OEL 5 machines, JRockit 1.6.0_26 JDK. Created a cluster with three instances, one instance per a machine. Deployed richAccess, restarted a cluster and find out that richAccess was not available for all instances, i.e. for some instances was seen 404 error. Then I've made a lot of experiments, i.e. stopped/started instances many times using different start commands, as start-cluster, start-instance, start-local-instance. I would say that, if instance is started using start-instance command, then in most cases richAccess would not be available. If to use start-local-instance, it would be available. For example, I've started one instance, using start-local-instance and two instances, using start-instance, richAccess was available only through the instance that used start-local-instance. I saw that issue for other app, for example stateless-simple, but for richAccess the issue was seen more often than for sateless-simple. I've attached rich richAccess, I did not see any errors in server.logs when richAccess was not available. But when this problem happened in the correspondent log file from instance start, I saw such messages:
[#|2011-11-18T16:55:13.445-0800|SEVERE|glassfish3.1.2|javax.enterprise.system.core.com.sun.enterprise.v3.server|_ThreadID=10;_ThreadName=Thread-2;|Cannot start container ejb
org.jvnet.hk2.component.ComponentException: injection failed on org.glassfish.ejb.startup.EjbContainerStarter.ejbContainerUtilImpl with interface com.sun.ejb.containers.EjbContainerUtil
I've attached such log file.