[rabbitmq-discuss] How to debug queue memory leak

Matthias Radestock matthias at rabbitmq.com
Mon Jan 28 23:28:03 GMT 2013


On 28/01/13 22:43, Max Kuznecov wrote:
> Here's again my data: [...]

Cheers.

This confirms what I thought. Your setup is pathological ;) The 
combination of a large number of slow-moving (but never slow enough to 
hibernate) queues with persistent messages causes a substantial amount 
of data to be held in memory for a long period of time. The memory usage 
should be bounded at a few tens of Mb per queue, but that's obviously 
quite a lot when there are many queues.

I will file a bug to investigate what we can do about this.

Meanwhile, one way to reduce the memory usage is to configure rabbit 
with a lower queue index journal size by putting s.t. like the following 
into your rabbitmq.config:

[{rabbit, [{queue_index_max_journal_entries, 65536}]}].

The default value here is 262144, and the above should cut your max 
per-queue memory usage to about 1/4th.

Regards,

Matthias.


More information about the rabbitmq-discuss mailing list