GPU transcoding options not showing in TrueNAS Scale

[thekingpin 18]

Hello 

I am running into this issue again that the GPU is recognised by Emby but the options for hardware transcoding are not showing. 

This issue was previously fixed with Version 4.8.10.0 but appears to be happening again.

Just to be clear, Passthrough available (non-NVIDIA) is selected from TrueNAS.

Hardware logs attached:

 

hardware_detection-63867970936.txt

Edited November 23, 2024 by thekingpin
Added screenshot of issue.

[Brudertac 18]

Yeah, i see that to. (With other Errors after restoring a Backup)

[thekingpin 18]

OK, I have fixed the issue myself and posting here if anyone else has the issue. 

Im not saying this is the best or the only way to fix it. In my case, with a 1 user system this works for me.

In the logs I had error:

"Message": "Failed to open the drm device /dev/dri/renderD128"

So immediately I thought it must be permission issue. 

I just chomd 777 the whole /dev/dri directory and now emby can access it. 

Now the GPU is useable again and transcoding from Emby is possible 

[tollgate 6]

I understand the correct way to fix this is to make the emby user a  member of video and render groups. (with thanks to member Q-Droid)

[Brudertac 18]

1 hour ago, tollgate said:

I understand the correct way to fix this is to make the emby user a  member of video and render groups. (with thanks to member Q-Droid)

Not working here. I have added the "Apps" user (this is the Docker user on Truenas Scale) to Groups render and video. Still no Hardware Transcoding.

[thekingpin 18]

56 minutes ago, Brudertac said:

Not working here. I have added the "Apps" user (this is the Docker user on Truenas Scale) to Groups render and video. Still no Hardware Transcoding.

It's working for me. You don't need to add app user to same group as video/render but rather change the Group inside the container to match the one of video. 

 

You can find the group number by running the below command in shell

cat /etc/group

[thekingpin 18]

2 hours ago, tollgate said:

I understand the correct way to fix this is to make the emby user a  member of video and render groups. (with thanks to member Q-Droid)

Thank you very much this worked better for me because the permissions reset after a reboot. 

[Brudertac 18]

So in my Case i have UserID and GroupID 568 (Apps User and Group).

I tried to change the GroupID to 44 (Video Group) but this does not work.

(The thing is, under Jellyfin this works perfect with UserID/GroupID 568....)

Edited November 24, 2024 by Brudertac

[thekingpin 18]

12 minutes ago, Brudertac said:

So in my Case i have UserID and GroupID 568 (Apps User and Group).

I tried to change the GroupID to 44 (Video Group) but this does not work.

(The thing is, under Jellyfin this works perfect with UserID/GroupID 568....)

I added render group and not video, which in my case is 107. Can you try that?

Also what GPU are you using?

Edit: Can you also post your hardware logs from Emby.

Edited November 24, 2024 by thekingpin

[Brudertac 18]

Yeah, with 107 it works! Thank you so much. 

Oh, it is an Intel Arc 310 Card.

Edited November 24, 2024 by Brudertac

[thekingpin 18]

Great happy to hear that!

[DAVe3283 7]

I see the exact same issue unless I do `chmod 777 /dev/dri` inside the container then tell Emby Server to restart. It also loses that setting when I reboot the NAS.

If setting the group to 107 works, why is that not the default? I feel this is a bug, and making Emby run as the `render` group is a bad hack, not the fix. Or if it is the fix, shouldn't it be part of the instructions here? https://emby.media/truenas-scale-server.html

Though that already has the wrong URL on the guide (should be port 9096 not 8096) so I suspect it does not get updated much.

[Luke 39664]

On 12/2/2024 at 11:43 PM, DAVe3283 said:

I see the exact same issue unless I do `chmod 777 /dev/dri` inside the container then tell Emby Server to restart. It also loses that setting when I reboot the NAS.

If setting the group to 107 works, why is that not the default? I feel this is a bug, and making Emby run as the `render` group is a bad hack, not the fix. Or if it is the fix, shouldn't it be part of the instructions here? https://emby.media/truenas-scale-server.html

Though that already has the wrong URL on the guide (should be port 9096 not 8096) so I suspect it does not get updated much.

@DAVe3283did you try group 107?

[DAVe3283 7]

I ended up setting a startup script to do `chmod 777 /dev/dri` as that was the only thing that seemed to work reliably. Setting the group didn't work, but I am not 100% sure I applied it correctly. The security implications are more theoretical than practical, and this system belongs to a friend that I don't get constant physical access to, so I went with a solution I can set and forget for now.

[thekingpin 18]

10 minutes ago, DAVe3283 said:

I ended up setting a startup script to do `chmod 777 /dev/dri` as that was the only thing that seemed to work reliably. Setting the group didn't work, but I am not 100% sure I applied it correctly. The security implications are more theoretical than practical, and this system belongs to a friend that I don't get constant physical access to, so I went with a solution I can set and forget for now.

