Mailing-List: contact users-help@activemq.apache.org; run by ezmlm
Precedence: bulk
Reply-To: users@activemq.apache.org
Date: Sun, 26 Jul 2015 18:04:59 -0700 (PDT)
From: Cadmean <hzcadmean@hotmail.com>
To: users@activemq.apache.org
Message-ID: <1437959099009-4699892.post@n4.nabble.com>
In-Reply-To: 
 <CAPVVMa_Pq5RqFpCHma64hc1pE5V=bm=gtjHfiPk4Af4rV7Go1w@mail.gmail.com>
References: <1437119658901-4699427.post@n4.nabble.com>
 <1437703899838-4699769.post@n4.nabble.com>
 <CAPVVMa_Pq5RqFpCHma64hc1pE5V=bm=gtjHfiPk4Af4rV7Go1w@mail.gmail.com>
Subject: Re: message stucks in 4 brokers network
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit

I just found by setting "duplex=true" connection for every broker, some
channels between brokers are blocked.

So here is my problem again:

I have brokers A,B,C,D and only 1 consumer connected to A, according to the
network connections, all the messages from B,C and D should be forwarded to
A and get consumed. 

But after running for a while, I restart the consumer and make it connect to
B. Some messages just stay in A and not being forwarded to B. (when I look
into it, I found those stuck messages in A are all from broker B,C and D.
They have been forwarded to A before I restart the consumer). 

In this case, how to adapt the borkers to make all the messages consumed?


Tim Bain wrote
> If all brokers connect to all other brokers using duplex="false", then you
> have a complete graph and messages can pass from any broker.  So I don't
> think duplex has anything to do with your problem.  (Though if you have a
> duplex connection from every broker to every other broker, then there are
> two bi-directional connections between any two nodes, and I don't know how
> that would behave; where I've worked with networks of brokers, I've always
> avoided having more than one identical network connector between any pair
> of nodes.)
> 
> Even if you have connectivity between all nodes, you won't forward
> messages
> if 1) they've exceeded the max TTL (note: I'm talking about network hops,
> NOT JmsExpiration), or 2) they've already been sent to that node and would
> now need to be sent back to it.  #2 got addressed by your
> replayWhenNoConsumer setting.  For #1 (which didn't get discussed here),
> you might want to check that you've set your TTL high enough to support
> the
> maximum number of forwards the message would need in your absolute
> worst-case scenario (clients repeatedly hopping between brokers, etc).
> 
> Tim
> 
> On Thu, Jul 23, 2015 at 8:11 PM, Cadmean &lt;

> hzcadmean@

> &gt; wrote:
> 
>> This problem has been solved by :
>>
>> 1. setting F5 load balancer as :  least connection & session keeping
>>
>> 2. setting networkConnector as : remove localhost, duplex="true",
>> dynamicOnly="true"
>>
>> 3. setting destination policy as : replayWhenNoConsumer="true"
>>
>> I think the main problem of my question is duplex="false", which means
>> the
>> messages forwarded from other brokers cannot replay.
>>
>>
>>
>> --
>> View this message in context:
>> http://activemq.2283324.n4.nabble.com/message-stucks-in-4-brokers-network-tp4699427p4699769.html
>> Sent from the ActiveMQ - User mailing list archive at Nabble.com.
>>


--
View this message in context: http://activemq.2283324.n4.nabble.com/message-stucks-in-4-brokers-network-tp4699427p4699892.html
Sent from the ActiveMQ - User mailing list archive at Nabble.com.