?
2) Under what circumstances will the cached remote interface for an Entity Bean be invalidated?
3) Assuming that I cache Stateful session bean remote interfaces in a Servlet HTTPSESSION, under what circumstances will the cached remote interface for Stateful session bean be invalidated?
4) Finally, do I have to utilise home handles instead of remote interfaces for any of these situations? If so, why?
Thanks

I understand that caching the Home Interface or HomeHandle of this, is the solution for
reconnection to EJB's problems.
But I do not succeed in getting a working solution for handling the reconnection
to a restarted EJB server after some downtime.
First the HomeHandle of the Home Interface is successfully saved with:
"myHomeHandle = myHome.getHomeHandle();"
Later the HomeHandle is used to restore the Home Interface with:
"MyHome myHome = (MyHome) myHomeHandle.getEJBHome();"
But this gives an Exception like
Caused by: javax.naming.NoInitialContextException: Need to specify class name in environment or system property,
or as an applet parameter, or in an application resource file: java.naming.factory.initial
I have followed the recommandations in this postings, but in some way I lack a bit information.
Has somebody a code example in Java of how to do this?
Here is my code which throws the exception
InitialContext context = new InitialContext( properties );
Object obj = context.Lookup( "My" );
MyHome myHome = (MyHome) PortableRemoteObject.narrow( obj, MyHome.class );
myBean = myHome.create();
myHomeHandle = myHome.getHomeHandle();
// At this point everything fine, and a myBean is created.
// Now I try to recreate the myHome Interface
MyHome myHome = (MyHome) myHomeHandle.getEJBHome();
// In order to (re)create the myBean EJB.
myBean = myHome.create();
// But the second create is never executed, as the method call "...myHomeHandle.getEJBHome();"
throws the above Exception.
So, any help I will Appreciate.

Location of stubs, skeletons, remote interface ,client and server files

I have a question to the subject mentioned above :
In a training example for RMI I red the following :
.... stubs, remote interface and client files into
client location
... skeletons, remote interfaces, server files and stubs
into server location ....
I understand that with the exception that the stubs must also be located at server side. My understanding is
that it is enough if those reside only at client side. Is that
right or wrong ?
Thanks for your comment.
Regards,
Oskar

The stubs must be available (in the classpath) on the server side. This is because when a remote server object is exported, the RMI runtime creates an instance of the stub (instead of the server) and sends it whenever a reference to the server is encountered. In order to create (and then serialize) the stub, its class file must be available on the server.

