[rabbitmq-discuss] Debugging RabbitMQ memory usage

[Simon MacMullen]

First of all, could you be seeing the same as this:

http://rabbitmq.1065348.n5.nabble.com/memory-leak-rabbitmq-3-1-5-td29924.html 
?

The smoking gun is

   $ rabbitmqctl eval \
     'length(supervisor:which_children(rabbit_stomp_client_sup_sup)).'

returning a number much larger than the number of open STOMP connections 
on the leaky node.

Cheers, Simon

On 01/10/13 16:55, Mariusz Gronczewski wrote:
> Hi,
>
> I have 2 clusers, 2 nodes each with same configuration. One node is
> serving most of the traffic (about 100 req/sec via STOMP/SSL, basically
> zero messages staying in queues) while other is "backup". Plugins are
> management, management_federation, federation and stomp)
>
> I have noticed that on clusterA, master node eats above 2GB RAM (after 2
> days of uptime, slowly growing), mostly in "Plugins" and "Binaries", and
> in clusterB both nodes are below 200MB RAM usage. Only difference is
> that on clusterA stats are gathered on same node serving traffic while
> on clusterB stats are on "secondary" node.
>
> Software configuration is basically identical (I used puppet to deploy
> both at same time), and clusterA is on slower machines, but by no means
> overloaded (using maybe 5% CPU).
>
> version: RabbitMQ 3.1.5.40927, Erlang R15B01, but same thing occurs on
> current stable 3.1.5
>
> Is there any way to debug what uses it and why it is leaking memory ?
>
> Cheers
> Mariusz
>
>
> _______________________________________________
> rabbitmq-discuss mailing list
> rabbitmq-discuss at lists.rabbitmq.com
> https://lists.rabbitmq.com/cgi-bin/mailman/listinfo/rabbitmq-discuss
>


-- 
Simon MacMullen
RabbitMQ, Pivotal