[Serusers] Memory leak?

Andrei Pelinescu-Onciul pelinescu-onciul at fokus.fraunhofer.de
Wed May 28 18:22:49 CEST 2003


On May 28, 2003 at 08:46, Jamin W. Collins <jcollins at asgardsrealm.net> wrote:
> On Wed, May 28, 2003 at 01:31:45PM +0200, Andrei Pelinescu-Onciul wrote:
> > On May 27, 2003 at 13:41, Jamin W. Collins <jcollins at asgardsrealm.net> wrote:
[...]
> > 
> > Probably your debug level is < 4 (4 generates lots of debugging
> > messages).
> 
> Yea, it's at the default of 3.

I just changed the log level for the mem. status messages to 3 on cvs,
for the new version.  (I think it's better to get the mem. status on
program exit and still be able to filter out the other dbg. messages).
> 
> > Could you send us the whole log, (or  grep ser syslog), the cfg. file
> > and the packet dumps you mentioned? Use serhelp at lists.iptel.org (no limits and
> > not public, only ser developers on it). Maybe we can figure out where
> > the mem. leak is.
> 
> Sent them in to serhelp.

thanks
> 
> > You could try to start ser with maximum debuging (e.g debug=9) let it
> > run a few days under your expected load, then stop it /send SIGUSR1 and
> > look in the log for the lines:
> > qm_status  
> > heap size= ...
> > used= ..., used+overhead=..., free
> > max used (+overhead)= ...
> > 
> > for each process.
> > 
> > max used will be very interesting.
> 
> How much (if any) would this impact the performance of the proxy?
> Trying to limit the impact to the customer.

It should work without any visible impact for 100-200 cps. It might be
ok for more  (depends also on syslog speed, disk a.s.o).
It will also write huge logs (so if your /var partition is small ...).
You might consider also setting log_stderr=yes and starting ser with
 ser 2>/tmp/ser_log.

BTW: if you want maximum speed from ser, change Makefile.defs:
 remove -DDBG_QM_MALLOC and add -DF_MALLOC -DNO_DEBUG. -DNO_DEBUG will
 kill most of the debug messages (but you don't see them anyway since
 you are running ser with debug=3). -DF_MALLOC will use a much faster
 malloc version, w/o any debugging code.



Andrei




More information about the sr-users mailing list