[
Date Prev][
Date Next][
Thread Prev][
Thread Next][
Date Index][
Thread Index]
[
List Home]
Re: [smila-dev] new router/listener
|
Hi Juergen,
Juergen.Schumacher@xxxxxxxxxxx wrote:
Hi,
I agree with Sebastian that a listener should only read those message
it really wants to process and not all
messages, so we need to use message selectors. It may be "our" queue
with only "our" message, but still
we could have specialized workflow processors that provide only a
selection of all workflows, so they should
not read messages that they cannot process because they do not contain
the needed workflow.
Agree.
Why do you say that the syntax for message selectors is not defined in
JMS? There is a syntax specification
in [1]. Or am I confusing something and you mean something different?
So, I was wrong, many thanks.
Idea - we may try to extract and use message selector SQL evaluator
implementation from ActiveMQ to use in Router instead of beanshell.
Also, Derby is surely not an appropriate database when it comes to
really high volume scenarios, so we should
do tests with "real" database servers, too. BTW, why do we use Derby
at all? The default message store in
ActiveMQ should be a file based storage [2], which should provide
better performance (at least they claim it).
Yes you are right its file based storage used here
I saw this performance problem and mix up ActiveMQ with other framework
- it was not a my day :(.
Cheers,
Juergen.
[1] http://java.sun.com/j2ee/sdk_1.3/techdocs/api/javax/jms/Message.html
[2] http://activemq.apache.org/amq-message-store.html
BTW: new router/listener already integrated and working, old version is
removed.
--
Regards, Ivan
------------------------------------------------------------------------
*From:* smila-dev-bounces@xxxxxxxxxxx
[mailto:smila-dev-bounces@xxxxxxxxxxx] *On Behalf Of *Sebastian Voigt
*Sent:* Thursday, July 31, 2008 5:16 PM
*To:* Smila project developer mailing list
*Subject:* RE: [smila-dev] new router/listener
We should use the Queue for Remote Communication and therefore we
should use Queries to get Entries from the Queue.
(instead to have a central listener that distribute the entries to
the workflow processes (distributed systems!)).
Please use for the first implementation configurable timeout values
(it should be no problem to use higher values than 1 second, or?)
Mit freundlichen Grüßen / Kind regards
Sebastian Voigt
*brox *IT-Solutions GmbH*
*
==================================
According to Section 80 of the German Corporation Act brox
IT-Solutions GmbH must indicate the following information.
Address: An der Breiten Wiese 9, 30625 Hannover Germany
General Manager: Hans-Chr. Brockmann
Registered Office: Hannover, Commercial Register Hannover HRB 59240
========== Legal Disclaimer ==========
*From:* smila-dev-bounces@xxxxxxxxxxx
[mailto:smila-dev-bounces@xxxxxxxxxxx] *On Behalf Of *Ivan Churkin
*Sent:* Wednesday, July 30, 2008 4:12 PM
*To:* smila-dev@xxxxxxxxxxx
*Subject:* [smila-dev] new router/listener
Hi Sebastian
Router/Listener based on new principles is working.
/(Its possible to manage working threads by JMX, using multiples
brokers, switching MQ implementations and specifying task
execution list in configuration.)
(Its not included into main build and not used by connectivity etc.)
/
For testing include into workspace all new bundles from trunc
/org.beanshell
org.eclipse.eilf.jms
org.eclipse.eilf.jms.activemq
org.eclipse.eilf.connectivity.queue.worker
org.eclipse.eilf.connectivity.queue.worker.test
/
Test it by running launch configuration
/org.eclipse.eilf.connectivity.queue.worker.test/WorkerTest.launch/
It was found a problem with ActiveMQ
/I pushed into queue just only 2500 messages before listening
and after that consumer was unable to receive messages.
I guess that it was because now consumer collect messages
synchronously in working thread
message = _consumer.receive(TIMEOUT);
(TIMEOUT was equals to 1 sec)
/*
I think that ActiveMQ is unable to find message in durable storage
by Condition in this timeout and always return NULL!*
Solutions:
/1) I did TIMEOUT configurable by Rule XML tag, increase it and to
forget.
2) We may try to use more fast database when Derby as ActiveMQ
durable message storage.
/But all that solutions are only partial.
There is totally different solution:
/To get message from consumer without condition and to check
Condition inside Listener.
That also helps to solve the second problem: "Condition syntax
does not specified in JMS and depends on broker implementation"
In this solution condition syntax in Rules will be independent
from broker implementation.
/
But, of course, there is also a disadvantage:
/Listener will get every message from queue and, if rule will not
be found, message will be send into DLQ, finally.
But I'm not sure that its a problem because its "our queue" and it
should contains only "our messages".
/
What do you think about it?
Folks, feedbacks are appreciated :).
--
Regards, Ivan
------------------------------------------------------------------------
_______________________________________________
smila-dev mailing list
smila-dev@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/smila-dev