One way audio in records made by MixMonitor

Asterisk Asterisk Support
1

Hi all,

We are using MixMonitor() to record calls in the call center.
The problem occurs on outbound calls made by Dial() and does not detected on incoming calls (usually passed to Queue()). Record is being done separately for receive and transmit party of the channel.

Sometimes (approx. every 10th call) we’re missing customer’s voice in the file being recorded as t()
The file contains valid WAV header and valid length, but filled with zeroes. The other file contains operator’s voice.

Comparing logs of valid recordings with invalid ones, I’ve noticed the precense of lines like this:

[Oct 8 13:03:39] DEBUG[123800][C-00000f58] audiohook.c: Failed to get 160 samples from write factory 0x7d22b800ce08
[Oct 8 13:03:39] DEBUG[123800][C-00000f58] audiohook.c: Read factory 0x7d22b800c3c8 and write factory 0x7d22b800ce08 both fail to provide 160 samples

These messages are repeating throughout the entire call.
At the same time operator and customer hear each other and speaking pretty well.

This problem looks like the same mentioned here:

Asterisk version: Asterisk 20.6.0~dfsg+~cs6.13.40431414-2build5
Being run on Ubuntu 24.04.1 in VMware environment.

Regards,
Dmitry

2

The same problem using version 20.8.1~dfsg+~cs6.14.40431414-1 (packages from Ubuntu 24.10 installed manually on Ubuntu 24.04.1)

3

I’ve spent many hours in reading the source code and the logs (see att.).

Can’t figure out what exactly causing the write factory to be exhausted, or its pointer is invalid?

It looks to me there are no packets in the buffer while audiohook reads frames from it.
The reading call stack:

apps/app_mixmonitor.c:mixmonitor_thread()
main/audiohook.c:ast_audiohook_read_frame_all()
main/audiohook.c:audiohook_read_frame_helper()
main/audiohook.c:audiohook_read_frame_both()

And here the line

usable_write = (ast_slinfactory_available(&audiohook->write_factory) >= samples ? 1 : 0);

sets usable_write to 0, and then the debug message is being generated:

		ast_debug(3, "Failed to get %d samples from write factory %p\n", (int)samples, &audiohook->write_factory);

And here is the call stack for putting frames into the write buffer:

main/channel.c:ast_write_stream()
main/audiohook.c:ast_audiohook_write_list()
main/audiohook.c:audio_audiohook_write_list()
main/audiohook.c:ast_audiohook_write_frame()
main/slinfactory.c:ast_slinfactory_feed()
include/asterisk/linkedlists.h:AST_LIST_INSERT_TAIL()

Unfortunately this bug is being reproduced only on production system under the load, so it is not possible to debug with gdb and catch the real stack traces and variables’ values.

The calling context in extensions.conf is using MixMonitor() and Dial() as follows:

[DLPN_Internet_prov1_company2] ;company2 OUT  from prefix *2

exten => _X.,1,NoOp(${EXTEN:-10},${CALLERID(num)},${CALLERID(name)},${CHANNEL},${SIPCALLID} ----- company2 OUT CALL XXXX to ${EXTEN})

same => n,Gosub(sub-recordcall-ecom,s,1(${UNIQUEID},out))

same => n,Set(CDR(userfield)=company2 OUT)
same => n,Set(ORIGCALLERID=${CALLERID(num)})
same => n,Set(QUEUENUMBER=company2)
same => n,Set(EXTENNUMBER=${EXTEN:-10})
same => n,Set(CALLERID(num)=4051234567) ; MAIN FOR 0800XXXXXX
same => n,Dial(SIP/prov1_Trunk/${EXTEN:-10},,Tt)
same => n,StopMixMonitor()
same => n,Hangup

exten => h,1,StopMixMonitor()
exten => h,n,AGI(pyagi_is_out.py)

[sub-recordcall]
;
; Basic call recording subroutine
;
; ${ARG1} - UNIQUEID
; ${ARG2} - path
; ${ARG3} - Call direction: [in|out]
;
exten => s,1,Set(CALLFILENAME=${ARG1})
same  =>   n,Set(CALLFILEPATH=${ARG2})
same  =>   n,Set(FULLPATH=${RECORDS_DIR}/${CALLFILEPATH})
same  =>   n,Set(CALLFILENAME_MP3=${CALLFILENAME}.mp3)
same  =>   n,Set(CALLFILENAME_WAV=${CALLFILENAME}.wav)
same  =>   n,Set(WAV_R=${IF($["${ARG3}" = "in"]?R:L)})
same  =>   n,Set(WAV_L=${IF($["${ARG3}" = "in"]?L:R)})
same  =>   n,NoOp(${WAV_R} -> ${WAV_L})
same  =>   n,Set(CDR(filename)=${CALLFILEPATH}/${CALLFILENAME_MP3})
same  =>   n,Set(MONOPT=/usr/bin/nice -n 19 /usr/bin/ffmpeg -y -i ${FULLPATH}/${CALLFILENAME}.R.wav -i ${FULLPATH}/${CALLFILENAME}.L.wav -filter_complex "[0:a][1:a]join=inputs=2:channel_layout=stereo[a]" -map "[a]" -codec:a libmp3lame -qscale:a 5 ${FULLPATH}/${CALLFILENAME_MP3} && rm -f ${FU
LLPATH}/${CALLFILENAME}.[LR].wav)
same  =>   n,MixMonitor(,br(${FULLPATH}/${CALLFILENAME}.${WAV_R}.wav)t(${FULLPATH}/${CALLFILENAME}.${WAV_L}.wav),${MONOPT})
same  =>   n,Return()

I would appreciate if someone will drop a hint how to find a root cause of this problem and fix it.

PS: Configs and logs are obfuscated due to data privacy

log_bad.txt (2.6 MB)
log_good.txt (3.0 MB)

4

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.