Failover in clustered environment does not work with JSF 2, Richfaces 4, Tomcat 7 - apache

Question

I have a clustered environment with Apache 2.2.6 and mod_proxy pointing to Tomcat 7.0.26 through AJP13 with Sticky Sessions.
The httpd.conf configuration is like this:
<Proxy balancer://mycluster2>
BalancerMember ajp://192.168.0.1:8009 route=tomcat1
BalancerMember ajp://192.168.0.2:8009 route=tomcat2
ProxySet lbmethod=byrequests
</Proxy>
ProxyPass /MyApp balancer://myCluster2/MyApp stickysession=JSESSIONID
ProxyPassReverse /MyApp https://apache_server/MyApp
In my tomcat server.xml file, I have properly configured the cluster inside the <Host> tag (posted just the tomcat1 file, tomcat2 is the same changing only the ip):
<Executor name="tomcatThreadPool" namePrefix="catalina-exec-"
maxThreads="150" minSpareThreads="4"/>
<Connector executor="tomcatThreadPool"
port="8080" protocol="HTTP/1.1"
connectionTimeout="20000"
redirectPort="8443" />
<Connector port="8009" protocol="AJP/1.3" redirectPort="8443" />
...
<Cluster className="org.apache.catalina.ha.tcp.SimpleTcpCluster"
channelSendOptions="8">
<Manager className="org.apache.catalina.ha.session.DeltaManager"
expireSessionsOnShutdown="false" notifyListenersOnReplication="true" />
<Channel className="org.apache.catalina.tribes.group.GroupChannel">
<Membership className="org.apache.catalina.tribes.membership.McastService" address="228.0.0.4" port="45564" frequency="500" dropTime="3000" />
<Receiver className="org.apache.catalina.tribes.transport.nio.NioReceiver" address="192.168.0.1" port="4000" autoBind="100" selectorTimeout="5000" maxThreads="6" />
<Sender className="org.apache.catalina.tribes.transport.ReplicationTransmitter">
<Transport className="org.apache.catalina.tribes.transport.nio.PooledParallelSender" />
</Sender>
<Interceptor className="org.apache.catalina.tribes.group.interceptors.TcpFailureDetector" />
<Interceptor className="org.apache.catalina.tribes.group.interceptors.MessageDispatch15Interceptor" />
<Interceptor className="org.apache.catalina.tribes.group.interceptors.ThroughputInterceptor" />
</Channel>
<Valve className="org.apache.catalina.ha.tcp.ReplicationValve" filter="" />
<Valve className="org.apache.catalina.ha.session.JvmRouteBinderValve" />
<ClusterListener className="org.apache.catalina.ha.session.JvmRouteSessionIDBinderListener" />
<ClusterListener className="org.apache.catalina.ha.session.ClusterSessionListener" />
</Cluster>
This configuration works like a charm with any jsp webapp, it replicates sessions and works perfectly on failover with the classic failover step test case:
1.- Tomcat1 starts.
2.- Tomcat2 starts.
3.- A request is processed by the balancer: https://apache_server/MyApp and sent to Tomcat1.
4.- Some operations are performed (i. e. refresh page with a counter as session attribute).
5.- Tomcat1 is killed.
6.- User refresh page and the session counter follows counting in Tomcat2.
So, at that point I have clear that there are no misconfigurations neither on apache nor in Tomcat. Then I go with MyApp. First of all, it has the <distributable/> tag in web.xml.
Next, I deploy it on Tomcat1 and Tomcat2 successfully and I see that Tomcat is multicasting and sharing information between nodes for my application:
INFO: Gestor [/MyApp], requiriendo estado de sesión desde org.apache.catalina.tribes.membership.MemberImpl[tcp://{192, 168, 0, 1}:4000,{192, 168, 0, 1},4000, alive=5113068, securePort=-1, UDP Port=-1, id={-31 113 14 29 99 -58 77 -75 -111 66 -103 86 102 -108 120 61 }, payload={}, command={}, domain={}, ]. Esta operación se agotará si no se recibe estado de sesión dentro de 60 segundos.
19-sep-2013 18:49:51 org.apache.catalina.tribes.group.interceptors.ThroughputInterceptor report
INFO: ThroughputInterceptor Report[
Tx Msg:1 messages
Sent:0,00 MB (total)
Sent:0,00 MB (application)
Time:0,00 seconds
Tx Speed:0,12 MB/sec (total)
TxSpeed:0,12 MB/sec (application)
Error Msg:0
Rx Msg:1 messages
Rx Speed:0,00 MB/sec (since 1st msg)
Received:0,00 MB]
19-sep-2013 18:49:51 org.apache.catalina.ha.session.DeltaManager waitForSendAllSessions
INFO: Gestor [/MyApp]; estado de sesión enviado a las 19/09/13 18:49 recibido en 106 ms.
I try to reproduce the before mentioned navigation, and I can see in Tomcat Managers for both nodes my domain objects getting replicated (all of them implement Serializable).
For some reason, my JSF com.sun.faces.renderkit.ServerSideStateHelper.LogicalViewMap does not replicate all the objects. In fact, the Backup node has always one object less than Primary node in that session attribute.
At point 6, after killing Tomcat1 and refreshing page, session is not recovered and user is sent to logout screen invalidating the session.
MyApp worked before in a non-clustered environment. Here is the configuration for STATE-SAVING of MyApp web.xml:
<context-param>
<param-name>javax.faces.STATE_SAVING_METHOD</param-name>
<param-value>server</param-value>
</context-param>
Even I tried to upgrade from JSF 2.1.4 to 2.1.21 and I get the same error. I can't upgrade to 2.2.3 without changing many things in MyApp (this is in production and it's a long time development project, so I must try all before considering the refactorization of the whole project).
I tried to put also this in my web.xml and it is even worse, since it replicates less objects in LogicalViewMap:
<context-param>
<param-name>com.sun.faces.serializeServerState</param-name>
<param-value>true</param-value>
</context-param>
My faces-config.xml has nothing special.
I also tried to upgrade JSF to 2.2.3 but in this case project wouldn't work at all since I'm using Richfaces 4.0.0-final and I would need to refactorizate some more code.
At this time I'm thinking that JSF 2 is not compatible with Tomcat Clustering. Does anybody configured a project with Tomcat Clustering and JSF 2?
Any help will be appreciated.

Answer 1

I had a similar issue but on a weblogic environment. Please see jsf session failover
Hopefully this might assist somehow.