[ejabberd] memory profiling
lee at dropio.com
Sat Mar 21 00:51:42 MSK 2009
Thanks for the advice. Looking at the top of the crash dump, I noticed
it confirms memory exhaustion:
Thu Mar 19 16:39:49 2009
Slogan: eheap_alloc: Cannot allocate 1271244 bytes of memory (of type "heap").
System version: Erlang (BEAM) emulator version 5.6.3 [source]
Compiled: Sun Jul 20 04:41:21 2008
2 billion processes seems a bit high. The server maintains about 1.4K
connected anonymous users with associated http_bind connections, and
around 40K online MUC rooms.
I'll do more digging tomorrow.
On Fri, Mar 20, 2009 at 5:30 PM, Christopher Zorn
<christopher.zorn at gmail.com> wrote:
> On Fri, Mar 20, 2009 at 5:23 PM, Lee Azzarello <lee at dropio.com> wrote:
>> I'm running ejabberd SVN revision 1978. My application is focused on
>> http_bind and mod_muc. A very large number of persistant rooms is
>> created and a script cleans them out using mod_muc_admin functions.
>> Unfortunately, RAM isn't getting freed and ejabberd has been crashing
>> reliably every three days. I'm looking for some advice on how to get
>> more detailed information out of ejabberd so I can target the problem.
>> RAM is the only resource that's scarce, load and CPU are almost
> You can try the eper tools. http://code.google.com/p/eper/
> But you can get some information about of the crash dump too. It will tell
> you why it crashed and you can find out where by looking at the heap or the
> best place usually is to look for "Message queue length:" in the process
> information of the dump.
> The dump is called erl_crash.dump and most of the time located in the
> ejabberd log directory.
> The erlang docs for crash dumps help too.
> I hope this helps. It is at least a place to start.
>> ejabberd mailing list
>> ejabberd at jabber.ru
> ejabberd mailing list
> ejabberd at jabber.ru
More information about the ejabberd