Maybe I hasn't illustrated my usage clearly.
My usage is two separate nested global transaction(jta), not just a single
transaction.
If only a single jta transaction, exception process and transaction could
work well together. But if new nested jta transaction declared(using
require-new), exception process will conflict with transaction ;
For example:
<from jms> //here jms start a jta
<transacted/> //default required policy will join the jms started
jta transaction
<dotry>
<policy ref="transaction-require-new>//db1,db2 require
consistency
<db1 save>
<db2 save>
</policy>
<doCatch> //will never be executed
..logging..
</doCatch>
</doTry>
And also you have said, moving policy before doTry. But if so,the process
in catch clause will join the rollback-ed transaction
Although in the simple case, rollback of catch clause has no effect, but
anyway maybe sometimes we need do some transactional operation in catch
clause.
----------------------------------------
Claus Ibsen [claus.ib...@gmail.com] wrote:
>Hi
>I have explain to you before that such a scenario is described in
>chapter 9 in the Camel in Action book.
>The source code is at a public and free site.
>Also this use case is explained on the Camel wiki pages, which does
>only rollback on B and let A commit
>http://camel.apache.org/transactional-client.html
>Also camel-spring has a shit load of unit tests for transactions and
>there would also be one for this use case.
On Fri, Apr 23, 2010 at 1:02 PM, ext2 <x...@tongtech.com> wrote:
>
> Hi:
>
> Sorry to raise the question again.
>
> Recently , I check the EJB specification and do some test with ejb's
> container managed transaction-policy , also the spring-transaction
> template(they both are same)
>
> After checking and testing EJB, I think the camel's current transaction is
> deficiency;
>
> For example, EJB container transaction support such usage:
> Assuming EJB A invoke B. A, B used separate transaction(both is
require-new
> policy);
> 1)If B failed then the B's transaction will rollback.
> 2) The caller A could choice to omit the B's exception and commit
> transaction, or not to omit B's exception and rollback transaction;
>
> But camel's transaction doesn't support such usage. This deficiency is
> caused by the camel's transaction-policy implementation mixed with
> error-handler.
>
> So if the camel's transaction implementation will not to use
error-handler,
> the deficiency will be correct.
> If this worth to do in camel?
>
> Or there are some other reasons that camel 's transaction must use
> error-handler?
>
> Thanks any suggestion;
>
>
>
--
Claus Ibsen
Apache Camel Committer
Author of Camel in Action: http://www.manning.com/ibsen/
Open Source Integration: http://fusesource.com
Blog: http://davsclaus.blogspot.com/
Twitter: http://twitter.com/davsclaus
