RTSP timeout

[turtledude01]

Which version are you using?

1.15.3

Which operating system are you using?

Linux amd64 standard

Describe the issue

This may be a "non issue" but a common use of MediaMTX is in VRChat for streams.

Specifically when using VRChat, MediaMtx v 1.15.1 seems to work perfectly (Though with some OBS config options that seem to be fixed in 1.15.2+)

However, on v1.15.2 and 1.15.3 we get the following type of behavior:

2025/10/24 18:27:01 INF [RTSP] [conn <Client IP>:65044] opened
2025/10/24 18:27:01 INF [RTSP] [session 95450350] created by <Client IP>:65044
2025/10/24 18:27:01 INF [RTSP] [session 95450350] is reading from path 'test', with TCP, 2 tracks (H264, MPEG-4 Audio)
2025/10/24 18:28:01 INF [RTSP] [conn 82.66.60.176:65044] closed: read tcp <Server IP>:554-><Client IP>:65044: i/o timeout
2025/10/24 18:28:01 INF [RTSP] [session 95450350] destroyed: not in use
2025/10/24 18:28:07 INF [RTSP] [conn <Client IP>:65058] opened
2025/10/24 18:28:07 INF [RTSP] [session 6ffb4409] created by <Client IP>:65058
2025/10/24 18:28:07 INF [RTSP] [session 6ffb4409] is reading from path 'test', with TCP, 2 tracks (H264, MPEG-4 Audio)

This happens consistently every minute, increasing the
readTimeout or writeTimeout to anything (I even tried 1000s on each) does not change the behavior.

As it works on a previous version, and broke specifically on 1.15.2, I must conclude it is something in a change between 1.15.1 and 1.15.2

Describe how to replicate the issue

1. start the MediaMTX
2. publish with rtmp
3. read with rtsp(t) (from within a VRChat video player)
4. Wait 1 minute

MediaMTX configuration

No response

MediaMTX logs

No response

Network dump

No response

[turtledude01]

Through more investigation, this appears to be due to how VRChat (Windows Media) handles the 'timeout'
Through debug, I can see the 'Keep alive' coming through, but it only comes through once every minute, and every so often there appears to be a desync with when VRChat sends the keep-alive, (By less than a second) which in turn causes MediaMTX to (Correctly) drop the connection. A workaround that would be nice for MediaMTX to include would be to add a 'Keep alive variance' type of option, IE: allow a 5 second additional window past the keep alive.

I am unsure about the spec of RTSP if it says to only send the keep-alive at the exact moment the server has said to send it, as this could be either MediaMTX not following spec, or VRChat not following spec.

EDIT: This may affect ANYTHING using WMF as that is where the true issue is.

[aler9]

duplicate of #5090

[bluenviron-bot]

This issue is being locked automatically because it has been closed for more than 6 months.
Please open a new issue in case you encounter a similar problem.