Asterisk 13.10.0 RAM usage grows linearly

Hello everybody!
I have an issue with RAM utilization with Asterisk 13.10.0 running on a Xeon E7 with 32Gb RAM powered by gentoo. Asterisk was originally built with emerge - version 13.9 and later upgraded to 13.10.0 and recompiled from source code with MALLOC_DEBUG flag to monitor RAM utilization.
Asterisk is configured with Realtime res/mysql and has about 450 registered SIP peers and ~20-25 active calls at one moment.
After restarting asterisk core RAM utilization becomes minimal, it grows linearly until the damaged calls are detected.
(In 13.10.0 version RAM grows slower but still permanent compared with 13.9 version)
“memory show allocations” shows summary allocation - 20 МB but the real ram utilization is about 10G now and grows up.
I suppose that asterisk with less than 10Mbps traffic and about 20 realtime calls could’t use so many memory. Any ideas?
http://pastebin.com/Tut3yYUt - output of memory show summary

pmap -x output:
/usr/sbin/asterisk
Address Kbytes RSS Dirty Mode Mapping


total kB 21361176 16965480 16945328

Memory summary total:
19313766 bytes in all allocations
721762 bytes in deferred free large allocations
40938 bytes in deferred free small allocations
762700 bytes in deferred free allocations
20076466 bytes in all allocations and deferred free allocations

I have the same problem on a 13.5 system. It has a somewhat low call volume, but allocates at 8% of 16 GB of RAM daily. I upgraded one server to 13.8 (there are many fixes for memory leaks, and the ‘core show taskprocessors’ command will actually give meaningful data). I haven’t upgraded the live server yet, but probably will tomorrow night. I’ll try and remember to post back here what results I find.

Hi,

Sorry for reviving this old thread but we have the exact same problem with Asterisk 13.16.0.

Any config recommendations to try?.

Thanks.

My problem had to do with using realtime voicemail on 13.5 and 13.8. They patched for that problem. I’ve been running 13.12 for over a year on multiple servers and have not had that issue since.