Calls not working on updated instances past v2.2.22
-
I have two Mirotalks and both stopped working after upgrading past v2.2.22. I'm guessing some changes in this commit didn't work well on migrated instances. Participants can see other people join, but the loading circles never disappear, so no video and/or audio.
However, copying the
.env
settings, installing a fresh Mirotalk and pasting those settings worked out in the end, but it's not great if the rolling updates crash existing instances. -
Hello @msbt
Thanks for reporting. Under normal circumstances this should not cause any issues.
But I have reproduced the issue.
Installing version 2.2.22 and upgrading to 2.3.1 not touching anything leads to a meeting where the feed is just loading endlessly.After changing the ports to
25000
in the location settings it worked again.
I'm currently investigating why the old default of40000
is now leading to issues. -
TL;DR
Changing the port from
40000
to25000
in the location view after the update to2.3.1
fixes this issue.
Some debugging details
It seems with the new defaults of
25000
the old settings of40000
is now causing ICE connection failures.
This can be viewed, when using Chromium withchrome://webrtc-internals
and with Firefoxabout:webrtc
.Chromium view:
I can see the packages arriving on the Linux system with
tcpdump portrange 40000-401000-i eth0 and udp
andtcpdump portrange 25000-25100 and udp -i br-e1aa4ec2a833
After changing the port to
25000
in version2.3.1
I can see my desktop and mobile device in thetcpdump
and the ICE connection is successful for the ICE channel for audio and video.eth0
interface:14:21:48.079032 IP 61.8.145.254.10753 > my.cloudron.dev.25023: UDP, length 1171 14:21:48.079032 IP 61.8.145.254.10753 > my.cloudron.dev.25023: UDP, length 1154 14:21:48.079032 IP 61.8.145.254.10753 > my.cloudron.dev.25023: UDP, length 1154 14:21:48.079033 IP 61.8.145.254.10753 > my.cloudron.dev.25023: UDP, length 1154 14:21:48.079094 IP 61.8.145.254.10753 > my.cloudron.dev.25023: UDP, length 1154 14:21:48.079094 IP 61.8.145.254.10753 > my.cloudron.dev.25023: UDP, length 1154 14:21:48.079201 IP my.cloudron.dev.25099 > ip5f5ae97c.dynamic.kabel-deutschland.de.3826: UDP, length 1162 14:21:48.079256 IP my.cloudron.dev.25099 > ip5f5ae97c.dynamic.kabel-deutschland.de.3826: UDP, length 1162 14:21:48.079291 IP my.cloudron.dev.25099 > ip5f5ae97c.dynamic.kabel-deutschland.de.3826: UDP, length 1162 14:21:48.079326 IP my.cloudron.dev.25099 > ip5f5ae97c.dynamic.kabel-deutschland.de.3826: UDP, length 1162 14:21:48.079363 IP my.cloudron.dev.25099 > ip5f5ae97c.dynamic.kabel-deutschland.de.3826: UDP, length 1162
and the docker bridge interface
br-e1aa4ec2a833
14:21:46.994191 IP ip5f5ae97c.dynamic.kabel-deutschland.de.3824 > 172.18.16.195.25057: UDP, length 1130 14:21:46.994199 IP ip5f5ae97c.dynamic.kabel-deutschland.de.3824 > 172.18.16.195.25057: UDP, length 1130 14:21:46.994203 IP ip5f5ae97c.dynamic.kabel-deutschland.de.3824 > 172.18.16.195.25057: UDP, length 1130 14:21:46.994230 IP ip5f5ae97c.dynamic.kabel-deutschland.de.3824 > 172.18.16.195.25057: UDP, length 1130 14:21:46.994234 IP ip5f5ae97c.dynamic.kabel-deutschland.de.3824 > 172.18.16.195.25057: UDP, length 1130 14:21:46.994326 IP 172.18.16.195.25061 > 61.8.145.254.10763: UDP, length 1138 14:21:46.994395 IP 172.18.16.195.25061 > 61.8.145.254.10763: UDP, length 1138 14:21:46.994434 IP 172.18.16.195.25061 > 61.8.145.254.10763: UDP, length 1138 14:21:46.994472 IP 172.18.16.195.25061 > 61.8.145.254.10763: UDP, length 1138 14:21:46.994508 IP 172.18.16.195.25061 > 61.8.145.254.10763: UDP, length 1138
What is noticeable that the amount of "hops" when comparing
40000
and25000
is less.
Sadly, I lack the knowledge as of now to debug this any further than this.
Maybe @mirotalk-57bab571 has some spontaneous insights?
Also while looking into it.
It seems MiroTalk SFU can not be configured to use STUN/TURN, right? -
TL;DR
Changing the port from
40000
to25000
in the location view after the update to2.3.1
fixes this issue.
Some debugging details
It seems with the new defaults of
25000
the old settings of40000
is now causing ICE connection failures.