Multiple inheritance in remote interfaces for EJB 3.0 session beans on Webl

Hi All,
We started migration from EJB 2.1(WLS 8.1) to EJB 3.0(WLS 10.3.2) and identified few serious problems. One of them is related with multiple business interfaces inheritance. I wrote simple example that presents point of the problem.
we have session bean AImpl:
+@Stateless(name="A")+
+@Remote({A.class})+
+@TransactionAttribute(TransactionAttributeType.REQUIRED)+
+public class AImpl implements A {+
+@Override+
+public void writeA() {+
System.out.println("A");
+}+
+@Override+
+public void writeB() {+
System.out.println("B");
+}+
+@Override+
+public void writeC() {+
System.out.println("C");
+}+
+}+
with remote interface A:
+@Remote+
+@JNDIName(A.JNDI_NAME)+
+public interface A extends B, C {+
public static String JNDI_NAME = "A_JNDI_NAME";
void writeA();
+}+
As you can see A extends B, and C. Definition of both interfaces is very simple:
+public interface B {+
void writeB();
+}+
+public interface C {+
void writeC();
+}+
Everything looks nice until we want to invoke some method on AImpl bean. For above implementation code:
A a = ctx.lookup(A. JNDI_NAME);
a.writeA();
a.writeB();
a.writeC();
writes down ”A \n B” and throws exception:
caused by: java.lang.NoSuchMethodException: pl.gov.arimr.zszik.bazowe.slowniki.ejb.A_vt0zts_AImpl_1032_WLStub.*writeC()*
at java.lang.Class.getMethod(Class.java:1605)
at weblogic.ejb.container.internal.RemoteBusinessIntfProxy.getTargetMethod(RemoteBusinessIntfProxy.java:165)
at weblogic.ejb.container.internal.RemoteBusinessIntfProxy.invoke(RemoteBusinessIntfProxy.java:57)
so.. in stub generated by WLS there is no method from interface C ! What more interesting after small change in interface A rely on change in interface implementation order from B, C to C, B (+public interface A extends C, B {+) server writes down only A and I have stack like below:
Caused by: java.lang.NoSuchMethodException: pl.gov.arimr.zszik.bazowe.slowniki.ejb.A_vt0zts_AImpl_1032_WLStub.*writeB()*
at java.lang.Class.getMethod(Class.java:1605)
at weblogic.ejb.container.internal.RemoteBusinessIntfProxy.getTargetMethod(RemoteBusinessIntfProxy.java:165)
at weblogic.ejb.container.internal.RemoteBusinessIntfProxy.invoke(RemoteBusinessIntfProxy.java:57)
After this experience I came up with suspicion that Weblogic 10.3 does not support inheritance from multiple interfaces in one “generation”. Instead of that it takes only the first interface from the list.
Does anybody have some experience with such a situation? maybe someone have an idea how to work around this problem?

This is Not Supported in WebLogic that the Remote Interface extends other Interfaces. Because Annotation Processor just looks up inside the implemented interface methods. The actual interface which is Implemented by the Bean Class. So the Methods declared inside the Interface B and Interface C will be ignored and will not be available as part of the generated Stubs. Thats why u are getting NoSuchMethodError.
You can even contact Oracle Support on this...there are 3-4 Cases on it. And the Solution is Work As Designed.
Workaround is : edit your interface A as following
Declare all the Business Methods only in the Remote Interface and not inside it's Super Interfaces.
Example:
@Stateless(name="A")
@Remote({A.class})
@TransactionAttribute(TransactionAttributeType.REQUIRED)
public class AImpl implements A {
@Override
public void writeA() {
System.out.println("A");
@Override
public void writeB() {
System.out.println("B");
@Override
public void writeC() {
System.out.println("C");
@Remote
@JNDIName(A.JNDI_NAME)
public interface A extends B, C {
public static String JNDI_NAME = "A_JNDI_NAME";
void writeA();
void writeB();
void writeC();
Thanks
Jay SenSharma
http://jaysensharma.wordpress.com (WebLogic Wonders Are Here)

A problem while getting a EJB remote interface from SJSAS 9.0

I hava deployed a session bean in SJSAS 9.0
I wrote some codes to get the remote interface as follow:
Context ctx = null;
Hashtable env = new Hashtable();
env.put ("java.naming.factory.initial","com.sun.jndi.cosnaming.CNCtxFactory");
env.put("java.naming.provider.url","iiop://127.0.0.1:3700");
try {
ctx = new InitialContext(env);
} catch (NamingException ex) {
ex.printStackTrace();
try {
Object cs =ctx.lookup(ejb.MySessionBean);
} catch (NamingException ex) {
ex.printStackTrace();
A exception occured during the lookup operation.
javax.naming.NameNotFoundException [Root exception is org.omg.CosNaming.NamingContextPackage.NotFound: IDL:omg.org/CosNaming/NamingContext/NotFound:1.0]
at com.sun.jndi.cosnaming.ExceptionMapper.mapException(ExceptionMapper.java:44)
at com.sun.jndi.cosnaming.CNCtx.callResolve(CNCtx.java:453)
at com.sun.jndi.cosnaming.CNCtx.lookup(CNCtx.java:492)
at com.sun.jndi.cosnaming.CNCtx.lookup(CNCtx.java:470)
at javax.naming.InitialContext.lookup(InitialContext.java:351)
at demo.Main.run(Main.java:46)
at demo.Main.main(Main.java:62)
Caused by: org.omg.CosNaming.NamingContextPackage.NotFound: IDL:omg.org/CosNaming/NamingContext/NotFound:1.0
at org.omg.CosNaming.NamingContextPackage.NotFoundHelper.read(NotFoundHelper.java:72)
at org.omg.CosNaming._NamingContextExtStub.resolve(_NamingContextExtStub.java:406)
at com.sun.jndi.cosnaming.CNCtx.callResolve(CNCtx.java:440)
... 5 more
Anyone can solve this problem for me???
Thanks a lot

