Since nobody replied I am posting it to JIRA. Apostolos Pantsiopoulos wrote:
Hi I noticed that after a day of relatively moderate traffic (about 400 simultaneous channels average) the memory used by FS reached 1.3 GB of RAM. I tried to trace the leak (if any) with valgrind and got that :
==18894== 572 bytes in 1 blocks are definitely lost in loss record 148 of 193 ==18894== at 0x40053C0: malloc (vg_replace_malloc.c:149) ==18894== by 0x7AFDE93: sofia_glue_do_invite (sofia_glue.c:1599) ==18894== by 0x7ACB191: sofia_on_init (mod_sofia.c:102) ==18894== by 0x405B124: switch_core_session_run (switch_core_state_machine.c:480) ==18894== by 0x4058204: switch_core_session_thread (switch_core_session.c:1064) ==18894== by 0x40CD6C5: dummy_worker (thread.c:138) ==18894== by 0xB5346A: start_thread (in /lib/libpthread-2.5.so) ==18894== by 0xAAADBD: clone (in /lib/libc-2.5.so)
So, I assummed that this happens for every call. I tried testing it again by placing two calls before shutting down FS, but it only came up once. I wanted to get a second opinion before posting this to JIRA as an issue.
I used revision 14269 of the SVN. I am attaching the valgrind output as well.
I also noticed that only one of my CPU cores gets really busy when dealing with moderate traffic. From what I read in the mailing list users are encouraged to use 64bit multi core servers for FS because it scales up better. But this is not what I am seeing in practice. Could the single threaded architecture of libsofia be the cause of that behavior?
------------------------------------------------------------------------
_______________________________________________ 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
-- ------------------------------------------- Apostolos Pantsiopoulos Kinetix Tele.com R & D email: regs@kinetix.gr -------------------------------------------