[rabbitmq-discuss] RabbitMQ and Stomp topic behavior

Massimo Paladin Massimo.Paladin at cern.ch
Mon Aug 8 16:19:11 BST 2011

I tried in two machines, in one I had 0.5, in the other I did not set it
which means the default is 0.4.

They both crashed and I get many memory alarms and clears in the log.

The memory goes up and down through the test, and soon or later it fills
both memory and swap until it gets to crash.

Massimo Paladin

email: massimo.paladin at gmail.com
website: http://www.mpaladin.com
flickr's page: http://flickr.com/photos/massimop

On Mon, Aug 8, 2011 at 5:06 PM, Matthias Radestock <matthias at rabbitmq.com>wrote:

> Massimo,
> On 07/08/11 20:52, Massimo Paladin wrote:
>> Why does the memory goes up if there are no consumers attached to the
>> topic?
> Messages are processed in a fully asynchronous pipeline, and the decision
> on whether to discard a message is made at a later stage.
> Having this pipeline allows rabbit to absorb spikes in message load, i.e.
> it can read messages off the socket quickly and pass them to the next stage
> of the pipeline.
> However, if the load is not a spike and instead is consistently high then
> obviously rabbit will run out of memory eventually. To prevent that, memory
> usage is monitored and producers are temporarily blocked to relieve memory
> pressure. So...
>  The scenario provided makes memory increasing and if you let it run for
>> some time at some point rabbitmq crashes because it can not allocate
>> memory.
> That shouldn't happen. Do you get any memory alarms in the rabbit logs?
> Also, does Rabbit run out of memory with the default threshold of 0.4?
> Regards,
> Matthias.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.rabbitmq.com/pipermail/rabbitmq-discuss/attachments/20110808/7bb9a00a/attachment.htm>

More information about the rabbitmq-discuss mailing list