Trying to Get Second Level Cahce Entities

[serdar]

Hello,

I am a newbie in Jboss and Infinispan and this is my first post

I am interested in infinispan cahce and trying to learn its capabilities and use cases. Trying to understand what is going inside second-level-cache for hibernate. I am using JBoss "jboss-as-web-7.0.2.Final" and standalone.bat.

After a few days of searching, finally I could set the required configurations and inject the infinispan cache container to my code as follows:

@ManagedBean
public class InfinispanManager<K, V> {
  @Resource(lookup="java:jboss/infinispan/container/hibernate")
  private org.infinispan.manager.CacheContainer container;
  private org.infinispan.Cache<K, V> cache;
   
  @PostConstruct
  public void start() {
    this.cache = this.container.getCache("entity");
  }
  
  public org.infinispan.Cache<K, V> getCache()
  {
            return cache;
  }
}

I persist my @Cachable @Entity KVPair bean as:

kvPair = new KVPair(key, value);
entityManager.persist(kvPair);

I inject the infinispanManager as:

@Inject
private InfinispanManager<Object,Object> manager;

But entity cache always seems empty (After a entityManager.persist(kvPair) ) when I look as below:

manager.getCache().keySet().isEmpty()

Is my approach is correct ? Am I missing something ?

Here is my persistence.xml persistence-unit

   <persistence-unit name="bmtDatabase">
      <provider>org.hibernate.ejb.HibernatePersistence</provider>
       <!-- the datasource JNDI name as it is configured in the Application Server config files -->
      <jta-data-source>java:jboss/datasources/ExampleDS</jta-data-source>
      <properties>
          <property name="hibernate.cache.use_second_level_cache" value="true" />
          <property name="hibernate.cache.use_query_cache" value="true" />
          <property name="hibernate.cache.region.factory_class" value="org.hibernate.cache.infinispan.JndiInfinispanRegionFactory" />
          <property name="hibernate.cache.infinispan.cachemanager" value="java:jboss/infinispan/container/hibernate" />
          <property name="hibernate.hbm2ddl.auto" value="create-drop" />
          <property name="hibernate.show_sql" value="true" />
      </properties>
   </persistence-unit>

Here is my standalone.xml hibernate cahce container:

        <subsystem xmlns="urn:jboss:domain:infinispan:1.0" default-cache-container="hibernate">
            <cache-container name="hibernate" default-cache="local-query" jndi-name="java:jboss/infinispan/container/hibernate">
                <local-cache name="entity">
                    <eviction strategy="LRU" max-entries="10000"/>
                    <expiration max-idle="100000"/>
                </local-cache>
                <local-cache name="local-query">
                    <eviction strategy="LRU" max-entries="10000"/>
                    <expiration max-idle="100000"/>
                </local-cache>
                <local-cache name="timestamps">
                    <eviction strategy="NONE"/>
                </local-cache>
            </cache-container>
        </subsystem>

And this is the tricky part. I should include the infinispan dependicy to my deployment by manifest.mf. I do that by maven in my pom.xml as below:

         <plugin>
            <artifactId>maven-war-plugin</artifactId>
            <version>2.1.1</version>
            <configuration>
               <!-- Java EE 6 doesn't require web.xml, Maven needs to catch
                  up! -->
               <failOnMissingWebXml>false</failOnMissingWebXml>
                            <archive>
                                        <manifestEntries>
                                          <Dependencies>org.infinispan export</Dependencies>
                                        </manifestEntries>
                            </archive>               
            </configuration>
         </plugin>

This post is a question for me and I hope the examples above also may be helpful for someone else.

[galder.zamarreno]

Rather than inspecting the cache yourself, the easiest thing to do is to check the 2nd level cache stats via SessionFactory.getStatistics().getSecondLevelCacheStatistics(...) and pass the name of the entity. After a persist, the getPutCount should be 1.

I dunno details about JBoss Web, but with AS 7.1, you don't need to specify the region factory class or any of that stuff, see https://docs.jboss.org/author/display/AS71/JPA+Reference+Guide#JPAReferenceGuide-UsingtheInfinispansecondlevelcache - just set hibernate.cache.use_second_level_cache to true and that's all you need to do. AS7 takes care of the rest. This avoids misconfiguration.

Also, the standalone.xml in AS 7.1 already has a cache container for Hibernate 2LC, so nothing you need to enable there.

[serdar]

Thanks for the reply. I can already see the statistics of the 2LC from JBoss Admin console, Put Count increases. What I am corious about is to touch the entities by using the cache not using the EntityManager. Seocondly, I was not sure in default configuration hibernate cache container has a JNDI name. I have used the extra configuration in order to give it a JNDI name but maybe the container which I gave the jndi-name is not the one hibernate uses.

Now I removed the extra configuration in persistence.xml and using default standalone-ha.xml JBoss 7.1.x in a cluster and @Resource(lookup="java:jboss/infinispan/container/hibernate") still works, it is the default name I think, but still manager.getCache().keySet().isEmpty() = true always

[galder.zamarreno]

Accessing the cache in 2LC use case directly is something we do not recommend. The keySet is probably empty because the cache your accessing is not the 2LC one.

[galder.zamarreno]

If you explain what you're trying to achieve by accessing the cache entries directly, maybe we can explain an alternative way to achieve what you're trying to do.

[serdar]

Ok let me explain what I want to do:

1- I have a big amount of data loaded from database and should be persisted.

2- I have to process and modify that data many times in a second.

3- Many clients are trying to get the snapshot of the some part of the data in every second.

4- I want to distrubute that data so I can scale the process and clients.

5- I am planning to make that scalability by distributing my data by infinispan 2LC and parallel process the owned cache data in each node. That is why I want to get the cache in 2LC bypassing the EntityManager.

Maybe doing that by 2LC is not a good idea. Maybe I should use infinispan directly and make the database persistence by a infinispan store not by JPA ?

[serdar]

Galder, maybe I am trying to re-invent some kind of Distributed Execution Framework or MapReduce model .  I am working on these issues an trying to understand if distributed execution is my solition ? What do you thik ?