[rabbitmq-discuss] Producer eating up RAM

[Suhail Doshi]

I have actually traced to the error_logger process:
State: Waiting
Name: error_logger
Spawned as: proc_lib:init_p/5
Spawned by: <0.1.0>
Started: Mon Oct  5 20:22:42 2009
Message queue length: 113
Number of heap fragments: 0
Heap fragment data: 0
Link list: [<0.25.0>, <0.0.0>, {to,<0.25.0>,#Ref<0.0.0.254510>}]
Reductions: 615399844
Stack+heap: 356352655
OldHeap: 285082125
Heap unused: 50293735
OldHeap unused: 285082125
Program counter: 0x00007f49a1b7f108 (io:wait_io_mon_reply/2 + 56)
CP: 0x0000000000000000 (invalid)
arity = 0
=proc:<0.5.0>


Anyone else seen this problem? Seems like an erlang thing not a rabbitmq
thing.

On Mon, Oct 5, 2009 at 4:03 AM, Ben Hood <0x6e6562 at gmail.com> wrote:

> Suhail,
>
> On Mon, Oct 5, 2009 at 2:08 AM, Suhail Doshi <suhail at mixpanel.com> wrote:
> > Here's part of the queue code we use:
>
> A few observations:
>
> - Tickets and realms are toast in Rabbit now, so you can nuke that code;
> - You set a lot of fields that are identical to the default values,
> hence adding unnecessary extra code;
> - You are also using a fairly old version of the Erlang client - you
> are less likely to get any serious feedback when you are not using the
> latest version;
>
> In general the idea is to write the most concise piece of code that
> you can that replicates the issue, even if it does take 5 hours to
> reproduce - we can run the code for 5 hours locally if we need to as
> well. Otherwise we have little hope of being able to help you.
>
> Ben
>



-- 
http://mixpanel.com
Blog: http://blog.mixpanel.com
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.rabbitmq.com/pipermail/rabbitmq-discuss/attachments/20091005/fbc318a3/attachment.htm