Janus "crashes" because of libnice segfault

[Dennis H]

Hey guys,

I'm using Janus with the latest 0.1.18 release of libnice. We are using the video plugin and have about 200 concurrent users online. I've noticed, that the Janus process crashes after 1-2 weeks because of the following error:

Sep 17 02:07:30 ns541136 kernel: [139206.554923] hloop 820286986[32263]: segfault at 40 ip 00007f7d75ccd090 sp 00007f7c15fa23e8 error 4 in libnice.so.10.11.0[7f7d75ca0000+50000]

Sep 17 02:07:31 ns541136 systemd[1]: janus.service: Main process exited, code=killed, status=11/SEGV

Sep 17 02:07:31 ns541136 systemd[1]: janus.service: Failed with result 'signal'.

Sep 17 02:07:46 ns541136 systemd[1]: janus.service: Service hold-off time over, scheduling restart.

Sep 17 02:07:46 ns541136 systemd[1]: janus.service: Scheduled restart job, restart counter is at 5.

Sep 17 02:07:46 ns541136 systemd[1]: Stopped Janus WebRTC Server.

Sep 17 02:07:46 ns541136 systemd[1]: Started Janus WebRTC Server.

Sep 17 02:07:46 ns541136 janus[32435]: Janus commit: ece07002322b66463c615aad8781a85b148fd22e

Sep 17 02:07:46 ns541136 janus[32435]: Compiled on:  Mon Jul 26 02:52:29 CEST 2021

The problem is, that Janus doesn't restart correctly. One process is running, but Janus will not handle any connections anymore. 

Is this issue known? The big problem is currently that we only can recognize if the service has crashes, if we check how many users are online (0 if crashed). If I do a service janus status in the command line, it shows, that the service is running.

[Alessandro Toppi]

First of all, collect a full stack of the crash and share it in a pastebin.

The problem you have with the restart seems not related to janus though but maybe it due to some frontend that is unable to recover from a janus restart.

[Alessandro Toppi]

You are also using a version of janus from last July. Please try a recent one.