Closed jasonpstokes closed 3 years ago
ozett
commented
3 years ago looks like the same error at my testing of 09.x https://github.com/blakeblackshear/frigate/discussions/1502#discussioncomment-1204555
(your docker-compose shows that your are using "libva_driver" for decoding rtsp? cool)
sinamics
commented
3 years ago Can confirm same behavior on my dev pc. All clips from yesterday after 21:00 does not work for some reason. I get cors error in browser. All clips from today works.
Access to XMLHttpRequest at 'http://localhost:5000/vod/2021-08/19/21/picamera/index.m3u8' from origin 'http://localhost:8080' has been blocked by CORS policy: No 'Access-Control-Allow-Origin' header is present on the requested resource.
VIDEOJS: ERROR: (CODE:2 MEDIA_ERR_NETWORK) HLS playlist request error at URL: http://localhost:5000/vod/2021-08/19/21/picamera/index.m3u8.
MediaError {code: 2, status: 0, message: "HLS playlist request error at URL: http://localhost:5000/vod/2021-08/19/21/picamera/index.m3u8.", responseText: ""}
GET http://localhost:5000/vod/2021-08/19/21/picamera/index.m3u8 net::ERR_FAILED
jasonpstokes
commented
3 years ago (your docker-compose shows that your are using "libva_driver" for decoding rtsp? cool)
I'm using rtmp, but TBH I just copied that from the docs when I started with Frigate (without a Coral), trying to get the best setup for my AMD. But I didn't know what I was doing and never got far with the mesa-va-drivers on Debian, to know how to best use it or if it was even helping. When my Coral arrived and my CPU was no longer an issue, I moved on pretty quickly! :-)
jasonpstokes
commented
3 years ago An update: switching to the MQ (Balanced) stream for detect has made a big improvement for me. This is most evident at night, if (re)starting Frigate when the cameras are in B&W mode: using the LQ (fluent) stream for detect, Frigate would show green screens for up to a minute on several (random) cameras and log lots of broken frame errors - whereas the MQ streams show immediately!
I'm also using http streams as MarkGodwin suggested (here) and, whilst it's only been 4 days, overall these two changes have made a big improvement. Far fewer broken frames messages and the camera recordings work.
However, I am now seeing the detection process get stuck a few times a day - anyone else seen this ? EDIT: Switched to 0.8 (running the same docker and Frigate config) and within an hour got the same error.
[2021-09-01 22:05:16] ws4py INFO : Terminating websocket [Local => 127.0.0.1:5002 | Remote => 127.0.0.1:46980],
[2021-09-02 06:01:10] frigate.watchdog INFO : Detection appears to be stuck. Restarting detection process...,
[2021-09-02 06:01:10] root INFO : Waiting for detection process to exit gracefully...,
[2021-09-02 06:01:40] root INFO : Detection process didnt exit. Force killing...,
[2021-09-02 06:01:40] detector.coral INFO : Starting detection process: 32013,
[2021-09-02 06:01:40] frigate.edgetpu INFO : Attempting to load TPU as usb,
[2021-09-02 06:01:43] frigate.edgetpu INFO : TPU found,
[2021-09-02 07:23:00] frigate.watchdog INFO : Detection appears to be stuck. Restarting detection process...,
[2021-09-02 07:23:00] root INFO : Waiting for detection process to exit gracefully...,
[2021-09-02 07:23:30] root INFO : Detection process didnt exit. Force killing...,
[2021-09-02 07:23:30] detector.coral INFO : Starting detection process: 35948,
[2021-09-02 07:23:30] frigate.edgetpu INFO : Attempting to load TPU as usb,
[2021-09-02 07:23:33] frigate.edgetpu INFO : TPU found,
