Error invoking a remote EJB method from a JSF managed bean in a portlet - jboss

Question

I am trying to deploy a JSF portlet into a Liferay portal 6.2 running on top of a JBoss EAP 6.4, with OpenJDK 8. The portlet is deployed as a WAR file, but invokes some methods of an EJB through a remote interface. The EJB is deployed as an EAR file directly on JBoss.
On my developing machine it is working fine. But in a test machine it is not. There are exactly the same WAR and EAR files, the standalone.xml setup is almost the same for both machines, and also the JBoss modules are the same.
The stacktrace printed in the log is the following:
JBWEB000236: Servlet.service() for servlet Adminis Servlet threw exception: javax.portlet.faces.BridgeException: javax.faces.FacesException
at com.liferay.faces.bridge.BridgePhaseRenderImpl.execute(BridgePhaseRenderImpl.java:91) [liferay-faces-bridge-impl-3.2.4-ga5.jar:3.2.4-ga5 (Ephesus / Feb 15, 2014 AD)]
at com.liferay.faces.bridge.BridgeImpl.doFacesRequest(BridgeImpl.java:112) [liferay-faces-bridge-impl-3.2.4-ga5.jar:3.2.4-ga5 (Ephesus / Feb 15, 2014 AD)]
...
at java.lang.Thread.run(Thread.java:745) [rt.jar:1.8.0_121]
Caused by: javax.faces.FacesException
at com.sun.faces.context.ExceptionHandlerImpl.handle(ExceptionHandlerImpl.java:139) [jsf-impl-2.1.28.redhat-8.jar:2.1.28.redhat-8]
at org.icefaces.impl.application.ExtendedExceptionHandler.handle(ExtendedExceptionHandler.java:113) [icefaces-3.3.0.jar:]
... 194 more
Caused by: java.util.NoSuchElementException
at java.util.HashMap$HashIterator.nextNode(HashMap.java:1439) [rt.jar:1.8.0_121]
at java.util.HashMap$KeyIterator.next(HashMap.java:1461) [rt.jar:1.8.0_121]
at org.jboss.ejb.client.EJBClientContext.getEJBReceiver(EJBClientContext.java:725) [jboss-ejb-client-1.0.30.Final-redhat-1.jar:1.0.30.Final-redhat-1]
at org.jboss.ejb.client.EJBInvocationHandler.invoke(EJBInvocationHandler.java:146) [jboss-ejb-client-1.0.30.Final-redhat-1.jar:1.0.30.Final-redhat-1]
...
at com.sun.proxy.$Proxy459.obtenerUsuarioPorCodigo(Unknown Source)
at com.sofis.adminis.delegates.PermisosDelegate.obtenerUsuarioPorCodigo(PermisosDelegate.java:450) [adminis-delegate-1.0.jar:]
at com.sofis.security.web.UsuariosSessionBeanWeb.autenticar(UsuariosSessionBeanWeb.java:1927) [classes:]
at com.sofis.security.web.listeners.SofisAutologinLiferayPhaseListener.afterPhase(SofisAutologinLiferayPhaseListener.java:54) [classes:]
at com.sun.faces.lifecycle.Phase.handleAfterPhase(Phase.java:189) [jsf-impl-2.1.28.redhat-8.jar:2.1.28.redhat-8]
at com.sun.faces.lifecycle.Phase.doPhase(Phase.java:107) [jsf-impl-2.1.28.redhat-8.jar:2.1.28.redhat-8]
... 198 more
I rised the log level to "org.jboss.ejb.client" and now I see two interesting messages just before the stactrace:
22:13:16,684 DEBUG [org.jboss.ejb.client.EJBClientContext] (ajp-/0.0.0.0:8009-5) org.jboss.as.ejb3.remote.LocalEJBReceiverPreferringDeploymentNodeSelector#4bc9113e deployment node selector selected node for appname=SS-ADMINIS-ear-1.0,modulename=SS-ADMINIS-MAVEN-EJB-1.0,distinctname=
22:13:16,687 DEBUG [org.jboss.ejb.client.EJBClientContext] (ajp-/0.0.0.0:8009-5) Selected node doesn't belong to eligible receivers. Continuing with a random eligible receiver
Note that there is no node name in both messages (two blanks before "node").
In my developing machine, where it works, the first message does include a node name, and the second message does not appear:
18:56:37,902 DEBUG [org.jboss.ejb.client.EJBClientContext] (http-/0.0.0.0:8080-2) org.jboss.as.ejb3.remote.LocalEJBReceiverPreferringDeploymentNodeSelector#6db73304 deployment node selector selected spio node for appname=SS-ADMINIS-ear-1.0,modulename=SS-ADMINIS-MAVEN-EJB-1.0,distinctname=
The only diference is that in my machine I am going directly with the browser to the server ([Browser] -- HTTP --> [JBoss]) while in the testing machine I go through a load balancer ([Browser] -- HTTP --> [Apache] -- AJP --> [JBoss]). That is reflected in the messages (ajp-/0.0.0.0 vs http-/0.0.0.0).
The exception is thrown when I try to invoke a method from the EJB through the remote interface. The reference to the local stub is gotten by means of a lookup, and the result is not null. There are no errors at deploy and the lookup does not throw an error neither.
What can be happening or what am I missing?

Answer 1

A friend of mine found the solution.
Something a did't say is that the testing machine was in fact a docker container. And it seems that there is a bug in JBoss with docker making the former to not read the hostname to pick a node name, so de node name when it starts is empty. So the message above a empty name node being selected is shown, and then the error.
The workaround is to specify a node name with the -Djboss.node.name parameter to JBoss. I put that setting in the standalone.conf, restarted the JBoss and it worked fine. Now, the first message shows the name I specified, and the second message does not appear, nor the exception.