ClearCommand participates on global transaction and produces a lock! Is this wanted behavoiur or a bug?

[pb00067]

Hi,

I use Infinispan as 2L-cache implementation for Hibernate.

I assumed that general evict actions are handled as atomic actions even if being in transaction (transactionmanager.getStatus == Status.STATUS_ACTIVE).

Indeed such actions (for instance sessionfactory.getCache.evictEntityRegion(Apersistent.class)) do participate on the global transaction,

and produce a lock which is hold until the transaction ends.

userTransaction.begin();

sessionfactory.getCache.evictEntityRegion(A.class); // action participates on global transaction and produces a lock

TransactionManager().suspend(); // suspend current global transaction

sessionfactory.getCache.evictEntityRegion(A.class); // this action fails as it goes in timeout after 15 seconds

org.infinispan.util.concurrent.TimeoutException: Unable to acquire lock after [15 seconds] on key [hello.A#1] for requestor [Thread[main,5,main]]! Lock held by [GlobalTransaction:<localhost-8797>:3]
    at org.infinispan.factories.EntryFactoryImpl.acquireLock(EntryFactoryImpl.java:198)
    at org.infinispan.factories.EntryFactoryImpl.wrapEntryForWriting(EntryFactoryImpl.java:143)
    at org.infinispan.interceptors.LockingInterceptor.visitClearCommand(LockingInterceptor.java:167)
    at org.infinispan.commands.write.ClearCommand.acceptVisitor(ClearCommand.java:55)
    at org.infinispan.interceptors.base.CommandInterceptor.invokeNextInterceptor(CommandInterceptor.java:118)
    at org.infinispan.interceptors.InvalidationInterceptor.visitClearCommand(InvalidationInterceptor.java:108)
    at org.infinispan.commands.write.ClearCommand.acceptVisitor(ClearCommand.java:55)
    at org.infinispan.interceptors.base.CommandInterceptor.invokeNextInterceptor(CommandInterceptor.java:118)
    at org.infinispan.interceptors.base.CommandInterceptor.handleDefault(CommandInterceptor.java:132)
    at org.infinispan.commands.AbstractVisitor.visitClearCommand(AbstractVisitor.java:69)
    at org.infinispan.commands.write.ClearCommand.acceptVisitor(ClearCommand.java:55)
    at org.infinispan.interceptors.base.CommandInterceptor.invokeNextInterceptor(CommandInterceptor.java:118)
    at org.infinispan.interceptors.TxInterceptor.enlistWriteAndInvokeNext(TxInterceptor.java:179)
    at org.infinispan.interceptors.TxInterceptor.visitClearCommand(TxInterceptor.java:141)
    at org.infinispan.commands.write.ClearCommand.acceptVisitor(ClearCommand.java:55)
    at org.infinispan.interceptors.base.CommandInterceptor.invokeNextInterceptor(CommandInterceptor.java:118)
    at org.infinispan.interceptors.InvocationContextInterceptor.handleAll(InvocationContextInterceptor.java:48)
    at org.infinispan.interceptors.InvocationContextInterceptor.handleDefault(InvocationContextInterceptor.java:34)
    at org.infinispan.commands.AbstractVisitor.visitClearCommand(AbstractVisitor.java:69)
    at org.infinispan.commands.write.ClearCommand.acceptVisitor(ClearCommand.java:55)
    at org.infinispan.interceptors.InterceptorChain.invoke(InterceptorChain.java:256)
    at org.infinispan.CacheDelegate.clear(CacheDelegate.java:369)
    at org.infinispan.CacheDelegate.clear(CacheDelegate.java:188)
    at org.hibernate.cache.infinispan.access.TransactionalAccessDelegate.evictOrRemoveAll(TransactionalAccessDelegate.java:110)
    at org.hibernate.cache.infinispan.access.TransactionalAccessDelegate.evictAll(TransactionalAccessDelegate.java:106)
    at org.hibernate.cache.infinispan.entity.TransactionalAccess.evictAll(TransactionalAccess.java:32)
    at org.hibernate.impl.SessionFactoryImpl$CacheImpl.evictEntityRegion(SessionFactoryImpl.java:1004)
    at org.hibernate.impl.SessionFactoryImpl$CacheImpl.evictEntityRegion(SessionFactoryImpl.java:995)

Please consider attached testcase.. (run: java JTAStandaloneExampleJBossTM)

[galder.zamarreno]

Guenther, this code has now changed and it's handled in a different way. It's now done as a local only clear and with 0 as lock timeout. This code has not yet been released but I can guide you on how to build it if you're interested in testing it. I don't know when next Hibernate 3.5 beta/cr will be released.

[galder.zamarreno]

In fact, it appears that Hibernate 3.5.beta3 will come out this week and that will fix your current issue. Keep an eye for release posts in http://in.relation.to

[galder.zamarreno]

FYI: Hibernate 3.5.0.beta3 has now been released and should have your issue fixed.

[pb00067]

Yes, I can confirm this.

Now in my testsuite, the org.infinispan.util.concurrent.TimeoutException is not appearing anymore

[galder.zamarreno]

Glad to hear that!