Think the latest version of the plex app has quite a few issues with gpu support at the momentThings appear to have changed.
plex no longer seems to detect the nvidia GPU
Think the latest version of the plex app has quite a few issues with gpu support at the momentThings appear to have changed.
plex no longer seems to detect the nvidia GPU
Multiple apps have simultaneously accessed it? Your screen shot shows a single session for a single appThis currently works for me. I have truenas scale 22.12.2 and plex in a container 1.32.3.7162_1.7.49. using an AMD processor with no GPU and a stand alone RTX 2600.
So I am using the GPU for the truenas install and for plex. I have no other GPUs. I could test another app if there is an easy one to setup. I do not know if multiple apps and truenas can all use the same GPU at the same time.Multiple apps have simultaneously accessed it? Your screen shot shows a single session for a single app
Do we have a brave beta tester of Cobia, now that it’s out, who can confirm GPU sharing is possible? :)According to advice from the Truechart Discord, this shared GPU trick is no longer possible with their new common chart. It seems TrueNas Scale has this feature planned natively in a future version (cobia) and thus it working in a TC chart unofficially is no longer possible.
Well that sounds new. Previously there was only one GPU resource to assign, so if you used that functionality you couldn’t share a single GPU with multiple apps. Sounds like they added functionality! Previously a single GPU could be shared by multiple apps by avoiding that resource assignment, and instead by adding/using environment variables. This worked with TrueChart apps for a bit then it stopped. I am not sure if there is any advantage of either design, but will be testing this when it hits GA!Just tried it on Cobia RC1. The UI is still a bit buggy showing 5 available GPU for allocation, but I was able to get multiple containers running nvidia-smi out of the box.
Apps can share the same GPU. I've been doing it forever. When your setting up the deployment, do not allocate GPU to it as it will not allow any other pod to use it.
Set your env variables like this on each deployment to utilize GPU:
View attachment 58020
The docker containers run as root while the truecharts ones typically don't. It's quite possible the user running the container doesn't have access to the /dev/dri device.I ungraded to Cobia. I run TC versions of Plex and Tdarr and added a single Nvidia GPU to each app. in plex GPU assist works, but when I add tdarr nvenv does not work.
According to advice from the Truechart Discord, this shared GPU trick is no longer possible with their new common chart. It seems TrueNas Scale has this feature planned natively in a future version (cobia) and thus it working in a TC chart unofficially is no longer possible.