Hi,
Loaded TrueNAS Scale on a box that I'm using for backups and would like to use for Channels as well. I've not dug into the "right" way to get it to work, but I've got it going simply using docker-compose. - Though note that the "runtime" option does not seem to be supported on the version of docker, but running "docker info" shows that the default runtime is nvidia, so I would think that should work.
Here's my compose file (created from looking at another post on the Channels forum):
version: '3.3'
services:
channels-dvr:
image: fancybits/channels-dvr:nvidia
container_name: channels-dvr
network_mode: host
environment:
- TZ=America/New_York
- NVIDIA_VISIBILE_DEVICES=all
- NVIDIA_DRIVER_CAPABILITIES=compute,utility,video
ports:
- "8089:8089"
restart: unless-stopped
devices:
- /dev/dri:/dev/dri
volumes:
- /mnt/Rabbit/appdata/Channels:/channels-dvr
- /mnt/Turtle/media/Channels:/shares/DVR
This works well, except for perhaps a couple of things. When I start the container, I get this at the CLI:
truenas# docker-compose up -d
Creating channels-dvr ... done
truenas# 2021 Mar 28 17:12:50 truenas Process 3205400 (ffmpeg) of user 0 dumped core.
Stack trace of thread 35:
#0 0x0000000001c24931 n/a (/channels-dvr/2021.03.16.1916/ffmpeg + 0x1824931)
#1 0x0000000000c2710e n/a (/channels-dvr/2021.03.16.1916/ffmpeg + 0x82710e)
#2 0x0000000000000038 n/a (n/a + 0x0)
2021 Mar 28 17:12:50 truenas Process 3205455 (ffmpeg) of user 0 dumped core.
Stack trace of thread 36:
#0 0x0000000001c24931 n/a (/channels-dvr/2021.03.16.1916/ffmpeg + 0x1824931)
#1 0x0000000000c2710e n/a (/channels-dvr/2021.03.16.1916/ffmpeg + 0x82710e)
#2 0x0000000000000038 n/a (n/a + 0x0)
Aside from the above errors, the GUI works, I was able to restore my DB, and all looks fine, except it only does software transcoding. I can connect to it from clients locally and it seems to work great.
If I run the curl -XPOST line to check what it finds for hardware support, I get two more ffmpeg core dumps, and no useful data.
Also, I got the same core dumps before I set the docker image to use the nvidia tag, so perhaps the nvidia tag isn't needed anymore?
Anyhow, I know I'm running on an Alpha version, so I'm not expecting much, but thought I would ask.