[2021-09-02 10:02:25] ws4py INFO : Managing websocket [Local => 127.0.0.1:5002 | Remote => 127.0.0.1:53556],Here's my 0.9 rc2 config if anyone else with Reolink cameras wants to try this-
config.yml
database:
path: /media/frigate/database/frigate.db
detectors:
coral:
type: edgetpu
device: usb
mqtt:
host: 192.168.1.5
port: 1883
user: mqtt
password: "{FRIGATE_MQTT_PASSWORD}"
birdseye:
mode: continuous
width: 1792
height: 1344
detect:
width: 896
height: 672
fps: 5
ffmpeg:
input_args:
- -avoid_negative_ts
- make_zero
- -fflags
- nobuffer
- -fflags
- +genpts+discardcorrupt
- -flags
- low_delay
- -strict
- experimental
- -rw_timeout
- '5000000'
- -f
- live_flv
motion:
mask:
- 265,650,265,625,635,625,635,650 # timestamp - mine is set to bottom centre in the cameras
record:
enabled: true
retain_days: 1
events:
enabled: true
pre_capture: 3
retain:
default: 15
snapshots:
enabled: true
clean_copy: false
bounding_box: true
retain:
default: 15
cameras:
Driveway:
ffmpeg:
inputs:
- path: http://192.168.1.71/flv?port=1935&app=bcs&stream=channel0_ext.bcs&user={FRIGATE_USER}&password={FRIGATE_PASSWORD}
roles:
- detect
- path: http://192.168.1.71/flv?port=1935&app=bcs&stream=channel0_main.bcs&user={FRIGATE_USER}&password={FRIGATE_PASSWORD}
roles:
- record
objects:
track:
- person
- vehicle
- dog
rtmp:
enabled: false
timestamp_style:
format: "%d/%m/%Y %H:%M:%S"
Entrance:
ffmpeg:
inputs:
- path: http://192.168.1.72/flv?port=1935&app=bcs&stream=channel0_ext.bcs&user={FRIGATE_USER}&password={FRIGATE_PASSWORD}
roles:
- detect
- path: http://192.168.1.72/flv?port=1935&app=bcs&stream=channel0_main.bcs&user={FRIGATE_USER}&password={FRIGATE_PASSWORD}
roles:
- record
objects:
track:
- person
- dog
rtmp:
enabled: false
timestamp_style:
format: "%d/%m/%Y %H:%M:%S"docker-compose.yml
services:
frigate:
container_name: frigate
image: blakeblackshear/frigate:0.9.0-rc2-amd64
hostname: frigate
restart: always
shm_size: '300m'
devices:
- /dev/bus/usb:/dev/bus/usb
volumes:
- /etc/localtime:/etc/localtime:ro
- /usr/docker/frigate/config.yml:/config/config.yml:ro
- /usr/docker/frigate/labelmap.txt:/labelmap.txt:ro
- /usr/docker/frigate/database:/media/frigate/database
- /mnt/nas_storage/frigate:/media/frigate
- type: tmpfs
target: /tmp/cache
tmpfs:
size: 2000000000
network_mode: bridge
ports:
- 5000:5000
environment:
FRIGATE_MQTT_PASSWORD: " "
FRIGATE_USER: " "
FRIGATE_PASSWORD: " "Running with http has made frigate a lot more stable for me.
Two of my six RLC-520A cameras throw errors every 10-20 seconds soon after startup, in both v8 and 9. Whilst I didn't notice any issues within Frigate in 8, in 9 these two cameras won't play anything from the Recordings interface - but will play the same Events from the Events interface. The other 4x cameras play Recordings and Events as expected.
Anyone seen this or have any ideas please?
All cameras have the same camera firmware and stream settings. I reset one of these cameras again this morning but no difference. Only Frigate config diff is that these two cameras also track 'vehicle' from my custom labelmap - but (not surprisingly) removing this didn't stop the error.
v9 screenshot:
Frigate is amazing, thanks for all your work on it!
Versions of frigate 0.8.4-5043040 0.9.0-3340952
Config file Driveway throws error, Entrance does not. Other four cameras repeat this config, except for some object tracking differences
labelmap.txt
Frigate container logs These errors are repeated every 10-20 seconds
after a reboot of Frigate
Frigate stats
Computer Hardware
docker-compose.yml