[rabbitmq-discuss] last_blocked_by resource

[Sean Allen]

On Sat, Mar 15, 2014 at 5:35 PM, Matthias Radestock
<matthias at rabbitmq.com>wrote:

> On 15/03/14 16:21, Sean Allen wrote:
>
>> Question about connections getting blocked for resource levels.
>>
>> Simplifying what we have.
>>
>> I have an exchange "event"
>>
>> Queue: handle.event is bound the "event" on node 1.
>>
>> Node 2 of the cluster is over memory water mark.
>>
>> I have a connection that is connected to Node 3 of the cluster that is
>> publishing to the "event" exchange. How is it possible for that
>> connection to end up being blocked for resource reasons?
>>
>
> RabbitMQ cannot know where future messages published on a connection might
> get routed to. So, in your example, the connection to Node 3 could be used
> to publish messages which end up getting enqueued in queues on Node 2.
> Which would be bad.
>
> The upshot is that a resource alarm on any node blocks publishing
> connections on all nodes.



Does that apply when an exchange is federated?

If I have cluster A and cluster B

and messages are published to an exchange in cluster A and is in turn
federated to cluster B

does an alarm in cluster A prevent cluster B from getting said message?
does an alarm in cluster B prevent cluster A from getting said message?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.rabbitmq.com/pipermail/rabbitmq-discuss/attachments/20140317/5ce7dacf/attachment.html>