You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@qpid.apache.org by Alexandre Trufanow <al...@gmail.com> on 2016/04/25 18:25:57 UTC

Configuring multiple dispatchers

Hi,

I am setting up a test environment with dispatchers and I have a couple of
questions regarding the configuration. I am using dispatch-0.6.0-beta2.

The current architecture is two dispatchers in front of multiple brokers,
and messages can be sent or received from either dispatcher
indiscriminately.

The documentation mentions this use case but I haven't found an sample
configuration. In order to make this work, I had to do some guesswork based
on how to configure a single dispatcher, can someone confirm that I am
doing this right?

* Each dispatcher must be given a unique containerName (in type=container)
and routerId (in type=router). This is to avoid clashes on the brokers.
* There needs to be two separate listeners, one for communication with
brokers (on-demand) and one for communication with the dispatcher
(inter-router).
* On each dispatcher, a connector is created towards the other dispatcher
as well as an autoLink for each address in the way as for connecting to a
broker.

* When messages are sent to a dispatcher, I would expect to be able to
retrieve them directly from the brokers. However, It is necessary to read
the messages from a dispatcher most of the time, as they are no longer
present on the brokers. Is this expected behavior?

Thanks for your help,

Alexandre
www.murex.com

Re: Configuring multiple dispatchers

Posted by Alexandre Trufanow <al...@gmail.com>.
Thanks a lot for you answers Ted. I'm looking through the configurations in
the tests.



>> * When messages are sent to a dispatcher, I would expect to be able to
>> retrieve them directly from the brokers. However, It is necessary to read
>> the messages from a dispatcher most of the time, as they are no longer
>> present on the brokers. Is this expected behavior?
>>
>
> I'm not sure I understand your question.  If you want to use
> dispatch-router to send messages to a broker but you want to retrieve the
> messages directly from the broker, don't configure an autoLink to retrieve
> messages from the queue in question.  This will leave the messages on the
> broker.
>
>
I was asking if it was possible to retrieve messages from the brokers if a
dispatcher was configured for subscriptions on the queue. Things are
clearer now.

Re: Configuring multiple dispatchers

Posted by Ted Ross <tr...@redhat.com>.

On 04/25/2016 12:25 PM, Alexandre Trufanow wrote:
> Hi,
>
> I am setting up a test environment with dispatchers and I have a couple of
> questions regarding the configuration. I am using dispatch-0.6.0-beta2.
>
> The current architecture is two dispatchers in front of multiple brokers,
> and messages can be sent or received from either dispatcher
> indiscriminately.
>
> The documentation mentions this use case but I haven't found an sample
> configuration. In order to make this work, I had to do some guesswork based
> on how to configure a single dispatcher, can someone confirm that I am
> doing this right?
>
> * Each dispatcher must be given a unique containerName (in type=container)
> and routerId (in type=router). This is to avoid clashes on the brokers.

Each router must have a unique routerId in order for the 
route-computation function to work correctly.  We are actually planning 
to remove the "container" type and default the containerName to routerId 
to simplify the configuration.

> * There needs to be two separate listeners, one for communication with
> brokers (on-demand) and one for communication with the dispatcher
> (inter-router).

When building a network of routers, the connections between routers 
(listener and connector) _must_ have the role of inter-router.  The role 
for connectors/listeners for brokers should be route-container 
(on-demand is still supported but is deprecated).

> * On each dispatcher, a connector is created towards the other dispatcher
> as well as an autoLink for each address in the way as for connecting to a
> broker.

The create a network of routers, each connection between routers 
involves one connector and one listener.  There is no need to have 
connectors pointing in both directions.

If you look in the tests directory in the source, you will see several 
directories containing configuration files.  These are named things like 
config-2, config-3-linear, config-6, etc.  These are examples of 
multi-router configurations.

>
> * When messages are sent to a dispatcher, I would expect to be able to
> retrieve them directly from the brokers. However, It is necessary to read
> the messages from a dispatcher most of the time, as they are no longer
> present on the brokers. Is this expected behavior?

I'm not sure I understand your question.  If you want to use 
dispatch-router to send messages to a broker but you want to retrieve 
the messages directly from the broker, don't configure an autoLink to 
retrieve messages from the queue in question.  This will leave the 
messages on the broker.

>
> Thanks for your help,
>
> Alexandre
> www.murex.com
>

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@qpid.apache.org
For additional commands, e-mail: users-help@qpid.apache.org