This can be viewed, when using Chromium withchrome://webrtc-internals
and with Firefoxabout:webrtc
.Chromium view:
I can see the packages arriving on the Linux system with
tcpdump portrange 40000-401000-i eth0 and udp
andtcpdump portrange 25000-25100 and udp -i br-e1aa4ec2a833
After changing the port to
25000
in version2.3.1
I can see my desktop and mobile device in thetcpdump
and the ICE connection is successful for the ICE channel for audio and video.eth0
interface:14:21:48.079032 IP 61.8.145.254.10753 > my.cloudron.dev.25023: UDP, length 1171 14:21:48.079032 IP 61.8.145.254.10753 > my.cloudron.dev.25023: UDP, length 1154 14:21:48.079032 IP 61.8.145.254.10753 > my.cloudron.dev.25023: UDP, length 1154 14:21:48.079033 IP 61.8.145.254.10753 > my.cloudron.dev.25023: UDP, length 1154 14:21:48.079094 IP 61.8.145.254.10753 > my.cloudron.dev.25023: UDP, length 1154 14:21:48.079094 IP 61.8.145.254.10753 > my.cloudron.dev.25023: UDP, length 1154 14:21:48.079201 IP my.cloudron.dev.25099 > ip5f5ae97c.dynamic.kabel-deutschland.de.3826: UDP, length 1162 14:21:48.079256 IP my.cloudron.dev.25099 > ip5f5ae97c.dynamic.kabel-deutschland.de.3826: UDP, length 1162 14:21:48.079291 IP my.cloudron.dev.25099 > ip5f5ae97c.dynamic.kabel-deutschland.de.3826: UDP, length 1162 14:21:48.079326 IP my.cloudron.dev.25099 > ip5f5ae97c.dynamic.kabel-deutschland.de.3826: UDP, length 1162 14:21:48.079363 IP my.cloudron.dev.25099 > ip5f5ae97c.dynamic.kabel-deutschland.de.3826: UDP, length 1162
and the docker bridge interface
br-e1aa4ec2a833
14:21:46.994191 IP ip5f5ae97c.dynamic.kabel-deutschland.de.3824 > 172.18.16.195.25057: UDP, length 1130 14:21:46.994199 IP ip5f5ae97c.dynamic.kabel-deutschland.de.3824 > 172.18.16.195.25057: UDP, length 1130 14:21:46.994203 IP ip5f5ae97c.dynamic.kabel-deutschland.de.3824 > 172.18.16.195.25057: UDP, length 1130 14:21:46.994230 IP ip5f5ae97c.dynamic.kabel-deutschland.de.3824 > 172.18.16.195.25057: UDP, length 1130 14:21:46.994234 IP ip5f5ae97c.dynamic.kabel-deutschland.de.3824 > 172.18.16.195.25057: UDP, length 1130 14:21:46.994326 IP 172.18.16.195.25061 > 61.8.145.254.10763: UDP, length 1138 14:21:46.994395 IP 172.18.16.195.25061 > 61.8.145.254.10763: UDP, length 1138 14:21:46.994434 IP 172.18.16.195.25061 > 61.8.145.254.10763: UDP, length 1138 14:21:46.994472 IP 172.18.16.195.25061 > 61.8.145.254.10763: UDP, length 1138 14:21:46.994508 IP 172.18.16.195.25061 > 61.8.145.254.10763: UDP, length 1138
What is noticeable that the amount of "hops" when comparing
40000
and25000
is less.
Sadly, I lack the knowledge as of now to debug this any further than this.
Maybe @mirotalk-57bab571 has some spontaneous insights?
Also while looking into it.
It seems MiroTalk SFU can not be configured to use STUN/TURN, right?@james said in Calls not working on updated instances past v2.2.22:
It seems MiroTalk SFU can not be configured to use STUN/TURN, right?
Wasn't it previously configued to use the Clouron STUN/TURN? I'm not at all certain, but I thought that was the case
(otherwise what's the different between the SFU and P2P versions?)
-
TL;DR
Changing the port from
40000
to25000
in the location view after the update to2.3.1
fixes this issue.
Some debugging details
It seems with the new defaults of
25000
the old settings of40000
is now causing ICE connection failures.
