On 18/11/2007, Roman Kalukiewicz <[EMAIL PROTECTED]> wrote:
> > However, I've found out that the Aggregator might help me out of this. It
> > has a time window. It needn't Spring's support, which makes it more doable
> > for us.
>
> Just be warnined that aggregator and its timeout are not
> message-based. I mean that aggregator timeouts purely on time basis
> (1st, 2nd, 3rd second after context starts - not 1st, 2nd, 3rd second
> after first message arrives).
> That mean that you cannot predict how long this timeout will be for
> given message - it depends on the time this message arrives.

Agreed.

We might wanna support a persistent (JPA) Aggregator as an alternative
to using the batch approach.

The batch approach works great when using JMS transactions where you
want high performance without using a database; though it does require
on aggregation within batches which is not always possible.

So you might wanna process each message one by one, using a database
to keep track of the state (which in general is a better approach as
it will work with any kind of message stream).

There's an outstanding JIRA if anyone fancies trying to write a little
JPA based Aggregator...
http://issues.apache.org/activemq/browse/CAMEL-217

We've already got a JPA based Idempotent Consumer; a JPA based
aggregator is kinda similar; apart from it'll be storing some blob of
data for each message processed etc.

-- 
James
-------
http://macstrac.blogspot.com/

Open Source Integration
http://open.iona.com

Reply via email to