[HW Accel Support]: use cpu and gpu for recording from rtsp #14506
-
Describe the problem you are havingneed some streams to be processed by the GPU and others by the CPU. Can you help me figure out how to achieve this? In my case, even if I don’t specify "global_args" for a specific camera but simply pass the argument in docker-compose to enable the GPU, everything still gets processed on the GPU, and naturally, the GPU crashes with the error: "[2024-10-22 15:42:02] ffmpeg.cam64.detect ERROR: [AVHWDeviceContext @ 0x560bd16f80c0] cu->cuCtxCreate(&hwctx->cuda_ctx, desired_flags, hwctx->internal->cuda_device) failed -> CUDA_ERROR_OUT_OF_MEMORY: out of memory." I have around 60 cameras, but I’ve simplified the config for readability. Version0.14.1-f4f3cfa Frigate config filedetect:
enabled: false
database:
path: /config/frigate.db
record:
enabled: true
retain:
days: 14
logger:
logs:
frigate.record.maintainer: debug
cameras:
cam102:
ffmpeg:
inputs:
- path: rtsp://puk:pwd@10.3.33.102/axis-media/media.amp
input_args: preset-rtsp-restream
roles:
- record
cam103:
ffmpeg:
inputs:
- path: rtsp://puk:pwd@10.3.4.81/axis-media/media.amp
input_args: preset-rtsp-restream
roles:
- record
cam105:
ffmpeg:
global_args: -hide_banner -loglevel warning -threads 2 -hwaccel_device 0
inputs:
- path: rtsp://puk:pwd@10.3.33.105/axis-media/media.amp
input_args: preset-rtsp-restream
roles:
- recorddocker-compose file or Docker CLI commandversion: "3.9"
services:
frigate:
privileged: true # this may not be necessary for all setups
image: ghcr.io/blakeblackshear/frigate:0.14.1
deploy:
resources:
reservations:
devices:
- driver: nvidia
device_ids: ['0']
count: 1 # number of GPUs
capabilities: [gpu]
shm_size: "40000MB" # update for your cameras based on calculation above
volumes:
- /etc/localtime:/etc/localtime:ro
- ./config:/config
- ./storage:/media/frigate
- type: tmpfs # Optional: 1GB of memory, reduces SSD/SD Card wear
target: /tmp/cache
tmpfs:
size: 10000000000
devices:
- /dev/nvidia-caps:/dev/nvidia-caps
- /dev/nvidia0:/dev/nvidia0
- /dev/nvidiactl:/dev/nvidiactl
- /dev/nvidia-modeset:/dev/nvidia-modeset
- /dev/nvidia-uvm:/dev/nvidia-uvm
- /dev/nvidia-uvm-tools:/dev/nvidia-uvm-tools
ports:
- "5000:5000"
- "8554:8554" # RTSP feeds
- "8555:8555/tcp" # WebRTC over tcp
- "8555:8555/udp" # WebRTC over udp
environment:
FRIGATE_RTSP_PASSWORD: "S4fopae8"
networks:
- internet
restart: unless-stopped
networks:
internet:
external: trueRelevant Frigate log output"[2024-10-22 15:42:02] ffmpeg.cam105.detect ERROR: [AVHWDeviceContext @ 0x560bd16f80c0] cu->cuCtxCreate(&hwctx->cuda_ctx, desired_flags, hwctx->internal->cuda_device) failed -> CUDA_ERROR_OUT_OF_MEMORY: out of memory."
