[infinispan-issues] [JBoss JIRA] (ISPN-3086) Infinite loop when creating more than two Distributed Cache nodes

[Balazs Zsoldos (JIRA)]

     [ https://issues.jboss.org/browse/ISPN-3086?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Balazs Zsoldos updated ISPN-3086:
---------------------------------

    Attachment: infinispan-three-nodes.zip


Attached a very simple application that shows how to reproduce the problem. I tested it again with the newest versions:

With version 5.3.0.Final during creating the third cache UDP bandwidth is at max. forever.

With version 6.0.0.Alpha2 there is no UDP bandwidth at all but it is still waiting on the getCacheManager function call.

If you change the cache type in the xml to repl everything works well.

To change the version simply change the version of the dependency in the pom file.

If you need I can copy and paste the stacktrace here but I think it is easier
                
> Infinite loop when creating more than two Distributed Cache nodes
> -----------------------------------------------------------------
>
>                 Key: ISPN-3086
>                 URL: https://issues.jboss.org/browse/ISPN-3086
>             Project: Infinispan
>          Issue Type: Bug
>          Components: JCache
>    Affects Versions: 5.3.0.Beta1
>            Reporter: Balazs Zsoldos
>            Assignee: Galder Zamarreño
>            Priority: Critical
>              Labels: cache, distribution
>             Fix For: 6.0.0.Final
>
>         Attachments: infinispan-three-nodes.zip
>
>
> I created a very simple application based on https://docs.jboss.org/author/display/ISPN/Using+Infinispan+as+a+JCache+provider#UsingInfinispanasaJCacheprovider-ClusteringJCacheinstances
> I changed the code to created distributed caches instead of replicated ones. It works until two nodes but if there are three nodes it starts an infinite loop.
> *The modified XML contains:*
>    <namedCache name="namedCache">
>       <clustering mode="dist" />
>    </namedCache>
> *The modified java code contains:*
> CacheManager cacheManager1 = Caching.getCacheManager(new TestClassLoader(tccl), "infinispan-jcache-cluster.xml");
> CacheManager cacheManager2 = Caching.getCacheManager(new TestClassLoader(tccl), "infinispan-jcache-cluster.xml");
> CacheManager cacheManager3 = Caching.getCacheManager(new TestClassLoader(tccl), "infinispan-jcache-cluster.xml");
> *Symptom:*
> The code waits on the third command while there is an infinite loop on a background thread. I guess the benefit of distribution would come with more than two nodes but I cannot test it due to the problem above.
> Although I used infinispan via JCache I think this issue is not related to the JCache API but it comes from the infinispan core.
> *BTW:* Distributed caches seem to me about 30% slower than replicated caches concerning to put and delete functions with two nodes. I guess the benefit would come out between the two topologies with more nodes.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira