input.ffmpeg - "We must catchup xx seconds!"

[JonathanThorpe]

Hi All,

I have two instances of LiquidSoap running (2.2.4 and also a rolling release from a couple of weeks ago but I've seen this on earlier releases too) which exhibits: [clock.main:2] We must catchup xx seconds! which seems to increment as you might expect.

I have two hosts that are running Liquidsoap - a production one and a staging one. Each host runs two instances of Liquidsoap - one using input.ffmpeg as the source for a UDP stream coming from a Barix Instreamer and the other using input.external.rawaudio which runs a VBAN Receptor (Voicemeeter). Only the ffmpeg.input one (on both hosts) exhibits the "We must catch up" hehaviour and I'm at a loss to fix it.

For reference, the input for input.ffmpeg is configured as follows:

input.ffmpeg(file, string_args=[("protocol_whitelist", "file,rtp,udp")])

Where the SDP file referenced above is as follows:

v=0
o=- 0 0 IN IP4 192.168.20.11
s=xxxx
c=IN IP4 192.168.10.25
t=0 0
m=audio 5004 RTP/AVP 103
a=rtpmap:103 L16/48000/2
a=recvonly

The working VBAN configuration is as follows:

input.external.rawaudio(
          restart=false,
          id=id,
          samplerate=48000,
          channels=2,
          max=10.0,
          buffer=.5,
          "vban_receptor -i #{interface} -p #{port} -s #{stream} -b file -d /dev/stdout")

I finally tried wrapping input.ffmpeg around a buffer.adaptative as follows, however it made no difference:

mksafe(buffer.adaptative(buffer=1., max=4., fallible=true, source_root))

It's worth mentioning that I do occasionally get this in the logs, so this could be an ffmpeg / source issue:

[sdp @ 0x7f6f1c635000] max delay reached. need to consume packet
[sdp @ 0x7f6f1c635000] RTP: missed 1 packets

This has as of yet, caused any noticeable issues, but I'm concerned what might happen long term.

Thoughts?