Network of brokers and starvation problem

[Mathias Herberts <ma...@gmail.com>]

Hi,

I have set up a network of two brokers (which connect bidirectionnaly
to each other after discovering themselves via multicast).

Broker A talks to broker B and B talks to A.

Consumer C1 is connected to B, subscribing to a queue Q.
Producer P1 connects to A and posts a message M to a queue Q.
C1 receives M, then dies without committing/acking M.
Consumer C2 connects to B and subscribes to Q.
C2 will NEVER receive M thus leading to a starvation situation.

Shouldn't the connection of a consumer to a broker part of a network
of brokers trigger an event sent to all brokers part of the network
and forcing brokers with messages for a given destination and no
consumer for that destination to forward the messages back to a broker
with consumers?

Mathiias.

Re: Network of brokers and starvation problem

[Rob Davies <ra...@gmail.com>]

The TTL parameter can be infinite - and it can be set on a network  
basis - e.g. every broker to broker network connection could have a  
different value. Its the connector where the message/subscription is  
generated that controls the TTL.
On 3 Apr 2009, at 16:46, L.Daigremont wrote:

>
> Is it possible to toggle dynamically that "TTL" parameter ? provided  
> our
> network of broker can have a variable depth.
> Is there an "infinite value" for that setting ?
>
>
> Hiram Chirino wrote:
>>
>> There is a TTL setting on the network connector the limits the number
>> of broker hops a messages is allowed to take.  By default it is set  
>> to
>> 1.  This prevents a message from ever looping around nodes in a
>> network.
>>
>> You could change the setting to be higher, but be advised that you
>> risk having messages loop.
>>
>> On 5/22/06, Mathias Herberts <ma...@gmail.com> wrote:
>>> Hi,
>>>
>>> I have set up a network of two brokers (which connect  
>>> bidirectionnaly
>>> to each other after discovering themselves via multicast).
>>>
>>> Broker A talks to broker B and B talks to A.
>>>
>>> Consumer C1 is connected to B, subscribing to a queue Q.
>>> Producer P1 connects to A and posts a message M to a queue Q.
>>> C1 receives M, then dies without committing/acking M.
>>> Consumer C2 connects to B and subscribes to Q.
>>> C2 will NEVER receive M thus leading to a starvation situation.
>>>
>>> Shouldn't the connection of a consumer to a broker part of a network
>>> of brokers trigger an event sent to all brokers part of the network
>>> and forcing brokers with messages for a given destination and no
>>> consumer for that destination to forward the messages back to a  
>>> broker
>>> with consumers?
>>>
>>> Mathiias.
>>>
>>
>>
>> -- 
>> Regards,
>> Hiram
>>
>>
>
> -- 
> View this message in context: http://www.nabble.com/Network-of-brokers-and-starvation-problem-tp4501784p22871125.html
> Sent from the ActiveMQ - Dev mailing list archive at Nabble.com.
>

Re: Network of brokers and starvation problem

["L.Daigremont" <ld...@axway.com>]

Is it possible to toggle dynamically that "TTL" parameter ? provided our
network of broker can have a variable depth.
Is there an "infinite value" for that setting ?


Hiram Chirino wrote:
> 
> There is a TTL setting on the network connector the limits the number
> of broker hops a messages is allowed to take.  By default it is set to
> 1.  This prevents a message from ever looping around nodes in a
> network.
> 
> You could change the setting to be higher, but be advised that you
> risk having messages loop.
> 
> On 5/22/06, Mathias Herberts <ma...@gmail.com> wrote:
>> Hi,
>>
>> I have set up a network of two brokers (which connect bidirectionnaly
>> to each other after discovering themselves via multicast).
>>
>> Broker A talks to broker B and B talks to A.
>>
>> Consumer C1 is connected to B, subscribing to a queue Q.
>> Producer P1 connects to A and posts a message M to a queue Q.
>> C1 receives M, then dies without committing/acking M.
>> Consumer C2 connects to B and subscribes to Q.
>> C2 will NEVER receive M thus leading to a starvation situation.
>>
>> Shouldn't the connection of a consumer to a broker part of a network
>> of brokers trigger an event sent to all brokers part of the network
>> and forcing brokers with messages for a given destination and no
>> consumer for that destination to forward the messages back to a broker
>> with consumers?
>>
>> Mathiias.
>>
> 
> 
> -- 
> Regards,
> Hiram
> 
> 

-- 
View this message in context: http://www.nabble.com/Network-of-brokers-and-starvation-problem-tp4501784p22871125.html
Sent from the ActiveMQ - Dev mailing list archive at Nabble.com.

Re: Network of brokers and starvation problem

[Hiram Chirino <hi...@hiramchirino.com>]

There is a TTL setting on the network connector the limits the number
of broker hops a messages is allowed to take.  By default it is set to
1.  This prevents a message from ever looping around nodes in a
network.

You could change the setting to be higher, but be advised that you
risk having messages loop.

On 5/22/06, Mathias Herberts <ma...@gmail.com> wrote:
> Hi,
>
> I have set up a network of two brokers (which connect bidirectionnaly
> to each other after discovering themselves via multicast).
>
> Broker A talks to broker B and B talks to A.
>
> Consumer C1 is connected to B, subscribing to a queue Q.
> Producer P1 connects to A and posts a message M to a queue Q.
> C1 receives M, then dies without committing/acking M.
> Consumer C2 connects to B and subscribes to Q.
> C2 will NEVER receive M thus leading to a starvation situation.
>
> Shouldn't the connection of a consumer to a broker part of a network
> of brokers trigger an event sent to all brokers part of the network
> and forcing brokers with messages for a given destination and no
> consumer for that destination to forward the messages back to a broker
> with consumers?
>
> Mathiias.
>


-- 
Regards,
Hiram