[Plex] NVIDIA GPU Stopped Working

ohsnapword commented 5 months ago

chart Name

Plex

Operating System

TrueNAS SCALE 23.10.2

Deployment Method

TrueNAS SCALE charts

Chart Version

17.7.2

Kubernetes Events

2024-03-30 10:09:41
Created container plex
2024-03-30 10:09:41
Started container plex
2024-03-30 10:09:34
Add eth0 [172.16.0.153/16] from ix-net
2024-03-30 10:09:34
Add net1 [10.22.0.12/24] from ix-plex/ix-plex-0
2024-03-30 10:09:34
Container image "ghcr.io/onedr0p/plex:1.40.1.8227-c0dd5a73e@sha256:4ca8db0ffa1bee5a2976a0385cc43368112e81a2485b80d0fff3b04c9d5bd1ca" already present on machine
2024-03-30 10:09:33
Scaled up replica set plex-8686477569 to 1
2024-03-30 10:09:33
Created pod: plex-8686477569-47jm7
2024-03-30 10:09:33
Successfully assigned ix-plex/plex-8686477569-47jm7 to ix-truenas
2024-03-30 10:09:32
Scaled down replica set plex-5b68dcc4b7 to 0 from 1
2024-03-30 10:09:32
Deleted pod: plex-5b68dcc4b7-f67m5
2024-03-30 10:09:32
Stopping container plex
2024-03-30 09:59:18
Started container plex
2024-03-30 09:59:17
Created container plex
2024-03-30 09:59:11
Add eth0 [172.16.0.152/16] from ix-net
2024-03-30 09:59:11
Add net1 [10.22.0.12/24] from ix-plex/ix-plex-0
2024-03-30 09:59:11
Container image "ghcr.io/onedr0p/plex:1.40.1.8227-c0dd5a73e@sha256:4ca8db0ffa1bee5a2976a0385cc43368112e81a2485b80d0fff3b04c9d5bd1ca" already present on machine
2024-03-30 09:59:09
Scaled up replica set plex-5b68dcc4b7 to 1
2024-03-30 09:59:09
Created pod: plex-5b68dcc4b7-f67m5
2024-03-30 09:59:09
Successfully assigned ix-plex/plex-5b68dcc4b7-f67m5 to ix-truenas
2024-03-30 09:58:55
Readiness probe failed: Get "http://172.16.0.134:32400/identity": context deadline exceeded (Client.Timeout exceeded while awaiting headers)
2024-03-30 09:58:48
Scaled down replica set plex-fd7b594 to 0 from 1
2024-03-30 09:58:48
Deleted pod: plex-fd7b594-kftwz
2024-03-30 09:58:48
Stopping container plex

chartlication Logs

2024-03-30 10:09:41.404953-04:00Setting customConnections to: https://plex.domain.me
2024-03-30 10:09:41.420382-04:00Setting allowedNetworks to: 172.16.0.0/12,10.0.0.0/8,192.168.0.0/16
2024-03-30 10:09:54.599265-04:00Critical: libusb_init failed

Chart Configuration

Add NVIDIA GPUs (Experimental) = 1

Describe the bug

NVIDIA transcoding previously worked in Plex. Running nvidia-smi in the container shows the GPU, however when attempting to transcode, it crashes or doesn't use HW transcoding.

To Reproduce

Install latest Plex app.
Assign NVIDIA GPU

Expected Behavior

Plex uses HW transcoding.

Screenshots

N/A

Additional Context

GPU is a Quadro P620 which has worked flawlessly in previous versions.

HW transcoding seems to have stopped working 2/20/2024.

I see the following lines in the Plex console: [Req#94d/Transcode] [FFMPEG] - Cannot load libnvidia-encode.so.1 [Req#94d/Transcode] [FFMPEG] - The minimum required Nvidia driver for nvenc is 455.28 or newer [Req#94d/Transcode] [FFMPEG] - Cannot load libnvcuvid.so.1 [Req#94d/Transcode] [FFMPEG] - Failed loading nvcuvid. [Req#94d/Transcode] [FFMPEG] - Failed setup for format cuda: hwaccel initialisation returned error.

If I run nvidia-smi in the container, the version returned is 535.54.03.

I've read and agree with the following

[X] I've checked all open and closed issues and my issue is not there.
[X] I've prefixed my issue title with [Chart-Name]

ohsnapword commented 5 months ago

Additional info:

I checked the chart config on another instance and it seems that the GPU config changed from a dropdown to a text field. This instance has an Intel GPU. I changed the allocate Intel i915 GPU field to 1 from 0 and it fixed the HW transcoding on that machine.

ohsnapword commented 5 months ago

After updating the app to 17.7.5, NVIDIA GPU transcoding seems to be working.

truecharts / charts