We don't recommend explicitly instantiating the CosNaming provider within a stand-alone java client when accessing beans within the Java EE SDK. We have a simpler approach that involves just instantiating the no-arg InitialContext. Details are in our EJB FAQ :
https://glassfish.dev.java.net/javaee5/ejb/EJB_FAQ.html
--ken

Remote interface distribution

Hi,
I am just learning to use RMI, have a typical setup, Server and Client and everything works fine, BUT.......
I have compiled my remote classes, ie RmtServer, RmtServerImpl, then run the rmiregistry
and then started the service, so far so good.
I then compile etc the Client class and run
My question is this, I am currently working from the same directory for all classes, I would like to truly
distribute the client and server. Do I need to put a copy of the RmtServer interface on the client?
Currently in my client I have the line:
RmtServer server = (RmtServer)Naming.lookup("rmi://" + serverHost + "/ProjectServer");
I have been reading about dynamic class loading but I am slightly confused!!!
It says that the stubs created by running "rmic" on the server classes can be loaded dynamically, thats great, but the stubs created and the RmtServer interface are not the same thing. If I have to have a "local" copy of every interface I want to use then it seems a bit limited.
My impression was that I could specify a location for a class repositry on a remote machine and then
instantiate classes from the server on the local machine! So then I would only need to have one interface on the client to enable me to connect to the server for the first time!!!
Im really confused as people can probably tell. Any help would be really appreciated

The purpose of dynamic loading is really so that you don't have to distribute the stub, or the implementation classes of any interfaces.
You can certainly load the remote interface dynamically from the codebase in the client, e.g. with URLClassLoader, but you will also have to have already loaded all the classes that use the remote interface the same way, rather than via the system class loader. Otherwise they won't load (NoClassDefFoundError). You can download the entire client actually, and this is not a bad way to go: see the the RMI 1.1 Specification, 'Bootstrapping the client', for details (if you can find it: try http://java.sun.com/products/archive/jdk/1.1/index.html), and the RMI-USERS archives of about five-six years ago for discussions (see http://archives.java.sun.com/archives/rmi-users.html). The description of this was removed from the 1.2 specification for some reason, and I did encounter problems migrating a 1.1 bootstrap client to 1.2, so beware.

TraderService Remote Interface

Why is the TraderService back end stateless session EJB implemented with remote
interfaces instead of local interfaces. If I'm not mistaken, upon reaching this
service, all processing will be co-located on one server, therefore, wouldn't
it be more performant to implement this EJB with local interfaces?

Hello,
The examples are simply starting points to assist your own development
work.
You might want to take a look at the web service design guide [1] in the
docs. If you are deploying to a cluster of WebLogic Server instances,
by default the deployment targets all server instances in the cluster.
This corresponds to homogenous module deployment, which is recommended
in most clusters. See [2].
Regards,
Bruce
[1]
http://edocs.bea.com/wls/docs81/webserv/design.html
[2]
http://edocs.bea.com/wls/docs81/deployment/overview.html
Sean Cohan wrote:
>
Why is the TraderService back end stateless session EJB implemented with remote
interfaces instead of local interfaces. If I'm not mistaken, upon reaching this
service, all processing will be co-located on one server, therefore, wouldn't
it be more performant to implement this EJB with local interfaces?

