EHCache Replication using JGroups over TCP ?

As we transition to the new Hyland Community, some post links may be temporarily disrupted as we finalize redirects. We apologize for the inconvenience and are working to resolve this as quickly as possible. Details.

deepak1987

What is the difference between Default EHCache Replication using JGroups over TCP protocol and Alfresco's implemenation for EHCache Replication using JGroups over TCP protocol ?


I mean to say,
In alfresco-global.properties file,  I have configured JGroups for Ehcache Replication using TCP protocol:

alfresco.cluster.name = alfrescocluster
alfresco.jgroups.defaultProtocol = TCP
alfresco.tcp.initial_hosts = host_1[7800],host_2[7800]
alfresco.jgroups.bind_address = host_1
alfresco.tcp.start_port = 7800
‍‍‍‍‍‍‍

and

In ehcache-custom.xml file,
1) CacheManagerPeerProviderFactory class is org.alfresco.repo.cache.AlfrescoCacheManagerPeerProviderFactory.
2) CacheManagerPeerListenerFactory class is net.sf.ehcache.distribution.RMICacheManagerPeerListenerFactory.
3) CacheEventListenerFactory class is net.sf.ehcache.distribution.RMICacheReplicatorFactory.

<cacheManagerPeerProviderFactory
         class="org.alfresco.repo.cache.AlfrescoCacheManagerPeerProviderFactory"
         properties="heartbeatInterval=5000,
                          peerDiscovery=automatic,
                          multicastGroupAddress=230.0.0.1,
                          multicastGroupPort=4446"
/>

<cacheManagerPeerListenerFactory
                                  class="net.sf.ehcache.distribution.RMICacheManagerPeerListenerFactory"
                                  properties="socketTimeoutMillis=10000"
/>

<defaultCache
       …
   >
             <cacheEventListenerFactory
                         class="net.sf.ehcache.distribution.RMICacheReplicatorFactory"
                          properties="replicatePuts = false,
                                replicateUpdates = true,
                                replicateRemovals = true,
                                replicateUpdatesViaCopy = false,
                                replicateAsynchronously = false"/>
</defaultCache>
‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍

If I use default classes of Ehcache for replication using JGroups over TCP
as
1) CacheManagerPeerProviderFactory = net.sf.ehcache.distribution.jgroups.JGroupsCacheManagerPeerProviderFactory.
2) CacheEventListenerFactory = net.sf.ehcache.distribution.jgroups.JGroupsCacheReplicatorFactory.

I have got the following configuration from
http://www.ehcache.org/documentation/replication/jgroups-replicated-caching#example-configuration-using-tcp-unicast

 
<cacheManagerPeerProviderFactory
          class="net.sf.ehcache.distribution.jgroups.JGroupsCacheManagerPeerProviderFactory"
          properties="connect=TCP(bind_port=7800):
                            TCPPING(initial_hosts=host_1[7800],host_2[7800];port_range=10;timeout=3000;
                                          num_initial_members=3):
                            VERIFY_SUSPECT(timeout=1500):
                            pbcast.NAKACK(retransmit_timeout=3000):
                            pbcast.GMS(join_timeout=5000;print_local_addr=false)"  
           propertySeparator="::"
/>

<defaultCache
             ….
         >  
            <cacheEventListenerFactory
                       class="net.sf.ehcache.distribution.jgroups.JGroupsCacheReplicatorFactory"
                       properties="replicatePuts = false,
                                         replicateUpdates = true,
                                         replicateRemovals = true,
                                         replicateUpdatesViaCopy = false,
                                         replicateAsynchronously = false" />

</defaultCache>

‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍

Will it impact on performance while replicating caches on nodes in a cluster??
If YES then, how and why ?

I tried with above configuration using Default implementation of Ehcache instead of Alfresco's method and did testing by following instructions in https://wiki.alfresco.com/wiki/Cluster_Configuration_V2.1.3_and_Later#Testing_the_cluster. It worked, but not sure about pros and cons.

Has anybody tried or did such kind of experiment??

Your inputs are highly appreciated.

jspuchau

Hi, I dont know if that stuff is going to work, but you can check them more extensively using the jconsole in both nodes to check if the objects are correctly replicating from one node to another.

If you test with persisted objects you can obtain false positives as the information is being persisted. Not persisted objects as user sessions are better indicators to test if the cache is replicating.

Regards
JSP