Re: Does opensmppbox and smsbox have serious memory issues?

Andreas Fink <[email protected]> Sun, 27 Apr 2014 08:27:54 +0200
Newsgroups	gmane.comp.mobile.kannel.devel
Message-ID	<[email protected]>
If there's a memory leak (which it looks like) it should be relatively easy to spot and fix.
valgrind should be able to tell you where the blocks are allocated which are not freed.

On 26.04.2014, at 22:31, [email protected] wrote:

Hi Rene and other Kannel developers,

Hanh posted his Valgrind research to the user group for smsbox and
opensmppbox.  His results seem interesting and so I'm copying them to this
thread so the Kannel developers can view them.
These results can be viewed by following the thread on Wed, Apr 23, 2014 at
3:41 AM, by Hanh Le Bich <[email protected]> with the Subject: Re: 2
Questions re Redis/Debian.
(The email subject is not related to this issue.)

His research shows that opensmppbox and smsbox may have serious memory
issues.
I use the word "may" as until others have confirmed his results, there could
be a mistake somewhere.
Is there anyone who has a test environment that can follow his approach and
confirm for the Kannel community if opensmppbox and smsbox have serious
memory issues?

His approach is:
Let me describe a little bit for my application back end. It's  pretty
simple: i make a loop that for each second, it push an sms via kannel
CGI for 1K mobile numbers, that mean throughput is 1000 msg/sec.
My kannel configuration is simple too, it's only smsbox -> bearerbox
-> SMSC (via smpp), no file storage, no SQL, no dlr (actually
dlr-mask=8).

For smsbox:
In the pass when using ver 1.4.3, it was fine for years. After
upgrade to 1.5.0, after each few days, i realized smsbox is reset,
then i found it exhaust my memory. It's funny that smsbox consume the
mem and doesn't release. Example, if it occupies 50% your mem and you
stop sms pushing, it will 50% forever except the box restarting.
That's all, same server with no other tasks, same back end, just
different kannel version.

Just paste the valgrind sum in here:

==27581== LEAK SUMMARY:
==27581==    definitely lost: 1,077,904 bytes in 67,369 blocks
==27581==    indirectly lost: 673,660 bytes in 67,366 blocks
==27581==      possibly lost: 160 bytes in 13 blocks
==27581==    still reachable: 1,240 bytes in 39 blocks
==27581==         suppressed: 0 bytes in 0 blocks
==27581== Reachable blocks (those to which a pointer was found) are
not shown.
==27581== To see them, rerun with: --leak-check=full
--show-leak-kinds=all ==27581== ==27581== For counts of detected and
suppressed errors, rerun with: -v ==27581== ERROR SUMMARY: 3 errors
from 3 contexts (suppressed: 45 from 10)

For opensmppbox
opensmppbox  drains your memory 10 times faster than smsbox
==31087== LEAK SUMMARY:
==31087==    definitely lost: 78,944 bytes in 4,882 blocks
==31087==    indirectly lost: 4,911,232 bytes in 4,859 blocks
==31087==      possibly lost: 48,496 bytes in 74 blocks
==31087==    still reachable: 3,124,401 bytes in 26,735 blocks
==31087==         suppressed: 0 bytes in 0 blocks
==31087== Reachable blocks (those to which a pointer was found) are
not shown.
==31087== To see them, rerun with: --leak-check=full
--show-leak-kinds=all ==31087== ==31087== For counts of detected and
suppressed errors, rerun with: -v ==31087== ERROR SUMMARY: 10 errors
from 10 contexts (suppressed: 45 from 10)

Regards









Andreas Fink

CEO DataCell ehf
CEO Backbone ehf

---------------------------------------------------------------
Tel: +41-61-6666330 Fax: +41-61-6666331  Mobile: +41-79-2457333
Address: Clarastrasse 3, 4058 Basel, Switzerland
E-Mail:  [email protected]
www.datacell.com, www.backbone.is, www.finkconsulting.com www.fink.org
---------------------------------------------------------------
Jabber/XMPP: [email protected]
ICQ: 8239353 Skype: andreasfink
signature.asc (application/pgp-signature, 841 B)