hls stops working: force fragment split

[vvidic]

Any idea why this configuration stops working after a few days:

worker_processes 4;
pid /run/nginx.pid;
#error_log  /var/log/nginx/debug.log debug;

events {
     worker_connections 1024;
     multi_accept on;
 }

 rtmp_auto_push on;
 rtmp {
    server {
        listen 1935;

        application student {
            live on;

            hls on;
            hls_path /run/shm/hls-student;

            dash on;
            dash_path /run/shm/dash-student;

            # pull to start hls export
            pull rtmp://lb.av.carnet.hr/student name=one static;
         }
}

2015/01/23 09:22:45 [error] 16433#0: *1 hls: force fragment split: 682283.103 sec, , client: lb.av.carnet.hr/student, server: ngx-relay
2015/01/23 09:22:45 [error] 16433#0: *1 hls: force fragment split: -682275.068 sec, , client: lb.av.carnet.hr/student, server: ngx-relay
2015/01/23 09:22:53 [error] 16433#0: *1 hls: force fragment split: 682275.068 sec, , client: lb.av.carnet.hr/student, server: ngx-relay
2015/01/23 09:22:53 [error] 16433#0: *1 hls: force fragment split: -682267.034 sec, , client: lb.av.carnet.hr/student, server: ngx-relay
2015/01/23 09:23:01 [error] 16433#0: *1 hls: force fragment split: 682267.034 sec, , client: lb.av.carnet.hr/student, server: ngx-relay
2015/01/23 09:23:01 [error] 16433#0: *1 hls: force fragment split: -682259.000 sec, , client: lb.av.carnet.hr/student, server: ngx-relay
2015/01/23 09:23:09 [error] 16433#0: *1 hls: force fragment split: 682259.000 sec, , client: lb.av.carnet.hr/student, server: ngx-relay
2015/01/23 09:23:09 [error] 16433#0: *1 hls: force fragment split: -682250.943 sec, , client: lb.av.carnet.hr/student, server: ngx-relay
2015/01/23 09:23:17 [error] 16433#0: *1 hls: force fragment split: 682250.943 sec, , client: lb.av.carnet.hr/student, server: ngx-relay
2015/01/23 09:23:17 [error] 16433#0: *1 hls: force fragment split: -682242.909 sec, , client: lb.av.carnet.hr/student, server: ngx-relay

[mknwebsolutions]

I believe this is the same issue @ https://github.com/arut/nginx-rtmp-module/issues/346

[vvidic]

On Sat, Jan 24, 2015 at 01:42:56PM -0800, Michael Mikhjian wrote:

I believe this is the same issue @ https://github.com/arut/nginx-rtmp-module/issues/346

It could be, however I'm using the latest git version (7a35372). Restarting nginx doesn't help as the errors start appearing right away. I have to restart the upstream RTMP server to get HLS working again but this is not very practical for live streams. I have a capture of the failing RTMP stream so maybe you can use this to figure out why the HLS generator stops working.

Valentin

[dburianov]

I have it problem too and nginx create more small files about 300-15000 bytes size.

2016/02/14 14:05:53 [error] 952#0: *18459 hls: force fragment split: -78.252 sec, , client: 127.0.0.1, server: 0.0.0.0:1935 2016/02/14 14:05:53 [error] 952#0: *18460 hls: force fragment split: 78.391 sec, , client: 127.0.0.1, server: 0.0.0.0:1935

It work only in this config: hls on; hls_path /usr/local/nginx/html/video; hls_fragment 10s; hls_playlist_length 60s; hls_continuous on; hls_fragment_naming timestamp; dash off; dash_path /usr/local/nginx/html/dash; dash_fragment 15s; dash_playlist_length 60s;

dash off and dash* present If I comment or remove dash* in config I have this errors.

Sometimes I have nest errors in ffmpeg logs:

Error while decoding stream #0:1: Invalid data found when processing input [mpegts @ 0x3868740] PES packet size mismatchme=00:21:51.90 bitrate= 828.8kbits/s speed=0.933x [mpeg2video @ 0x38db0a0] 00 motion_type at 3 32 [mpeg2video @ 0x38db0a0] 00 motion_type at 6 24 [mpeg2video @ 0x38db0a0] 00 motion_type at 5 25 [mpeg2video @ 0x38db0a0] mb incr damaged [mpeg2video @ 0x38db0a0] 00 motion_type at 12 27 [mpeg2video @ 0x38db0a0] 00 motion_type at 12 33 [mpeg2video @ 0x38db0a0] 00 motion_type at 4 28 [mpeg2video @ 0x38db0a0] 00 motion_type at 25 34 [mpeg2video @ 0x38db0a0] 00 motion_type at 5 16 [mpeg2video @ 0x38db0a0] 00 motion_type at 26 35 [mpeg2video @ 0x38db0a0] ac-tex damaged at 29 20 [mpeg2video @ 0x38db0a0] 00 motion_type at 4 21 [mpeg2video @ 0x38db0a0] invalid mb type in P Frame at 21 22 [mpeg2video @ 0x38db0a0] 00 motion_type at 2 23 [mpeg2video @ 0x38db0a0] 00 motion_type at 5 8 [mpeg2video @ 0x38db0a0] mb incr damaged [mpeg2video @ 0x38db0a0] 00 motion_type at 18 9 [mpeg2video @ 0x38db0a0] 00 motion_type at 1 13 [mpeg2video @ 0x38db0a0] 00 motion_type at 15 10 [mpeg2video @ 0x38db0a0] 00 motion_type at 9 14 [mpeg2video @ 0x38db0a0] 00 motion_type at 9 11 [mpeg2video @ 0x38db0a0] 00 motion_type at 2 15 [mpeg2video @ 0x38db0a0] 00 motion_type at 11 7 [mpeg2video @ 0x38db0a0] 00 motion_type at 7 1 [mpeg2video @ 0x38db0a0] mb incr damaged [mpeg2video @ 0x38db0a0] 00 motion_type at 8 3 [mpeg2video @ 0x38db0a0] 00 motion_type at 1 4 [mpeg2video @ 0x38db0a0] invalid cbp -1 at 2 5 [mpeg2video @ 0x38db0a0] 00 motion_type at 10 6 [mpeg2video @ 0x38db0a0] invalid cbp -1 at 9 7 [mpeg2video @ 0x38db0a0] 00 motion_type at 6 29 [mpeg2video @ 0x38db0a0] ac-tex damaged at 5 30 [mpeg2video @ 0x38db0a0] ac-tex damaged at 9 17 [mpeg2video @ 0x38db0a0] 00 motion_type at 13 31 [mpeg2video @ 0x38db0a0] invalid cbp -1 at 9 19 [mpeg2video @ 0x38db0a0] 00 motion_type at 12 19 [mpeg2video @ 0x38db0a0] Warning MVs not available [mpeg2video @ 0x38db0a0] concealing 1440 DC, 1440 AC, 1440 MV errors in P frame [mp2 @ 0x386ad60] Header missing Error while decoding stream #0:1: Invalid data found when processing input

