[JBoss-dev] jboss-head status

Scott M Stark scott.stark at jboss.org
Thu Aug 24 11:48:54 EDT 2006


One example of a mixed mbean service notion causing reploy problems is 
the org.jboss.ejb.plugins.CMPPersistenceManager. This implements the 
Service contract, but does not extend ServiceMBeanSupport and so direct 
calls to start by the EjbModule.startService have no guarentee that the 
CMPPersistenceManager has been created.

Scott M Stark wrote:
> There are still a number of sporadic 
> InstanceNotFoundException/NullPointerException errors are due to the 
> inconsistent handling of mbean services as services to be managed by 
> the ServiceController. In a number places there is code like this in 
> the EntityContainer:
>
>         // Try to register the instance cache as an MBean
>         try
>         {
>            ObjectName containerName = super.getJmxName();
>            Hashtable props = containerName.getKeyPropertyList();
>            props.put("plugin", "cache");
>            ObjectName cacheName = new 
> ObjectName(containerName.getDomain(), props);
>            server.registerMBean(instanceCache, cacheName);
>         }
>         catch(Throwable t)
>         {
>            log.debug("Failed to register cache as mbean", t);
>         }
>
> Subsequent use of this mbean as a jboss mbean service can result in 
> the ServiceController taking ownership of the mbean and removing it on 
> undeployment, and this can cause the InstanceNotFoundException when a 
> container redundantly tries to remove the mbean it registered. This 
> has to be cleaned up to only use the ServiceController to manage the 
> mbeans. The issue is, this is all deployer code that needs to be 
> updated to the vdf anyway. Maybe this should be done as it will have 
> to be done regardless, but I don't want to waste too much time away 
> from getting deployers ported to the VDF.
>
> The NullPointerExceptions showing up tend to result in a failure at 
> one point triggering a cascade of destroys on services that were never 
> fully created. Its sporadic so I don't know if its because of some 
> lifecycle calls not going through the controller, configuration of 
> mbeans happening outside of lifecycle, etc.
>
> _______________________________________________
> jboss-development mailing list
> jboss-development at lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/jboss-development




More information about the jboss-development mailing list