One remote interface vs. lots of interfaces

Hi,
I am now designing an API using RMI.
I can do it in 2 ways: One remote interface which has lots of functions,
or seperate it to lots of logical interfaces, which the client needs to get a remote interface of all of them.
Does anyone face this problem before? what do you think?
Thanks.

If those are your only options, I would recommend using several smaller logically grouped interfaces, rather than one large one, because any changes to an interface require the generation and distribution of a new stub. In some cases, your client may not even need all the interfaces.
However, there is another way to do it.
You could use only one interface, containing only one method, which can support unlimited client interfaces, with no stubs. That is my favourite way to do it! Want to see more? Please come and visit, it is very easy to do, and it is completely free:
https://cajo.dev.java.net/overview.html

Remote Interface

Hey. I was wondering if there was a way to enumerate the remote interface used by an RMI server if you do not have the RMI Client. Ive been looking at reflection but from what ive seen it assumes you know what type of object you will be dealing with.

You can use reflection on the server or the stub to enumerate the interfaces it implements. You have to filter them so you only look at interfaces which extend java.rmi.Remote.

Cache Event polling S_NQ_EPT in Cluster Env.

Hi,
How to use event polling mechanism in Clustered Env to purge cache ? It works fine in non - clustered env.
I'm getting errors -
2009-12-16 15:40:45
[55003] The cache polling SELECT query failed for table S_NQ_EPT.
2009-12-16 15:40:45
[nQSError: 17001] Oracle Error code: 933, message: ORA-00933: SQL command not properly ended
at OCI call OCIStmtExecute.
[nQSError: 17010] SQL statement preparation failed.
2009-12-16 15:40:45
[55005] The cache polling delete statement failed for table S_NQ_EPT.
Anything special we have to do ?
Thanks,
Ayaps

Here's the resolution:
Hdr: 6877602 N/A ADMIN_TOOL 10.1.3.3.2 IMPORT PRODID-2025 PORTID-289
Abstract: INCORRECT SCHEMA WHEN IMPORTING THE EVENT POLLING TABLE
*** JOKENNE 03/10/08 01:47 am ***
script in 10.1.3.3.2 we are getting the following error in the NQServer.log
file:
"[56001] The cache polling event table UET has an incorrect schema."
After some investigation on a Paint.rpd file we found that this occurs when
we import the schema using an OCI connection i.e. File > Import > From
database > OCI
When we import using ODBC 3.5 option it works fine.

How to lookup the remote interface in JNDI of the cluster (glassfish)

I have write a simple sessionbean:
package authority;
import javax.ejb.Stateless;
* Session Bean implementation class LoginSessionBean
@Stateless
public class LoginSessionBean implements LoginSessionBeanRemote, LoginSessionBeanLocal {
     * Default constructor.
    public LoginSessionBean() {
        // TODO Auto-generated constructor stub
    @Override
    public boolean login(String name, String password)
        boolean result = false;
        System.out.println("User: " + name + " is login with password: " + password);
        return result;
    @Override
    public LoginSessionBeanRemote create()
        // TODO Auto-generated method stub
        return this;
}And I write a simple client to test it.
package test;
import javax.naming.InitialContext;
import javax.naming.NamingException;
import javax.rmi.PortableRemoteObject;
import authority.LoginSessionBean;
import authority.LoginSessionBeanRemote;
public class SessionBeanTestClient
    public static void main(String[] args)
        try
            InitialContext context = new InitialContext();
            Object obj = context.lookup(LoginSessionBean.class.getName());
            LoginSessionBeanRemote loginService = (LoginSessionBeanRemote)PortableRemoteObject.narrow(obj, LoginSessionBeanRemote.class);
            loginService.login("Jason", "password");
        catch (NamingException e)
            // TODO Auto-generated catch block
            e.printStackTrace();
}And it is OK, but after I deploy the sessionbean into a cluster with two instances on local host, it can not find the remote interface:
javax.naming.NameNotFoundException: authority.LoginSessionBean not found
     at com.sun.enterprise.naming.TransientContext.doLookup(TransientContext.java:216)Is it different for cluster (for IIOP port?), I want to test the HA solution, how could I start it?

