Atomikos Forum

Hi,

Sorry if this is a little off topic - I'm just trying to get my head around how this works.

I have a simple architecture with 2 processes on different hosts. Each process has its own JBoss cache instance. The pair of caches form a cluster with sync replication between them. Each process uses Atomikos TransactionEssentials (UserTransactionManager) to provide a JTA transaction manager to the cache. The logs and some simple testing appear to show this working OK.

What I don't get is how this actually works under the hood, so I'm a little reluctant to assume it is correctly configured!

My concern is that I understand Atomikos is an in-process JTA manager. So how can it co-ordinate replication with the cache instance in the other processes? If I had a central JTA server and pointed both cache processes at it that would seem OK. But as each process has its own JTA manager I'm a little confused. I was under the impression the cache replication uses JTA to ensure the participants in the replication (ie, cache instances) are consistent. Is this instead something that is handled separately by the cache layer and not actually using JTA?

Can anyone shed any light on how this is working (if it is indeed a correct configuration!).

Thanks,

Ben

Hi Ben,

TransactionsEssentials is intra-VM, ExtremeTransactions is also inter-VM.

However, even TransactionsEssentials becomes 'inter-VM' if the VMs are tied-in as XAResource instances via JDBC or JMS calls.

That is the bottom line. If you want we can look at your specific details within the scope of a developer ticket (available through our main site).

Best