ffmpeg get mpeg2 stream via http.

Sometimes I have next message: [libfdk_aac @ 0x388f220] Queue input is backward in time [flv @ 0x38d2880] Non-monotonous DTS in output stream 0:1; previous: 1437181, current: 1437173; changing to 1437181. This may result in incorrect timestamps in the output file.

but it does not lead to error

nginx 1.9.9 ffmpeg version N-78299-gf021030 - compiling from sources

I use script when monitoring folder and do next: stop nginx remove *.ts files with small size start nginx

[guoyi8]

@dburianov I also have this problem. Have you solved it?

[aaronpk]

I'm having this problem as well, but only when streaming from Larix Broadcaster on iOS. Streaming from OBS on my Mac works fine.

[alphacentauri82]

i have the same problem. And i have NO IDEA!

[dburianov]

It problem your based on audiocodec Sometimes it been repaired with change audiocodec or in serials encoding(as enc only video and audio copy and next video copy and audio to aac)

[alphacentauri82]

How do i fix it?. That's the question. I can see the stream just fine but it's jumpy and delayed at times.

[milesegan]

I'm seeing this issue too. Has anyone found a solution?

[campones]

i have this while using an rtmp application that push the original stream toward several servers. starts creating thousands or mini segments

[dburianov]

Now I use build ffmpeg from source code and do not have this problem ffmpeg 3.4.2 ./configure
--prefix="/usr/ffmpeg_build"
--pkg-config-flags="--static"
--extra-cflags="-I/usr/ffmpeg_build/include"
--extra-ldflags="-L/usr/ffmpeg_build/lib"
--extra-libs="-lpthread -lm"
--bindir="/usr/bin"
--enable-gpl
--enable-libass
--enable-libfdk-aac
--enable-libfreetype
--enable-libmp3lame
--enable-libopus
--enable-libtheora
--enable-libvorbis
--enable-libvpx
--enable-libx264
--enable-libx265
--enable-nonfree
--enable-filter=movie --enable-filter=drawtext --enable-libfreetype --enable-filter=overlay --enable-filter=yadif

[hgshoggins]

Having the same problem here. @dburianov, do you have an idea about what has changed from your distribution version to your own ffmpeg compilation? I'm using ffmpeg 4.0.4 from Fedora 29.

What happens to me looks like a (I know I'm not using the appropriate words) "desync" with the chunk size or something like that: 2019/05/04 15:25:05 [error] 969#0: *101 hls: force fragment split: 10.007 sec, , client: 127.0.0.1, server: 0.0.0.0:1935 2019/05/04 15:25:15 [error] 969#0: *101 hls: force fragment split: 10.008 sec, , client: 127.0.0.1, server: 0.0.0.0:1935 2019/05/04 15:25:25 [error] 969#0: *101 hls: force fragment split: 10.008 sec, , client: 127.0.0.1, server: 0.0.0.0:1935

[mkrn]

The issue is happening when the audio and video timestamps are starting from the different start timestamp. This is valid, however the nginx-rtmp library hls splitter does not account for this, taking dts for timestamp and assumes that they have the same start in audio and video.

To fix it I used -vf "setpts=PTS-STARTPTS" -af "asetpts=PTS-STARTPTS" as ffmpeg options.

Note: the nginx-rtmp interleave and sync options may mess around with timestamps even further.

[lve4code]

I am facing the same issue. Is there any updated solution to this error?

[IgorDePaula]

Now I use build ffmpeg from source code and do not have this problem ffmpeg 3.4.2 ./configure --prefix="/usr/ffmpeg_build" --pkg-config-flags="--static" --extra-cflags="-I/usr/ffmpeg_build/include" --extra-ldflags="-L/usr/ffmpeg_build/lib" --extra-libs="-lpthread -lm" --bindir="/usr/bin" --enable-gpl --enable-libass --enable-libfdk-aac --enable-libfreetype --enable-libmp3lame --enable-libopus --enable-libtheora --enable-libvorbis --enable-libvpx --enable-libx264 --enable-libx265 --enable-nonfree --enable-filter=movie --enable-filter=drawtext --enable-libfreetype --enable-filter=overlay --enable-filter=yadif

You are using ffmpeg fron source, but how, where?