This can be viewed, when using Chromium withchrome://webrtc-internals
and with Firefoxabout:webrtc
.Chromium view:
I can see the packages arriving on the Linux system with
tcpdump portrange 40000-401000-i eth0 and udp
andtcpdump portrange 25000-25100 and udp -i br-e1aa4ec2a833
After changing the port to
25000
in version2.3.1
I can see my desktop and mobile device in thetcpdump
and the ICE connection is successful for the ICE channel for audio and video.eth0
interface:14:21:48.079032 IP 61.8.145.254.10753 > my.cloudron.dev.25023: UDP, length 1171 14:21:48.079032 IP 61.8.145.254.10753 > my.cloudron.dev.25023: UDP, length 1154 14:21:48.079032 IP 61.8.145.254.10753 > my.cloudron.dev.25023: UDP, length 1154 14:21:48.079033 IP 61.8.145.254.10753 > my.cloudron.dev.25023: UDP, length 1154 14:21:48.079094 IP 61.8.145.254.10753 > my.cloudron.dev.25023: UDP, length 1154 14:21:48.079094 IP 61.8.145.254.10753 > my.cloudron.dev.25023: UDP, length 1154 14:21:48.079201 IP my.cloudron.dev.25099 > ip5f5ae97c.dynamic.kabel-deutschland.de.3826: UDP, length 1162 14:21:48.079256 IP my.cloudron.dev.25099 > ip5f5ae97c.dynamic.kabel-deutschland.de.3826: UDP, length 1162 14:21:48.079291 IP my.cloudron.dev.25099 > ip5f5ae97c.dynamic.kabel-deutschland.de.3826: UDP, length 1162 14:21:48.079326 IP my.cloudron.dev.25099 > ip5f5ae97c.dynamic.kabel-deutschland.de.3826: UDP, length 1162 14:21:48.079363 IP my.cloudron.dev.25099 > ip5f5ae97c.dynamic.kabel-deutschland.de.3826: UDP, length 1162
and the docker bridge interface
br-e1aa4ec2a833
14:21:46.994191 IP ip5f5ae97c.dynamic.kabel-deutschland.de.3824 > 172.18.16.195.25057: UDP, length 1130 14:21:46.994199 IP ip5f5ae97c.dynamic.kabel-deutschland.de.3824 > 172.18.16.195.25057: UDP, length 1130 14:21:46.994203 IP ip5f5ae97c.dynamic.kabel-deutschland.de.3824 > 172.18.16.195.25057: UDP, length 1130 14:21:46.994230 IP ip5f5ae97c.dynamic.kabel-deutschland.de.3824 > 172.18.16.195.25057: UDP, length 1130 14:21:46.994234 IP ip5f5ae97c.dynamic.kabel-deutschland.de.3824 > 172.18.16.195.25057: UDP, length 1130 14:21:46.994326 IP 172.18.16.195.25061 > 61.8.145.254.10763: UDP, length 1138 14:21:46.994395 IP 172.18.16.195.25061 > 61.8.145.254.10763: UDP, length 1138 14:21:46.994434 IP 172.18.16.195.25061 > 61.8.145.254.10763: UDP, length 1138 14:21:46.994472 IP 172.18.16.195.25061 > 61.8.145.254.10763: UDP, length 1138 14:21:46.994508 IP 172.18.16.195.25061 > 61.8.145.254.10763: UDP, length 1138
What is noticeable that the amount of "hops" when comparing
40000
and25000
is less.
Sadly, I lack the knowledge as of now to debug this any further than this.
Maybe @mirotalk-57bab571 has some spontaneous insights?
Also while looking into it.
It seems MiroTalk SFU can not be configured to use STUN/TURN, right? -
@james said in Calls not working on updated instances past v2.2.22:
It seems MiroTalk SFU can not be configured to use STUN/TURN, right?
Wasn't it previously configued to use the Clouron STUN/TURN? I'm not at all certain, but I thought that was the case
(otherwise what's the different between the SFU and P2P versions?)
@jdaviescoates said in Calls not working on updated instances past v2.2.22:
Wasn't it previously configued to use the Clouron STUN/TURN?
The MiroTalk SFU App had the turn/stun add-on in its manifest.
@nostrdev said in Participants can't see eachother?:
It appears the turn server is no longer used as an addon in SFU - could this explain why the app just does not work for many of our participants for several months now?
https://git.cloudron.io/packages/mirotalksfu-app/-/commit/05013ccfb0cbd9342433783e8ffe7c3fb14fa600
But it seems MiroTalk SFU itself has no option to configure TURN/STUN.
-
otherwise what's the different between the SFU and P2P versions?
-
MiroTalk P2P → Works without a central media server. All peers connect directly to each other. For this reason, it does need STUN/TURN configuration. STUN helps peers discover their public IP/port, and TURN relays media when direct connections aren’t possible (e.g. behind strict NAT/firewalls).
-
MiroTalk SFU → Works through a Selective Forwarding Unit (based on mediasoup). Here, all peers send their media streams to the SFU, which then forwards them to others. Because every client only needs to connect to the SFU, it doesn’t require TURN in most cases. The SFU is the “meeting point,” so peers don’t have to negotiate direct NAT traversal with each other. STUN/TURN isn’t part of its configuration by default, since mediasoup handles the media routing internally.
But it seems MiroTalk SFU itself has no option to configure TURN/STUN.
- P2P → needs STUN/TURN for NAT traversal.
- SFU → doesn’t usually require STUN/TURN, since the media server acts as the central hub for all streams. However, you must ensure that the necessary port ranges are open and not blocked by firewalls or the ISP.
-
-
Hello @mirotalk-57bab571
Thanks for the clarification.