[ejabberd] How to trace 'bad' users

Evgeniy Khramtsov xramtsov at gmail.com
Mon Feb 21 10:32:26 MSK 2011


21.02.2011 16:24, Peter Viskup wrote:
> Hi all,
> I am getting following error message from ejabberd since yesterday:
>
> [3:23:17 AM] jabber.sk: (ejabberd at localhost) The process<0.14889.165>  is
> consuming too much memory:
> [{old_heap_block_size,0},
>   {heap_block_size,228065700},
>   {mbuf_size,0},
>   {stack_size,83},
>   {old_heap_size,0},
>   {heap_size,182452621}]
> Process type: s2s_out
> S2S connection: from jabber.sk to conference.jabber.ru
> Possible reason: the process can't process messages faster than they
> arrive.  Current function is prim_inet:recvfrom0/3
> Command to kill this process: kill ejabberd at localhost<0.14889.165>
>
> How to trace which accounts are used to send such amount of messages to
> conference.jabber.ru?

 From remsh (aka `ejabberdctl debug`):
 > process_info(pid(0,14889,165), messages).

You will see "stanzas" in the process's queue. Those "stanzas" will be 
in Erlang format, but it is pretty straightforward.

-- 
Regards,
Evgeniy Khramtsov, ProcessOne.
xmpp:xram at jabber.ru.



More information about the ejabberd mailing list