[infinispan-dev] Readonly tx leak fix causing issues with Hibernate 2LC use case

[Mircea Markus]

On 21 Jan 2011, at 16:34, Galder Zamarreño wrote:

> Hi,
> 
> Re: https://issues.jboss.org/browse/ISPN-845
> 
> This is causing issues with use cases such as the Hibernate 2LC. Imagine this scenario:
> 
> tx.begin();
> cache1.get();
> cache2.put();
> cache3.remove();
> tx.commit();
> 
> At commit time, 3 prepares are generated of which, the 1st prepare is marked as readonly (no mods made on cache1), so that prepare is committed in advance.
> 
> Now, when tx.commit() is called it fails cos the first tx for cache1 has already been committed.
I guess this is because the XAResource doesn't find the tx in the transaction table. I guess the way to go is to delist it after committing, let me ping Jonathan to see :)
         if (trace) log.trace("no tx found for {0}", xid);
         throw new XAException(XAException.XAER_NOTA);

> 
> There's a test that simulates this kind of scenario in Infinispan 2LC testsuite.
> 
> A different solution is needed and seems like this was discussed this week in Newcastle? (https://issues.jboss.org/browse/ISPN-887)
> 
> Cheers,
> --
> Galder Zamarreño
> Sr. Software Engineer
> Infinispan, JBoss Cache
> 
> 
> _______________________________________________
> infinispan-dev mailing list
> infinispan-dev at lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/infinispan-dev