@MarcA711 commented on GitHub (Nov 18, 2023):
Apparently, the hardware presets were not the only issue. The problem still occurs. I will update the issue tomorrow with new log outputs etc.
@quantummobile commented on GitHub (Nov 18, 2023):
Okay I understand what is going on here. Your pre_capture is 120 seconds which is too long for the assumptions made by the recordings…
@NickM-27 commented on GitHub (Nov 18, 2023):
You can get a shell in the container with docker exec -it frigate /bin/bash then run ls /tmp/cache
@NickM-27 commented on GitHub (Nov 18, 2023):
That would indicate that ffmpeg is not receiving any data. I'm assuming it's the same when using the restream preset?
@NickM-27 commented on GitHub (Nov 18, 2023):
it would also be helpful to know at what time the event occurred that had missing recording segments
@MarcA711 commented on GitHub (Nov 18, 2023):
Yes, the restream preset doesn't work as well. This is confusing, since I can save the go2rtc stream to a mp4 file on my host machine that uses the…
@NickM-27 commented on GitHub (Nov 18, 2023):
Yes if you open http://frigate:5000/api/config you will find the ffmpeg_cmds
@NickM-27 commented on GitHub (Nov 18, 2023):
by the way, it looks like you kept the old motion detection values despite the release notes explaining motion detection has changed and these old…
@MarcA711 commented on GitHub (Nov 18, 2023):
Ah, I see the problem. My cam streams h265. I want to record h264. So, I transcode it using go2rtc.
Now, I used the hardware accel preset…
@quantummobile commented on GitHub (Nov 18, 2023):
You can get a shell in the container with
docker exec -it frigate /bin/bashthen runls /tmp/cache
ls /tmp/cache mainhallway-e1pro@2023…
@MarcA711 commented on GitHub (Nov 18, 2023):
I looked it up in the docs, I can set the hwaccel args for each input. This would solve the problem, I will try it out later.
@quantummobile commented on GitHub (Nov 18, 2023):
by the way, it looks like you kept the old motion detection values despite the release notes explaining motion detection has changed and these…
@NickM-27 commented on GitHub (Nov 18, 2023):
Okay I understand what is going on here. Your pre_capture is 120 seconds which is too long for the assumptions made by the recordings maintainer.…
@NickM-27 commented on GitHub (Nov 18, 2023):
There's no particular reason, just that the record and detect can be done in the same ffmpeg process so it didn't make sense to force it out for…
@quantummobile commented on GitHub (Nov 18, 2023):
While this is happening it will be good to know the contents of /tmp/cache
How would I get these?
@NickM-27 commented on GitHub (Nov 18, 2023):
Do you see anything being created in /tmp/cache ?
@MarcA711 commented on GitHub (Nov 18, 2023):
No, the folder is empty.
@quantummobile commented on GitHub (Nov 18, 2023):
Also there's no reason not to use a coral, it would only help
I am using coral on another machine, works like a charm, though the same…
@MarcA711 commented on GitHub (Nov 18, 2023):
This is the full log after some minutes:
@NickM-27 commented on GitHub (Nov 18, 2023):
Also there's no reason not to use a coral, it would only help