VoIP Mailing List Archives

[fraunhofer.lists.frees...]

Hello *,

a memory leak showed up in our loadtests. It's (still) the same setup as in the
http://jira.freeswitch.org/browse/MODSOFIA-22 bugfix.

One thing I'd like to add is that "fsctl shutdown restart" was unable
to shutdown freeswitch.
The last line printed is "switch_core_memory.c:567 Stopping memory pool queue."

attached file is a the collected and graphed output of some "ps waux"
command. "sz" should be
in "physical pages" (2k?).

I rerun the test and this time it coredumped trying to malloc() space
for some playback.

Anything else you need (full backtraces?) to dig into it?

Cheers
Beni.

_______________________________________________
FreeSWITCH-users mailing list
FreeSWITCH-users@lists.freeswitch.org
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
http://www.freeswitch.org

[anthony.minessale at g...]

600k is not a leak? FS can use as much as a gig of ram or more depending on what you are doing.
you may want to install a fresh copy of FS, removing all your old files etc and make sure they build clean.
We also have not had much luck running on ubuntu which is more of a desktop centric OS.
I recommend you try your application on 64 bit CentOS which is the platform all of our paid customers use.





On Thu, Aug 27, 2009 at 8:15 AM, Benedikt Fraunhofer <fraunhofer.lists.freeswitch-001@traced.net (fraunhofer.lists.freeswitch-001@traced.net)> wrote:

[fraunhofer.lists.frees...]

Hello Anthony,

2009/8/27 Anthony Minessale <anthony.minessale@gmail.com>:

[jaybinks at gmail.com]

Anthony can you ( or anyone else alao ). Please elaborate on what
makes centos 5.3 o much better for Freeswitch.

Is there some specific library vesiion on centos that makes a massive
difference ?

Reason I ask ... I personally only have a preference for debian,
but others may have policy mandated Os's
For their companies, and it would be great to have some info about this.

I'd imagine this actually boils down to a requirement for libfoo
version X ... And if we ran those library versions on other OS's we'd
be fine ?

Jay

[anthony.minessale at g...]

Mostly based on that is what we developed it with and continue to use and have had the best luck and least problems with.

I am sure you are right about it being certain version of the core toolchain
I will guess it's the combo of libc and kernel and the fact that CentOS has a purist attitude towards patching such code.

Some people use stable debian with no issues, other have reported problems resolved by switching to CentOS

so presumably you could roll your own debian that was similar but we don't have the resources to try to find out what makes what happen etc.

If you can configure debian to work, it's fine, if we get some inexplicable problem on debian we will start to wonder though.
it's happened more than once with people doing nothing other than changing OS =D



On Thu, Aug 27, 2009 at 4:55 PM, Jay Binks <jaybinks@gmail.com (jaybinks@gmail.com)> wrote:

[jason at jasonjgw.net]

Jay Binks <jaybinks@gmail.com> wrote:

[diego.viola at gmail.com]

Cool, feel free to report these bugs in Jira. If they are indeed FreeSWITCH bugs.

On Fri, Aug 28, 2009 at 12:10 AM, Jason White <jason@jasonjgw.net (jason@jasonjgw.net)> wrote:

[fraunhofer.lists.frees...]

Hello *,

still no luck. Removing the default dialplan with all those
hash-inserts made the memory-usage-curve flatter but it oom-coredumped
stating:

freeswitch: src/switch_core_memory.c:443: switch_core_perform_alloc:
Assertion `ptr != ((void *)0)' failed.

after eating up 3.2 gig ram+swap.

Any chance or idea to track that down or any hint where to look at?

The other two freeswitch-machines in this scenario play well, they're
dumb load-sinks. what makes this machine different is on one hand the
lua-scripts that are run on it and on the other hand it's the only
machine who gets it's commands via xml_rpc.

Cheers
Beni.

_______________________________________________
FreeSWITCH-users mailing list
FreeSWITCH-users@lists.freeswitch.org
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
http://www.freeswitch.org

[brian at freeswitch.org]

Use valgrind.

/b

On Aug 31, 2009, at 7:22 AM, Benedikt Fraunhofer wrote:

[rupa at rupa.com]

Isn't there a known issue with lua+sql leaking memory on some platforms?

On Mon, Aug 31, 2009 at 8:32 AM, Brian West<brian@freeswitch.org> wrote:

[fraunhofer.lists.frees...]

Hello *,

2009/8/31 Rupa Schomaker <rupa@rupa.com>:

[brian at freeswitch.org]

What are you doing in these lua scripts? Because there are a few things you can do in the lua script itself that will cause you to leak like crazy due to improper use.

/b

On Sep 2, 2009, at 7:59 AM, Benedikt Fraunhofer wrote:

[fraunhofer.lists.frees...]

Hello Brian,

2009/9/2 Brian West <brian@freeswitch.org>:

[anthony.minessale at g...]

You have to reduce the load when running valgrind.


On Wed, Sep 2, 2009 at 7:59 AM, Benedikt Fraunhofer <fraunhofer.lists.freeswitch-001@traced.net (fraunhofer.lists.freeswitch-001@traced.net)> wrote:

[anthony.minessale at g...]

run it slower and make sure it shuts down clean.

valgrind --tool=memcheck --log-file-exactly=vg.log --leak-check=full --leak-resolution=high --show-reachable=yes /path/to/freeswitch -vg


On Wed, Sep 2, 2009 at 10:11 AM, Anthony Minessale <anthony.minessale@gmail.com (anthony.minessale@gmail.com)> wrote: