Hardware transcoding issue

Plex Media Server
292

In the container, does PMS run as root also?

If not, let’s try: chmod 666 /dev/dri/*

This opens up permission on the device temporarily to everyone in that container

293

Same issue.

image
image1116×601 27.1 KB

Plex Media Server Logs_2023-03-16_07-53-03.zip (1.4 MB)

294

To confirm? You have two RTX cards in the machine?

295

Yes. RTX 8000

296

Now for the dum-dum question.

You can or cannot setup a quick & dirty native install on the host?
(remove the container from the list of variables)

297

@Vicerak

How skilled are you with invoking PMS in the Debugger ?

Asking because the team seeing something really whacky going on and has asked if you can set a breakpoint on std::invalid_argument::invalid_argument

This type stuff mean anything to you?

298

You mean a bare metal installation directly to the host? If that is so, i already have one and it works like a charm.

The problem is with the container .

The reason I have two Plex installations is because i want to differs the channels in two “two” groups.

I wish Plex would provide the option to “tag” them as a group but that is another whole new topic :joy:

299

Nope.

My setup is basically plex server running on docker with Portainer and xteve

300

What’s the base OS?

301

The host is ubuntu 22.10

302

PMS on Ubuntu is trivial. (Linux PMS is based on Ubuntu)

Can I get you to standup a test-only server?

The purpose of that test server, which goes away as soon as test complete, is to confirm PMS, on the base OS, using the base OS drivers, can talk to the cards

303

Of course! how that would work? What are the next steps?

304

The next steps are straight forward:

  1. You have the Nvidia drivers already installed on the host – Nothing additional required.

  2. Download PMS from https://plex.tv/downloads (Debian/Ubuntu 64 bit)

  3. Stop the Docker container on the Ubuntu host

  4. Install the DEB file you just downloaded (sudo dpkg -i filename.deb)

  5. You’ll see the installer run and install PMS on the host.

  6. When it’s done, it will automatically start PMS for you too.

  7. Now open open the browser, open a new incognito window

  8. In that window, open http://127.0.0.1:32400/web

  9. You’ll be setting up a new server.
    – Give it only 1 library section (small one we can use to confirm HW transcoding)
    – Give it a FRIENDLY NAME of “test” or anything else not to conflict with your production server’s name (the hostname)
    – Remember to go into the Server settings → Transcoder - Show Advanced
    – ENABLE all the HW transcoding options (by default one is not)
    – SAVE

  10. Test

305

That’s the thing. I already have a PMS installed on host, running 1.31.3.6819. It works completely fine.

Let me explain my setup a little bit.

I have Hardware Server that contains two PMS:

   -  "US" - contains local libraries such as movies and TV shows.  It also contain US a xteve with  US Channels . This PMS server is installed on the host or as we usually say "bare metal"

image
image1183×406 40.3 KB

  - "Brazil" -contains Brazilian channels , also running with xteve.  This is the problematic PMS server. It is running on Docker.

image
image1426×351 42.6 KB

This is the server where I have been testing. Everything works with version 1.29.2.6364 but when I upgrade the version to the latest it fails with errors I already mentioned in the previous messages.

The main reason for two servers is because I like to “split” the channels into two tabs

image
image1057×588 73.7 KB

306

You do? (sorry for missing that point)

So this is entirely a container issue?

307

That’s correct.

308

And It works with 1.29.2.6364! So it is not my setup… I think

309

I use docker , not portainer. I can’t advise you further there.

I do know that docker works flawlessly when you add:

--device=/dev/dri:/dev/dri

you didn’t forget that … or forget to run the container in privileged mode?

310

This is my docker-compose.

version: “3.5”
services:
brazil:
image: plexinc/pms-docker:1.29.2.6364-6d72b0cf6
runtime: nvidia
restart: unless-stopped
container_name: brazil
environment:
- UID=1000
- GID=1000
#- PLEX_CLAIM=claim-sGEdXdErut1V3Esfnzsp
- NVIDIA_VISIBLE_DEVICES=all
- NVIDIA_DRIVER_CAPABILITIES=all
- TZ=America/Los_Angeles
networks:
plex_enp166s0:
ipv4_address: 192.168.96.2
volumes:
- /data/containers/plex/brazil:/config
- /tv:/tv
- /movies:/movies
- /data/transcode/brazil:/transcode
- /data/downloads:/downloads
devices:
- /dev/dri:/dev/dri

networks:
plex_enp166s0:
external: true

the container is NOT in privileged mode. Let me test and and I will let you know in a few minutes.

311

Nope. Same freaking error
Plex Media Server Logs_2023-03-16_16-01-30.zip (930.8 KB)