[rabbitmq-discuss] Federation chokes on load

[Patrik Ring]

Yes. With only that binding, things are working and I can see the message
in test-receive-users queue. The problems start when I start adding
bindings and messages. We have had it working with a few bindings (5-10
approx) but when we start putting load on it (5000 bindings approx) it
seems to just die.

Best regards
Patrik Ring

Co-founder and CTO of Bloglovin.com
patrik.ring at bloglovin.com
+46 730 59 78 14
+1 347 859 7999


On Wed, Mar 5, 2014 at 7:05 AM, Simon MacMullen <simon at rabbitmq.com> wrote:

> Hi Patrik, thanks.
>
> So this looks like federation is configured correctly and should be
> working normally. The downstream exchange only has a single binding to a
> single queue, with routing key "user.2". Are messages being published
> upstream with that key?
>
> Cheers, Simon
>
>
> On 04/03/14 22:06, Patrik Ring wrote:
>
>> Rabbit version in use is: RabbitMQ 3.2.1, Erlang R14B04
>>
>> rabbitmqctl report on
>> upstream: http://pastebin.com/PNqTziGk
>> "slave" server: http://pastebin.com/4L6JFsiW
>>
>> *I have this in the exchanges list on the upstream:*
>>
>>
>> federation: users -> rabbit at rabbit3.c.rabbit-magic.internal B
>> <http://rabbit1:15672/#/exchanges/%2F/federation%3A%
>> 20users%20-%3E%20rabbit%40rabbit3.c.rabbit-magic.internal%20B>
>>
>>         x-federation-upstream   ha-all  D AD I Args
>>
>>
>> And this i the queues list:
>>
>> federation: users -> rabbit at rabbit3.c.rabbit-magic.internal
>> <http://rabbit1:15672/#/queues/%2F/federation%3A%20users%20-%3E%20rabbit%
>> 40rabbit3.c.rabbit-magic.internal>
>>
>>         rabbit at util1 +1                 Exp D Args
>>
>>
>> *For the slave:*
>> *
>> *
>> *Nothing.* Except on channels, where I find this:
>>
>> <rabbit at rabbit3.3.237.0> (1)
>> <http://162.222.179.163:15672/#/channels/%3Crabbit%
>> 40rabbit3.3.237.0%3E%20(1)>
>>
>> federation
>>
>>
>> Feel like we might be on to something?
>>
>> How do I set this up?
>>
>>
>>
>> As a side note: The queues and exchanges on the upstream might still be
>> there since when we tried making it run both ways, not sure about this.
>> We do however have the federation status on the SLAVE, and according to
>> that it's up and running.
>>
>>
>> Best regards
>> Patrik Ring
>>
>> Co-founder and CTO of Bloglovin.com
>> patrik.ring at bloglovin.com <mailto:patrik.ring at bloglovin.com>
>>
>> +46 730 59 78 14
>> +1 347 859 7999
>>
>>
>> On Tue, Mar 4, 2014 at 12:56 PM, Simon MacMullen <simon at rabbitmq.com
>> <mailto:simon at rabbitmq.com>> wrote:
>>
>>     Hi Patrik.
>>
>>     On 03/03/14 21:41, Patrik Ring wrote:
>>
>>         Some messages come through (but are heavily delayed, as in
>>         hours) but
>>         they seem completely random.
>>
>>
>>     That's definitely not right!
>>
>>     First of all, could you send the output of "rabbitmqctl report" for
>>     each broker?
>>
>>     Also, which version of RabbitMQ are you using?
>>
>>     Secondly, you might want to investigate whether the underlying
>>     infrastructure used by federation is wired up correctly. There
>>     should be a queue for each (exchange, downstream) pairing declared
>>     in the upstream broker (called "federation: [exchange] -> [node]").
>>     It should be bound (once) to an internal exchange with a similar
>>     name. That internal exchange should be bound to the upstream
>>     exchange with all the bindings that are in use downstream.
>>
>>     Does it look like that is happening? Is the infrastructure queue
>>     seeing the message rates you would expect?
>>
>>     Cheers, Simon
>>
>>     --
>>     Simon MacMullen
>>     RabbitMQ, Pivotal
>>
>>
>>
>
> --
> Simon MacMullen
> RabbitMQ, Pivotal
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.rabbitmq.com/pipermail/rabbitmq-discuss/attachments/20140305/0a895fbc/attachment.html>