[rabbitmq-discuss] Problem with a constantly increasing memory footprint
Gavin M. Roy
gmr at myyearbook.com
Wed Mar 9 14:55:36 GMT 2011
I've noticed this behavior as well. In my case, I was over saturating a single node in a cluster. I solved my problem by expanding my cluster size and balancing my publishers across the cluster. You might want to check the rabbitmq.log/rabbmq-sasl.log file and see if you notice any weird behavior in there, such as abnormal connection terminations.
Hope this helps,
On Wednesday, March 9, 2011 at 8:20 AM, Daniel Maher wrote:
> We are running a two node RabbitMQ cluster, and on _one_ of those nodes
> the memory footprint used by beam.smp grows steadily and constantly over
> time, regardless of the usage pattern. I've been pulling my hair out
> trying to figure out why, and while it's helped me to learn quite a bit
> about administrating RabbitMQ (which is a good thing), it hasn't
> resulted in a solution for the memory creep.
> The particulars:
> Debian Squeeze 64bit
> Rabbit 2.3.1 / Erlang R14A
> Mnesia storage to disc
> If I restart RabbitMQ on the problem node, the footprint will drop
> dramatically (logical), but will then continue to rise over time once
> I've been using Collectd to graph the footprint as reported by
> rabbitmqctl as well as my pmap directly. I don't know if this list
> supports attachments, so i've put the two graphs here:
> They are both graphs of a week in scope. The first graph is the memory
> usage as reported by rabbitmqctl in bytes, and the second is that
> reported by pmap in kilobytes.
> Any insight - or perhaps ideas on how to go about debugging the issue -
> would be greatly appreciated. Thank you, all.
> Daniel Maher
> « can't talk, too busy calculating computrons. »
> rabbitmq-discuss mailing list
> rabbitmq-discuss at lists.rabbitmq.com
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the rabbitmq-discuss