Starting replicated cache with NON_XA transaction fails in WildFly 8.0.0.CR1

[hchiorean]

Hi,

I'm attempting to start a Wildfly 8.0.0.CR1 server using a configuration similar to the default standalone-ha.xml, using the following clustered Infinispan cache:

            <!--Clustered cache container-->

            <cache-container name="modeshape-repository-clustered" default-cache="clustered-repo" module="org.modeshape">

                <transport lock-timeout="60000"/>

                <replicated-cache name="repo-clustered" mode="SYNC" batching="true">

                   <transaction mode="NON_XA"/>

                    <file-store path="modeshape/store/repo-clustered" passivation="false" purge="false"/>

                </replicated-cache>

            </cache-container>

While the <transaction/> element is present, this fails with:

14:40:48,298 ERROR [org.jboss.msc.service.fail] (MSC service thread 1-6) MSC000001: Failed to start service jboss.infinispan.modeshape-repository-clustered.repo-clustered.config: org.jboss.msc.service.StartException in service jboss.infinispan.modeshape-repository-clustered.repo-clustered.config: Failed to start service

    at org.jboss.msc.service.ServiceControllerImpl$StartTask.run(ServiceControllerImpl.java:1904) [jboss-msc-1.2.0.CR1.jar:1.2.0.CR1]

    at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1145) [rt.jar:1.7.0_40]

    at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:615) [rt.jar:1.7.0_40]

    at java.lang.Thread.run(Thread.java:724) [rt.jar:1.7.0_40]

Caused by: org.jboss.msc.inject.InjectionException: Value already set for this injector

    at org.jboss.msc.inject.RetainingInjector.inject(RetainingInjector.java:50) [jboss-msc-1.2.0.CR1.jar:1.2.0.CR1]

    at org.jboss.msc.inject.CastingInjector.inject(CastingInjector.java:55) [jboss-msc-1.2.0.CR1.jar:1.2.0.CR1]

    at org.jboss.msc.service.ServiceControllerImpl.doInject(ServiceControllerImpl.java:1672) [jboss-msc-1.2.0.CR1.jar:1.2.0.CR1]

    at org.jboss.msc.service.ServiceControllerImpl.access$2000(ServiceControllerImpl.java:51) [jboss-msc-1.2.0.CR1.jar:1.2.0.CR1]

    at org.jboss.msc.service.ServiceControllerImpl$StartTask.performInjections(ServiceControllerImpl.java:1917) [jboss-msc-1.2.0.CR1.jar:1.2.0.CR1]

    at org.jboss.msc.service.ServiceControllerImpl$StartTask.run(ServiceControllerImpl.java:1876) [jboss-msc-1.2.0.CR1.jar:1.2.0.CR1]

    ... 3 more

Removing the <transaction> element (defaulting to NONE) works. Is this a bug or am I missing some configuration option(s) ?

Thanks

[rhauch]

Anybody have any suggestions or comments?

[pferraro]

OK - I see the problem.  There is currently a configuration incompatibility between the transaction mode and batching.  To summarize, if you don't plan to use the batching API, don't use batching="true".  If you do plan to use the batching API, don't configure a transaction mode.

Really, batching should configured as a separate transaction mode (e.g. <transaction mode="BATCH"/>, since batches do not coordinate with WF's transaction manager, but use an internal tx mechanism.  Consolidating the 2 configuration options will make things unambiguous.

[hchiorean]

That did the trick, thanks Paul.

[pferraro]

FYI: [WFLY-3227] Replace batching="true|false" with <transaction mode="BATCH"/> in infinispan subsystem schema/model - JBoss …