Please look at http://otn.oracle.com/tech/java/oc4j/htdocs/oc4j-how-to.html. There is a How-To on local interface.
thanks
Debu

Configure webapp to use local or remote interface based on environment

Hi,
I have the following problem I am trying to solve (app server is Weblogic 10g by the way).
I have a webapp (war-project) and an ear project containing stateless session beans.
In our development environment I would like to deploy the war and ear separately, because that is easier for development (war-project can then be deployed as a directory, so changes to jsp's and such have immediate effect).
Deploying the war and ear separately means that the war-project can only call the session beans through their remote interfaces. For development this is not a problem.
In a production environment however we would like to package the war INSIDE the ear, so that the webapp can use the local interfaces to call the session beans, as this will improve performance.
So I am looking for a solution where I can configure the way beans are called. Local if the war is inside the ear, remote if the war is separate from the ear. I was thinking along the lines of packaging a properties file with the war that determines the mode (local or remote). Or maybe packaging a different deployment descriptor, if that is a possibility.
This is what I have so far:
Business Interface_
package be.cegeka.test.ejb3.service;
public interface TestService {
     public void doSomething();
Remote Interface_
package be.cegeka.test.ejb3.service;
public interface TestServiceRemote extends TestService {
Local Interface_
package be.cegeka.test.ejb3.service;
public interface TestServiceLocal extends TestService {
Bean implementation_
package be.cegeka.test.ejb3.service;
import javax.ejb.Local;
import javax.ejb.Remote;
import javax.ejb.Stateless;
@Stateless(mappedName="ejb/TestService")
@Local(TestServiceLocal.class)
@Remote(TestServiceRemote.class)
public class TestServiceBean implements TestService {
     @Override
     public void doSomething() {
          System.out.println("I'm doing something");
     @Override
     public void doSomethingLocal() {
          System.out.println("I'm doing something local");
JSF backing bean_
package be.cegeka.test.ui;
public class MyWebAppBackingBean {
     @EJB(mappedName="ejb/TestService", beanInterface=TestServiceRemote.class)
     private TestService testService;
     public String doSmth() {
          testService.doSomething();
          return "success";
}The part that I would like to be configurable is the "+beanInterface=TestServiceRemote.class+" part in the JSF backing bean. Depending on the environment (development vs. production), this should be "+beanInterface=TestServiceRemote.class+" for development and "+beanInterface=TestServiceLocal.class+" for production.
Is something like this possible?
I would like to avoid having to fill my web.xml with ejb-refs and having to do the JNDI-lookups myself if at all possible, to keep things as easy and clean as possible.
Any ideas are welcome!
Steven

Hey Manish,
comments inline.
"manish kumar" <[email protected]> wrote in message
news:1794338.1102557677925.JavaMail.root@jserv5...
So If they are supposed to be in the same JVM -
then
1. the cluster of 3 weblogic servers are three JVMs. So does that meanLOCAL INTERFACES CAN'T BE USED IN CLUSTERED ENVIORONMENT ?
Sure they can, but the calling component in an enterprise app will not call
an intance of the local bean on another node in the cluster. It will make a
by-reference call to the local component in the same EAR. Local interfaces
are not Remote, so they cannot be used with by-value RMI semantics.
>
2. ARE YOU SURE ABOUT THIS FACT THAT THAT LOCAL INTERFACES CAN BE USEDONLY INSIDE THE SAME APPLICATION, I THOUGHT IT CAN BE USED AS LONG AS CLINET
IS IN THE SAME JVM?
PLEASE CONFIRM.........!!!!!!!!!!!!!Dead sure:
http://e-docs.bea.com/wls/docs81/programming/classloading.html#1073506
http://e-docs.bea.com/wls/docs81/ejb/understanding.html#1126831
http://e-docs.bea.com/wls/docs61/ejb/cmp.html#1085452
http://www.theserverside.com/discussions/thread.tss?thread_id=14628
Also, search the ejb newsgroups. I know Rob Woollen has addressed this in
here somewhere. And, if Rob says it's so, trust me, it's so.
Bill

Home & remote interface

i deployed EJB with ORACLE deployejb tool,it succeed,from the sess_sh shell,i can see the bean itself.But where should i put the home & remote interface class,is that the web server defined class path just like using a javabean.i use tomcat as my webserver.i didn't do above,and when i import the Home & remote interface in my jsp file,it says can not compile the Home & remote interface class,is that the reason.
thanks

it did work!thans!
but i got another error in my jsp file:
"javax.naming.CommunicationException: Can't find SerialContextProvider "
it generate after the following code,can you tell me the reason:
try{
Hashtable env = new Hashtable();
env.put(Context.URL_PKG_PREFIXES, "oracle.aurora.jndi");
env.put(Context.SECURITY_PRINCIPAL, user);
env.put(Context.SECURITY_CREDENTIALS, password);
env.put(Context.SECURITY_AUTHENTICATION, ServiceCtx.NON_SSL_LOGIN);
ic = new InitialContext(env);
thanks!
null

Implement multiple remote interfaces

Hi there, I wanted to write an RMI application with an remote object which implement two different remote interfaces, each has its own remote methods.
eg:
public class TestImpl implements Test1, Test2 {
I only want to allow client to call the methods from Test1 remote interface (without Test2.class in the package). But exception NoClassDefFound of Test2 is caught. The reason I want to do like this is because I have a method in Test2 to unexport the remote object, call by server, so I want to hide it from client.
I implement the system with Activation, but I couldn't unexport the TestImpl from somewhere else on the same activation group by using Activatable.unexportObject(..). I only can unexport it inside TestImpl remote object itself by calling Activatable.unexportObject(this, true); otherwise, NoSuchObjectException is caught.
Thanks,
Jax

If X and A are in the same activation group, X can
unexport A as long as it has a reference to the actual
object A not its stub (because unexportObject() takes
an impl not a stub).Yes, it's true that X can unexport A because there are within the same JVM. But the problem is I can't get the actual object of A.
For example:
// assume this code is within X
ActivationGroupID agi = ActivationGroup.currentGroupID();
ActivationDesc desc = new ActivationDesc(agi, "A", location, data);
ActivationID id = ActivationGroup.getSystem().registerObject(desc);
Remote remote = id.activate(true);
The code above will only return a stub.
Even if I use Activatable.register(desc), it returns a stub as well.
How do I get the actual object?

RMI for multiple remote interfaces

We have a few classes that extend multiple remote interfaces. The stub and
skeleton created by the weblogic rmi compiler have strange names like
WLSkel3f2i4g44322n5l3i1r224p60697473f.class and
WLStub3f2i4g44322n5l3i1r224p60697473f.class.
Is this ok? How can we make it create "normal" class names?
Thanks,
Inbal

try using the -nomanglednames option on the weblogic.rmic compiler.
I think what you've got is OK - it's a result of the clever stubs weblogic
generates for rmi.
It's documented in the "Using weblogic RMI" guide.
-Dominic
Inbal Ronen wrote:
We have a few classes that extend multiple remote interfaces. The stub and
skeleton created by the weblogic rmi compiler have strange names like
WLSkel3f2i4g44322n5l3i1r224p60697473f.class and
WLStub3f2i4g44322n5l3i1r224p60697473f.class.
Is this ok? How can we make it create "normal" class names?
Thanks,
Inbal

Cache remote interface stub in clustered env

Similar Messages

Maybe you are looking for