2

We have multiple instances of Ofbiz/Opentaps running. All the instances talk to the same database. There are many tables that are rarely updated hence they are cached and all the instances maintain their individual copies of cache as a standard Ofbiz cache mechanism. But in rare situations when we update some entity using one of many instances then all other instances keep showing dirty cache data. So it requires a manual action to go and clear all the cache copies on other instances as well.

I want this cache clearing operation on all the instances to happen automatically. On Ofbiz confluence page here there is a very brief mention of "Distributed cache clearing". It relies on JMS it seems so whenever an instance's cache is cleared it sends notification over JMS to a topic and other instances subscribing to the same JMS topic clear their corresponding copies of cache upon this notification. But I could not find any other reference or documentation on how to do that? What are the files that need to be updated to set it all up in Ofbiz? An example page/link is what I'm looking for.

MohammadHosseinZiyaaddini
  • 658
  • 1
  • 6
  • 17
anubhava
  • 761,203
  • 64
  • 569
  • 643
  • 6
    @Downvoters: Wow 2 downvotes in a min on a 2+ year old question. Please have courtesy of leaving a comment at least. – anubhava Jun 28 '13 at 08:03

3 Answers3

3

Alright I believe I've figured it all out. I have used ActiveMQ as my JMS broker to set it up so here are the steps in Ofbiz to make it working:

1. Copy activemq-all.jar to framework/base/lib folder inside your Ofbiz base directory.

2. Edit File base/config/jndiservers.xml: Add following definition inside <jndi-config> tag: 

   <jndi-server name="activemq"
     context-provider-url="failover:(tcp://jms.host1:61616,tcp://jms.host2:61616)?jms.useAsyncSend=true&amp;timeout=5000"
     initial-context-factory="org.apache.activemq.jndi.ActiveMQInitialContextFactory"
     url-pkg-prefixes=""
     security-principal=""
     security-credentials=""/>

3. Edit File base/config/jndi.properties: Add this line at the end: 

topic.ofbiz-cache=ofbiz-cache

4. Edit File service/config/serviceengine.xml: Add following definition inside <service-engine> tag: 

    <jms-service name="serviceMessenger" send-mode="all">
       <server jndi-server-name="activemq"
       jndi-name="ConnectionFactory"
       topic-queue="ofbiz-cache"
       type="topic"
       listen="true"/>
    </jms-service>

5. Edit File entityengine.xml: Change default delegator to enable distributed caching: 

<delegator name="default" entity-model-reader="main" entity-group-reader="main" entity-eca-reader="main" distributed-cache-clear-enabled="true">

6. Edit File framework/service/src/org/ofbiz/service/jms/AbstractJmsListener.java: This one is probably a bug in the Ofbiz code

Change following line from:

this.dispatcher = GenericDispatcher.getLocalDispatcher("JMSDispatcher", null, null, this.getClass().getClassLoader(), serviceDispatcher);

To:

this.dispatcher = GenericDispatcher.getLocalDispatcher("entity-default", null, null, this.getClass().getClassLoader(), serviceDispatcher);

7. And finally build the serviceengine code by issuing following command:

ant -f framework/service/build.xml

With this entity data changes in Ofbiz on one instances are immediately propagated to all the other Ofbiz instances clearing cache line item on its own without any need of manual cache clearing.

Cheers.

anubhava
  • 761,203
  • 64
  • 569
  • 643
2

I have a added a page on this subject in OFBiz wiki https://cwiki.apache.org/OFBIZ/distributed-entity-cache-clear-mechanism.html. Though it's well explained here, the OFBiz wiki page adds other important information.

Note that the bug reported here has been fixed since, but another is currently pending, I should fix it soon https://issues.apache.org/jira/browse/OFBIZ-4296

Jacques

JacquesLeRoux
  • 577
  • 4
  • 10
  • Awesome, thanks a lot for putting it up on official docs. I had to make more changes in the Ofbiz JMS Service engine to handle exceptions gracefully and more importantly behave asynchronously inside `runAsync` method. Earlier code was rolling back Orders in case of JMS provider not available (a strict no-no for any e-commerce site I believe for cache clear operation). – anubhava May 28 '11 at 15:40
  • @Jacques: The linked page is now 404. – Amal Murali Feb 01 '14 at 18:02
  • Thanks Amal, nor sure why I did not get your message from/by stackoverflow. Here the actualised link https://cwiki.apache.org/confluence/display/OFBIZ/Distributed+Entity+Cache+Clear+(DCC)+Mechanism – JacquesLeRoux Aug 20 '17 at 14:41
2

Yes, I fixed this behaviour sometimes ago at http://svn.apache.org/viewvc?rev=1090961&view=rev. But it still needs another fix related to https://issues.apache.org/jira/browse/OFBIZ-4296.

The patch below fixes this issue locally, but still creates 2 listeners on clusters, not sure why... Still investigating (not a priority)...

Index: framework/entity/src/org/ofbiz/entity/DelegatorFactory.java
===================================================================
--- framework/entity/src/org/ofbiz/entity/DelegatorFactory.java (revision 1879)
+++ framework/entity/src/org/ofbiz/entity/DelegatorFactory.java (revision 2615)
@@ -39,10 +39,10 @@

             if (delegator != null) {
+                // setup the distributed CacheClear
+                delegator.initDistributedCacheClear();
+
                 // setup the Entity ECA Handler
                 delegator.initEntityEcaHandler();
                 //Debug.logInfo("got delegator(" + delegatorName + ") from cache", module);
-                
-                // setup the distributed CacheClear
-                delegator.initDistributedCacheClear();



             return delegator;

Please notify me using @JacquesLeRoux in your post, if ever you have something new to share.

JacquesLeRoux
  • 577
  • 4
  • 10