自由开关被阻止

Question

FreeSwitch software working well in a few days (~3 - 5 days), then new incoming call requests are accepted since FreeSwitch is blocked !! Ongoing calls continue their session, their calls seems not effected, but new calls are not accepted. I got FreeSwitch snapshot and analyzed it in GDB.

I have 601 therads & most of them are waiting

Thread 0x7f16bc55f700 (LWP 28544) pthread_cond_wait@@GLIBC_2.3.2 () at ../nptl/sysdeps/unix/sysv/linux/x86_64/pthread_cond_wait.S:185

When i apply "thread apply all bt" in gdb, I see most of the threads try to push events into queue (switch_queue_push )

Thread 600 (Thread 0x7f16bc55f700 (LWP 28544)):
#0 pthread_cond_wait@@GLIBC_2.3.2 () at ../nptl/sysdeps/unix/sysv/linux/x86_64/pthread_cond_wait.S:185
#1  0x00007f180cf9b87d in apr_thread_cond_wait (cond=<optimized out>, mutex=<optimized out>) at locks/unix/thread_cond.c:68
#2  0x00007f180cf92dd0 in apr_queue_push (queue=queue@entry=0x7f180db157a8, data=data@entry=0x7f16d3d5ec20) at misc/apr_queue.c:166
#3  0x00007f180cc958fb in switch_queue_push (queue=0x7f180db157a8, data=data@entry=0x7f16d3d5ec20) at src/switch_apr.c:1134
#4  0x00007f180cd17850 in switch_event_queue_dispatch_event (eventp=0x7f16bc55ec48) at src/switch_event.c:384
#5  switch_event_fire_detailed (file=file@entry=0x7f180cfb07ea "src/switch_channel.c", func=func@entry=0x7f180cfb2ba0 <__func__.18348> "switch_channel_perform_set_running_state", line=line@entry=2260, event=event@entry=0x7f16bc55ec48, user_data=user_data@entry=0x0) at src/switch_event.c:1986
#6  0x00007f180cc9f118 in switch_channel_perform_set_running_state (channel=0x7f17e3e7de00, state=CS_NEW, file=0x7f180cfbc590 "src/switch_core_state_machine.c", func=<optimized out>, line=543) at src/switch_channel.c:2260
#7  0x00007f180ccc87d0 in switch_core_session_run (session=0x7f17e3e7fd28) at src/switch_core_state_machine.c:543
#8  0x00007f180ccc36de in switch_core_session_thread (thread=<optimized out>, obj=0x7f17e3e7fd28) at src/switch_core_session.c:1629
#9  0x00007f180ccbf47d in switch_core_session_thread_pool_worker (thread=0x7f17e3e9abb0, obj=0x80) at src/switch_core_session.c:1692
#10 0x00007f180cfa1910 in dummy_worker (opaque=0x7f17e3e9abb0) at threadproc/unix/thread.c:151
#11 0x00007f180c1e0064 in start_thread (arg=0x7f16bc55f700) at pthread_create.c:309
#12 0x00007f180b8b862d in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:111

Why i am gonna getting this state? Any thoughts, tips, tricks would be much appreciated. Regards,

Answer 1

I've deeply researched and found the my solution,

You should understand " Event handler mechanism " of Freeswitch in order to solve this issue. Because there are many producer threads that generate and put its event to this queue, however only one consumer thread exist in this mechanism. Consumer thread which is as known event handler delivers event to the all listeners such as modules that listen appropriate event.

One or more of these listeners could hold up (via blocking) this consumer thread, and event queue could becoming get a full. When event queue getting a full, your feeswitch will be blocked.

There are three solution for solving theses issue:

1: In default configuration, event handler mechanism uses event queue. But, you can use thread solution for this with changing " events-use-dispatch=false " value in " post_load_switch.conf " file.

2: Inscreasing consumer threads number , because single consumer thread is not good solution for heavy load freeswitch server you could do it with " initial-event-threads=X " in " post_load_switch.conf " file where X represent initial thread count.

3: In your module, you could use your event handler mechanism too. When you gets a event from core of Freeswitch, create a new thread and assign it into the your newly created thread in order to not waiting Freeswitch's consumer thread.