and for other camRelevant go2rtc log outputdon't usingFFprobe output from your cameraroot@b3eb51f3cceb:/opt/frigate# ffprobe rtsp://puk:pwd@10.3.33.105/axis-media/media.amp
ffprobe version n5.1-2-g915ef932a3-20220731 Copyright (c) 2007-2022 the FFmpeg developers
built with gcc 12.1.0 (crosstool-NG 1.25.0.55_3defb7b)
configuration: --prefix=/ffbuild/prefix --pkg-config-flags=--static --pkg-config=pkg-config --cross-prefix=x86_64-ffbuild-linux-gnu- --arch=x86_64 --target-os=linux --enable-gpl --enable-version3 --disable-debug --enable-iconv --enable-libxml2 --enable-zlib --enable-libfreetype --enable-libfribidi --enable-gmp --enable-lzma --enable-fontconfig --enable-libvorbis --enable-opencl --enable-libpulse --enable-libvmaf --enable-libxcb --enable-xlib --enable-amf --enable-libaom --enable-libaribb24 --enable-avisynth --enable-libdav1d --enable-libdavs2 --disable-libfdk-aac --enable-ffnvcodec --enable-cuda-llvm --enable-frei0r --enable-libgme --enable-libass --enable-libbluray --enable-libjxl --enable-libmp3lame --enable-libopus --enable-mbedtls --enable-librist --enable-libtheora --enable-libvpx --enable-libwebp --enable-lv2 --enable-libmfx --enable-libopencore-amrnb --enable-libopencore-amrwb --enable-libopenh264 --enable-libopenjpeg --enable-libopenmpt --enable-librav1e --enable-librubberband --disable-schannel --enable-sdl2 --enable-libsoxr --enable-libsrt --enable-libsvtav1 --enable-libtwolame --enable-libuavs3d --enable-libdrm --enable-vaapi --enable-libvidstab --enable-vulkan --enable-libshaderc --enable-libplacebo --enable-libx264 --enable-libx265 --enable-libxavs2 --enable-libxvid --enable-libzimg --enable-libzvbi --extra-cflags=-DLIBTWOLAME_STATIC --extra-cxxflags= --extra-ldflags=-pthread --extra-ldexeflags=-pie --extra-libs='-ldl -lgomp' --extra-version=20220731
libavutil 57. 28.100 / 57. 28.100
libavcodec 59. 37.100 / 59. 37.100
libavformat 59. 27.100 / 59. 27.100
libavdevice 59. 7.100 / 59. 7.100
libavfilter 8. 44.100 / 8. 44.100
libswscale 6. 7.100 / 6. 7.100
libswresample 4. 7.100 / 4. 7.100
libpostproc 56. 6.100 / 56. 6.100
Input #0, rtsp, from 'rtsp://look:S4fopae8@10.3.33.105/axis-media/media.amp':
Metadata:
title : Media Presentation
Duration: N/A, start: 0.040000, bitrate: N/A
Stream #0:0: Video: h264 (Main), yuvj420p(pc, progressive), 1920x1080 [SAR 1:1 DAR 16:9], 25 fps, 25 tbr, 90k tbnInstall methodDocker Compose Object DetectorOther Network connectionWired Camera make and modelhikvision Screenshots of the Frigate UI's System metrics pages
Any other information that may be helpfulNo response |
Beta Was this translation helpful? Give feedback.
Answered by
NickM-27
Oct 22, 2024
Replies: 1 comment 5 replies
-
|
set to disable hwaccel |
Beta Was this translation helpful? Give feedback.
5 replies
-
|
Thank you very much, that helped. Could you please tell me why I might be losing some frames? Here are my CPU and GPU loads: |
Beta Was this translation helpful? Give feedback.
-
|
what do you mean losing frames? |
Beta Was this translation helpful? Give feedback.
-
|
The network is fine, i checked this. |
Beta Was this translation helpful? Give feedback.
-
|
where are you seeing that you only see 54 minutes of the hour? I'd also suggest not using debug logs and pasting the full set of normal logs so we can see what is going on during an hour where you are seeing missing segments |
Beta Was this translation helpful? Give feedback.
-
|
I'm sorry for the delayed response, but it seems that distributing the load between the GPU and CPU has solved my issue with missing some recording fragments. |
Beta Was this translation helpful? Give feedback.
Answer selected by
mahchet
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
set
to disable hwaccel