[manik]

FYI I have decided to cut a 3.0.3.GA. I expect 3.0.3.CR1 will be available for download in a few hours and 3.0.3.GA late next week if the CR looks good.

[nathanmesser]

Excellent news! I'm downloading to test now.

[nathanmesser]

I thought you might be interested in feedback.

In my local testing I was unable to reproduce the problem of orphaned locks, and after two days being in our product and tested here we've not had any problems.

[nathanmesser]

We've just had an orphaned lock again, so we may have an additional problem.

[nathanmesser]

This one is less obscure, we had an OutOfMemoryError during rollback.

####<19-Feb-2009 15:50:06 o'clock GMT> <cdlwlprod01> <prod_CDLWLPROD01> <[ACTIVE] ExecuteThread: '19' for queue: 'weblogic.kernel.Default (self-tuning)'> <<WLS Kernel>> <> <> <1235058606291> <BEA-000000> <182721306 [[ACTIVE] ExecuteThread: '19' for queue: 'weblogic.kernel.Default (self-tuning)'] WARN interceptors.TxInterceptor - Rollback had a problem
java.lang.OutOfMemoryError: allocLargeObjectOrArray - Object size: 199376, Num elements: 49840
at org.jboss.cache.util.ImmutableListCopy.(ImmutableListCopy.java:74)
at org.jboss.cache.util.Immutables.immutableListConvert(Immutables.java:66)
at org.jboss.cache.transaction.AbstractTransactionContext.getLocks(AbstractTransactionContext.java:177)
at org.jboss.cache.interceptors.MVCCLockingInterceptor.transactionalCleanup(MVCCLockingInterceptor.java:399)
at org.jboss.cache.interceptors.MVCCLockingInterceptor.handleRollbackCommand(MVCCLockingInterceptor.java:302)
at org.jboss.cache.interceptors.base.PrePostProcessingCommandInterceptor.visitRollbackCommand(PrePostProcessingCommandInterceptor.java:401)
at org.jboss.cache.commands.tx.RollbackCommand.acceptVisitor(RollbackCommand.java:49)
at org.jboss.cache.interceptors.base.CommandInterceptor.invokeNextInterceptor(CommandInterceptor.java:116)
at org.jboss.cache.interceptors.NotificationInterceptor.visitRollbackCommand(NotificationInterceptor.java:67)
at org.jboss.cache.commands.tx.RollbackCommand.acceptVisitor(RollbackCommand.java:49)
at org.jboss.cache.interceptors.base.CommandInterceptor.invokeNextInterceptor(CommandInterceptor.java:116)
at org.jboss.cache.interceptors.TxInterceptor.handleCommitRollback(TxInterceptor.java:459)
at org.jboss.cache.interceptors.TxInterceptor.runRollbackPhase(TxInterceptor.java:529)
at org.jboss.cache.interceptors.TxInterceptor$RemoteSynchronizationHandler.afterCompletion(TxInterceptor.java:816)
at org.jboss.cache.interceptors.TxInterceptor$LocalSynchronizationHandler.afterCompletion(TxInterceptor.java:968)
at org.jboss.cache.interceptors.OrderedSynchronizationHandler.afterCompletion(OrderedSynchronizationHandler.java:92)
at weblogic.transaction.internal.ServerSCInfo$CallAfterCompletionsAction.run(ServerSCInfo.java:1601)
at weblogic.transaction.internal.ServerSCInfo.callAfterCompletions(ServerSCInfo.java:1011)
at weblogic.transaction.internal.ServerTransactionImpl.callAfterCompletions(ServerTransactionImpl.java:2990)
at weblogic.transaction.internal.ServerTransactionImpl.afterRolledBackStateHousekeeping(ServerTransactionImpl.java:2872)
at weblogic.transaction.internal.ServerTransactionImpl.setRolledBack(ServerTransactionImpl.java:2848)
at weblogic.transaction.internal.ServerTransactionImpl.globalRetryRollback(ServerTransactionImpl.java:3087)
at weblogic.transaction.internal.ServerTransactionImpl.globalRollback(ServerTransactionImpl.java:2838)
at weblogic.transaction.internal.TransactionImpl$3.run(TransactionImpl.java:1937)
at weblogic.work.ExecuteThread.run(ExecuteThread.java:173)>

[manik]

"nathanmesser" wrote:

java.lang.OutOfMemoryError: allocLargeObjectOrArray - Object size: 199376, Num elements: 49840
at org.jboss.cache.util.ImmutableListCopy.<init>(ImmutableListCopy.java:74)
at org.jboss.cache.util.Immutables.immutableListConvert(Immutables.java:66)
at org.jboss.cache.transaction.AbstractTransactionContext.getLocks(AbstractTransactionContext.java:177)

That's a large collection you're copying. :-) Is this probable, you have a single transaction that touched almost 50,000 nodes in the cache?!?

[nathanmesser]

It is quite big isn't it?

The largest amount of nodes any transaction should have been able to hit would be around 22000, when we're bouncing the cache, although that would remove the node that was parent to them all, and then re-added them, so I guess that's 22000 removals and 22000 adds.