How did you go about setting the group? It should be set within the container "Section Help" area. Container -> Edit -> Section Help -> Group ID.

Maybe double check what other video render groups there are and verify the group ID number. Also what GPU are you using? 

I have Rx580 True NAS scale EE and setting the group to 107 works. 

 

[tech37 7]

In a similar boat but no luck yet - when I check "Use this GPU" from the Emby app settings in TrueNAS, I get the attached error messages talking about the UUID not being set.

However, when I manually set the GPU UUID in /mnt/.ix-apps/app_configs/emby/versions/1.2.17/user_config.yaml, it still doesn't appear in Emby, and when I try to check the box afterward, it clears the UUID field again.

Have tried changing the Emby app group ID to both video (44) and render (107), as well as add the emby 'apps' user to both groups, but no dice as of yet.

From within the container shell it has /dev but not /dev/dri, so not sure if it's an issue with the docker-compose.yml setup itself? This thread is one of the closest to my issue that I've seen so thought I'd try here

Any help appreciated!

truenas_use_gpu_error.txt user_config.yaml

[thekingpin 18]

44 minutes ago, tech37 said:

In a similar boat but no luck yet - when I check "Use this GPU" from the Emby app settings in TrueNAS, I get the attached error messages talking about the UUID not being set.

However, when I manually set the GPU UUID in /mnt/.ix-apps/app_configs/emby/versions/1.2.17/user_config.yaml, it still doesn't appear in Emby, and when I try to check the box afterward, it clears the UUID field again.

Have tried changing the Emby app group ID to both video (44) and render (107), as well as add the emby 'apps' user to both groups, but no dice as of yet.

From within the container shell it has /dev but not /dev/dri, so not sure if it's an issue with the docker-compose.yml setup itself? This thread is one of the closest to my issue that I've seen so thought I'd try here

Any help appreciated!

truenas_use_gpu_error.txt 7.31 kB · 0 downloads user_config.yaml 2.96 kB · 0 downloads

I don't think you can paasthrough a NVIdia GPU to containers in Scale. At least that option is not available directly when editing the container from the GUI (Electric eel)

You can only passthrough non-nvidia

[tech37 7]

5 minutes ago, thekingpin said:

I don't think you can paasthrough a NVIdia GPU to containers in Scale. At least that option is not available directly when editing the container from the GUI (Electric eel)

You can only passthrough non-nvidia

I would have thought that too, except scale has an option to install nvidia drivers and enable support, and other containers like immich have played with it nicely

f

[thekingpin 18]

3 minutes ago, tech37 said:

I would have thought that too, except scale has an option to install nvidia drivers and enable support, and other containers like immich have played with it nicely

f

I think this is the previous release of truenas, those options have changed in latest release. I would not edit the configs manually either and stick to GUI. What UID and GUI have you set for the container?

[thekingpin 18]

On second thought, have you tried stopping immich which is currently using the GPU. I think there was something in the previous TrueNas that you can only passthrough the GPU to a single container and not multiple at the same time. 

[tech37 7]

On 3/7/2025 at 12:05 PM, thekingpin said:

On second thought, have you tried stopping immich which is currently using the GPU. I think there was something in the previous TrueNas that you can only passthrough the GPU to a single container and not multiple at the same time. 

Hmm I actually had both Immich and Nextcloud 'using this GPU', but I disabled them anyway, and then the 'use this gpu' did get checked for the Emby container/app, even though it still didn't show anything in the Transcoding section or in the container shell.

Still got the same error when trying to hit 'update' on the app while the checkbox was configured, and no change after turning the app container off and on Am I missing something obvious?

Edited March 8 by tech37

[thekingpin 18]

17 minutes ago, tech37 said:

Am I missing something obvious?

I don't think so, but, from what your previous reply said, you edited the config manually:

GPU UUID in /mnt/.ix-apps/app_configs/emby/versions/1.2.17/user_config.yaml

 

I would suggest not doing that. The error shows something wrong with the config. 

Quickest fix is probably:

Take a backup of emby and reinstall the container without manually editing the config. 

[tech37 7]

1 hour ago, thekingpin said:

I don't think so, but, from what your previous reply said, you edited the config manually:

GPU UUID in /mnt/.ix-apps/app_configs/emby/versions/1.2.17/user_config.yaml

 

I would suggest not doing that. The error shows something wrong with the config. 

Quickest fix is probably:

Take a backup of emby and reinstall the container without manually editing the config. 

@thekingpinwe're in business! The user_config.yaml file looks to have been fine, but the backup, delete app/container, reinstall and restore did the trick!

I think I installed the Emby TrueNAS Scale app before I had checked the 'support Nvidia GPUs'  checkbox, so the app was installed without that support, but now the GPU is showing up in Emby as expected. Thanks for the help!