Clustering using Oracle ADF - CHECK_STATE_SERIALIZATION=all

Question

We are using Oracle ADF 11.1.1.6. We have started testing our Oracle ADF application for the cluster compatibility and used following document. We have coded all recommendations given in the document including enabling all applicable java classes to serialization and also setup adf_config.xml <adfc:adf-scope-ha-support>true</adfc:adf-scope-ha-support> and <adf:adf-scope-ha-support>true</adf:adf-scope-ha-support> exactly as per given in the document.
http://docs.oracle.com/cd/E12839_01/core.1111/e10106/adf.htm
Now, to check High Availability on a standalone web logic server for debugging purpose we configured Java Option "-Dorg.apache.myfaces.trinidad.CHECK_STATE_SERIALIZATION=all" as per the recommendation.
After setting this flag system started throwing bunch of errors similar to the error given below.
[2012-07-26T14:17:43.602-04:00] [OPERA_ADF1] [ERROR] [] [org.apache.myfaces.trinidadinternal.config.CheckSerializationConfigurator$MutatedBeanChecker] [tid: [ACTIVE].ExecuteThread: '17' for queue: 'weblogic.kernel.Default (self-tuning)'] [userId: edozer] [ecid: 64915ee7c1860d85:40a983d1:138bf4550b2:-7ffd-0000000000005fcb,0] [APP: OperaApplications_o90_smoke#V9.0.0] Failover error: Serialization of Session attribute:data has changed from oracle.adf.model.servlet.HttpBindingContext@365f8709 to oracle.adf.model.servlet.HttpBindingContext@36b6565b without the attribute being dirtied
The question is what are we missing here?
Please note that our application is still missing ControllerContext.getInstance().markScopeDirty(pageFlowScope); but when I tried to create a sample application and deliberately left the "markScopeDirty" using below code the above mentioned error was not reproduceable.
public class myBackingBean
     public void changeValueButton( ActionEvent ae )
     Map<String, Object> pageFlowScope = AdfFacesContext.getCurrentInstance().getPageFlowScope();
          pageFlowTestBean obj = ( pageFlowTestBean )pageFlowScope.get( "pageFlowTestBean" );
          System.out.println( "Old Value " + obj.getYVar() );
          obj.setYVar( "New Value" );
          /* ControllerContext.getInstance().markScopeDirty(pageFlowScope); */
Can anyone shade more light on it?
Also, I would like to know that can someone guide me what is the best practice to perform ControllerContext.getInstance().markScopeDirty(pageFlowScope);. Because developers can easily miss this statement and I didn't find any easyway to search the code that is missing this statement (I have already tried Jdeveloper's Audit feature and it works for very few cases).

Answer 1

We checked the ADF source (pasted below). It reads that markScopeDirty is getting skipped if ADF_SCOPE_HA_SUPPORT is not configured but when we debug our application the serialization fires after every pageScope/viewScope variable gets changed regardless of markScopeDirty() is being called or not. That signifies that markScopeDirty() call is not needed. Do we really need markScopeDirty() everywhere in our application or this is just a public helper method to manually trigger replication for some exceptional case? Can anyone from team Oracle explain this behavior?
@Override
public void markScopeDirty(Map<String, Object> scopeMap)
Boolean handleDirtyScopes =
(Boolean)ControllerConfig.getProperty(ControllerProperty.ADF_SCOPE_HA_SUPPORT); //checks for adf-scope-ha-support
if (handleDirtyScopes.booleanValue())
if (!(scopeMap instanceof SessionBasedScopeMap))
String msg = AdfcExceptionsArb.get(AdfcExceptionsArb.NOT_AN_ADF_SCOPE);
throw new IllegalArgumentException(msg);
((SessionBasedScopeMap)scopeMap).markDirty();
}