ari_verse

I thought this was a boost issue, but even using lemm.ee's web site, when browsing to my Nikon subscription (https://lemm.ee/c/nikon@lemmy.world) I don't see any post. If I go to the same community from lemmy.world's own site, I can see recent posts: https://lemmy.world/c/nikon I noticed the same problem in other communities (but I thought the issue was my lemmy client). Can this be fixed?

If you use Reolink wifi cams via the Frigate integration in homeassistant, you may be used to seeing tons of "ffmpeg has crashed unexpectedly" in your frigate logs. I have 3 older reolink wifi cams in frigate (510WA, 511WA) which most in the community seems to advise against and indeed since I've been running frigate, while they worked, they have been problematic since Frigate 0.12 and the arrival of go2rtc

Cutting down to the chase, when I was using the standard common configuration with these cameras:

go2rtc:
  streams:
    driveway:
      - "http://192.168.x.y/flv?port=1935&app=bcs&stream=channel0_main.bcs&user=USER&password=PASSWORD#video=copy"
    driveway_sub:
      - "http://192.168.x.y/flv?port=1935&app=bcs&stream=channel0_sub.bcs&user=USER&password=PASSWORD"

... frigate would generally work with them however in the logs I could see that the ffmpeg process that frigate creates for each re-stream would crash every few minutes, accumulating thousands of crashes over time. I assume that the crashes would also cause event detection to be unavailable for a brief period each time they happened.

With Frigate 0.13, it got even worse, as with each crash, the HA dashboard would show a black image with a "No frames received" message that would only go away after a manual dashboard refresh.

I believe the issue is that go2rtc was unable to properly handle the streams from these old cams, while in the past, ffmpeg directly in frigate could do it, adding some ffmpeg parameter.

The solution I found a few days ago that ALMOST COMPLETELY eliminated all the "ffmpeg crashed unexpectedly" situations (I went from THOUSANDS of errors to just one or two errors after a few days) is to change the go2rtc configuration so that it uses ffmpeg instead of it's own code to connect to the cams:

go2rtc:
  streams:
    driveway:
      - "ffmpeg:http://192.168.x.y/flv?port=1935&app=bcs&stream=channel0_main.bcs&user=USER&password=PASSWORD#video=copy"
    driveway_sub:
      - "ffmpeg:http://192.168.x.y/flv?port=1935&app=bcs&stream=channel0_sub.bcs&user=USER&password=PASSWORD"

I also have the following global ffmpeg configuration in frigate's config file, not sure if it helps or not:

ffmpeg:
  global_args: -hide_banner -loglevel error
  hwaccel_args: preset-vaapi
  input_args: -avoid_negative_ts make_zero -fflags +genpts+discardcorrupt -flags low_delay -strict experimental -analyzeduration 1000M -probesize 1000M -rw_timeout 5000000

This go2rtc configuration decreased CPU utilization significantly (frigate 0.13 itself also helped apparently). Hoping this will help others in the same situation, as the cams themselves are not bad, it's just their software that sucks, but that can be worked around quite nicely with ffmpeg.

Years on the waiting and finally a replacement for my fanless Apollo lake home server seems to be on the way! After the disappointments of Jasper lane and prior announcements this one seems for real. 3-4 times better compute metrics and half the power.

Wondering what are others in the community using these days for always-on home server/media streaming needs?