-
-
Notifications
You must be signed in to change notification settings - Fork 1.8k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[Support]: > 0.11-Beta-2 FFmpeg crashes after connection lost #3237
Comments
So restarting the Pi fixes this? ffmpeg complains of not being able to reach the camera which implies some sort of network / camera connection issue. |
Not the Docker-Compose restart, but the RPi restart. |
Okay, if docker compose restart doesn't fix it then that supports the idea that it is a network connection issue. Seems the rpi needs to get a fresh connection for some reason |
Ummm.. I'm not into the code very far - just played a bit with config and ffmpeg parameters here and there. But I guess normally TCP is used and UDP has to be forced - so I guess in my case TCP has to re-establish or get (better) a new connection. A working UDP solution making much more sense in many cases I guess (May I find some time in the coming days to test). |
This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions. |
Now in 0.11-Beta7:
@NickM-27 Still something went wrong reinitialising, the last error-sequence looped endless |
Not sure, I've not been able to reproduce this so it may also have to do with the camera as well. This shows it's not related to a specific ffmpeg version at least |
It looks like the hardware decoder get stuck in a bad state and can't recover. If restarting frigate's docker container doesn't resolve it, then I am not sure what to do about this. |
Maybe check gpu memory config for the Pi? |
Yea - same toughts here. Also there is no known reason why the input from the ip-cam (cable) is interrupted. I've the |
IDK if it make sense in RC to upgrade FFmpeg, but there have been an update recently: https://ffmpeg.org/index.html#news to Version 5.1. |
I had this exact same issue and it went away as soon as I disabled my hardware acceleration. I've been trying for three days now to get it working correctly. I'm on Ubuntu 21.04 LTS and the camera in question is a Raspberry Pi 4 running 32-bit Bullseye with the v2 camera module. I'm using libcamera, not the legacy option. |
Are you running 0.11 as well? |
I'm on 0.11.0beta7 It is worth mentioning that my pi is running an mjpg stream, and I'm using Nvidia hardware acceleration. The weird thing is, I had it working for a little bit. However, I can't get |
@NickM-27 Since last error updated the config.txt. So you may can explain whats happening now. Sometimes the TPU gets dected sometimes not? I've a cron-job for autoupdate and restart daily at 2:30 AM.
|
If you restart your host and you have the USB coral, it takes some time once booted for it to initialize its internal driver and report as the Google TPU (vs generic device). Typically when I restarted my system with the USB coral the first time frigate would fail due to no TPU and then wait 1 minute and try again it would work as expected and continue to work. |
Yes - but there also without restarting (random) issues/errors: detection appears to be stuck? Thanks for that info - but thats not the problem I guess. I mean I can test gracefull stop and start, if u think it can help. |
So I gathered more error logs: Seems it's Ffmpeg specific - but it's just easy said. So what I've gathered in the container:
Same event in docker logs (bevor has happend a lot, after this specific nothing more):
One hour before the madness begins:
Part 2:
...
|
@blakeblackshear In the above logs it says that RTSP "jitter buffer" is set to zero - may possible to increase to 500 or something? I think it's more like a workarround: https://ffmpeg.org/pipermail/ffmpeg-cvslog/2015-September/093812.html |
After some research I'm now sure it comes from the driver-implemention and has nothing to do with Frigate-Project. Basically there is currently a voltage handling issue which can implic error in the h264-handling which is ffmpeg using beside a lot of other minor error scenarios. So trying now with |
Describe the problem you are having
After restarting network (but not the RPi4B) FFmpeg is chrashing and not fixing itself. First issue probably on "sunrise". See Logs.
Version
0.11.0-d2c3cdc
Frigate config file
Relevant log output
FFprobe output from your camera
Frigate stats
No response
Operating system
Other
Install method
Docker Compose
Coral version
CPU (no coral)
Network connection
Wired
Camera make and model
HiLook IPC-D120
Any other information that may be helpful
Comlete Web-UI is "broken".
The text was updated successfully, but these errors were encountered: