[Plugin] Linuxserver.io - Unraid Nvidia


Recommended Posts

Recently I noticed many occurrences of many processes were 'unable to fork, resources not available' errors, and these processes failed to execute.  Suspecting out of memory issue (I have 32GB), I watch the top command for a while when these errors occurring, but it doesn't look that way.
'ps' command surprisingly shows there are 30k+ processes of nv_queue.

-- truncated --
32757 ?        S      0:00 [nv_queue]
32758 ?        S      0:00 [nv_queue]
32759 ?        S      0:00 [nv_queue]
32760 ?        S      0:00 [nv_queue]
32761 ?        S      0:00 [nv_queue]
32762 ?        S      0:00 [nv_queue]
32763 ?        S      0:00 [nv_queue]
32764 ?        S      0:00 [nv_queue]
32765 ?        S      0:00 [nv_queue]
32766 ?        S      0:00 [nv_queue]
32767 ?        S      0:00 [nv_queue]

# ps ax | grep nv_queue | wc -l
31198

This might come from nvidia driver. I have been using the nvidia build for a long time but the only different thing I do recently is the command

nvidia-smi -pm 1

to enable the persistence mode. Without this command, my graphics card (1080Ti) would always at 55W power when idle. With this command, it's down to 9-12W.

Does anyone encounter this issue?
I believe the very large numbers of nv_queue processes here has rendered the system unstable, depriving many other important processes of resources.

I have rebooted since, and NOT executing the -pm 1 command above. Many hours later now, no such issue. So more or less it's the command that is responsible for it. But of course now, the 55W idle power usage is back an issue.

 

Link to comment

Initially had 0 problems with this plugin and Plex, but now the containers are super touchy, sometimes working with --runtime=nvidia and sometimes not. The error I often get is:

 

root@localhost:# /usr/local/emhttp/plugins/dynamix.docker.manager/scripts/docker run -d --name='plex2' --net='bridge' --cpuset-cpus='25,26,27,28,29,30,31,57,58,59,60,61,62,63' -e TZ="America/Chicago" -e HOST_OS="Unraid" -e 'VERSION'='latest' -e 'PUID'='99' -e 'PGID'='100' -e 'NVIDIA_DRIVER_CAPABILITIES'='all' -e 'NVIDIA_VISIBLE_DEVICES'='GPU-1e9824c9-cf6f-9a6d-9b75-dccaa79ba9a1' -p '1901:1900/tcp' -p '32401:32400/tcp' -p '33400:33400/tcp' -v '/mnt/disks/union-thd-plex2/misc/':'/misc':'rw,slave' -v '/mnt/disks/union-thd-plex2/misc2/':'/misc2':'rw,slave' -v '/mnt/disks/union-thd-plex2/misc3/':'/misc3':'rw,slave' -v '/mnt/user/Photos/server2/':'/photos':'rw' -v '/mnt/cache/Transcode/plex2':'/transcode':'rw' -v '/mnt/disks/nvmepool/appdata/plex2':'/config':'rw,slave' --runtime=nvidia 'linuxserver/plex' 
598820f4e09e0ac0f5bdb8234210fecdc67487b3d3a3392e094b2683cccd2bdc
/usr/bin/docker: Error response from daemon: OCI runtime create failed: container_linux.go:346: starting container process caused "process_linux.go:449: container init caused "process_linux.go:432: running prestart hook 1 caused \"error running hook: exit status 1, stdout: , stderr: nvidia-container-cli: initialization error: cuda error: initialization error\\n\""": unknown.

The command failed.

EDIT: so when I run 'watch nvidia-smi' it comes up with my gpu, then later says no devices found.

Edited by d2dyno
Link to comment
30 minutes ago, d2dyno said:

Initially had 0 problems with this plugin and Plex, but now the containers are super touchy, sometimes working with --runtime=nvidia and sometimes not. The error I often get is:

 


root@localhost:# /usr/local/emhttp/plugins/dynamix.docker.manager/scripts/docker run -d --name='plex2' --net='bridge' --cpuset-cpus='25,26,27,28,29,30,31,57,58,59,60,61,62,63' -e TZ="America/Chicago" -e HOST_OS="Unraid" -e 'VERSION'='latest' -e 'PUID'='99' -e 'PGID'='100' -e 'NVIDIA_DRIVER_CAPABILITIES'='all' -e 'NVIDIA_VISIBLE_DEVICES'='GPU-1e9824c9-cf6f-9a6d-9b75-dccaa79ba9a1' -p '1901:1900/tcp' -p '32401:32400/tcp' -p '33400:33400/tcp' -v '/mnt/disks/union-thd-plex2/misc/':'/misc':'rw,slave' -v '/mnt/disks/union-thd-plex2/misc2/':'/misc2':'rw,slave' -v '/mnt/disks/union-thd-plex2/misc3/':'/misc3':'rw,slave' -v '/mnt/user/Photos/server2/':'/photos':'rw' -v '/mnt/cache/Transcode/plex2':'/transcode':'rw' -v '/mnt/disks/nvmepool/appdata/plex2':'/config':'rw,slave' --runtime=nvidia 'linuxserver/plex' 
598820f4e09e0ac0f5bdb8234210fecdc67487b3d3a3392e094b2683cccd2bdc
/usr/bin/docker: Error response from daemon: OCI runtime create failed: container_linux.go:346: starting container process caused "process_linux.go:449: container init caused "process_linux.go:432: running prestart hook 1 caused \"error running hook: exit status 1, stdout: , stderr: nvidia-container-cli: initialization error: cuda error: initialization error\\n\""": unknown.

The command failed.

EDIT: so when I run 'watch nvidia-smi' it comes up with my gpu, then later says no devices found.

Sounds like a hardware issue. Might be your power supply can't handle all your hardware.

Link to comment
On 2/9/2019 at 11:15 AM, linuxserver.io said:

Read this first

6.  We will produce one Nvidia build per Unraid release, we will not be updating the drivers multiple times for each Unraid version, unless there is a critical bug that demands this.  So please don't ask.

From the second post.

  • Like 1
Link to comment
On 5/22/2020 at 11:07 PM, saarg said:

Why are you posting in this thread??

Or maybe

"Hi Gee1" this thread is for specific issues with the unraid NVidia Plugin, to request changes to Dockers, see the support thread for that docker (should be in the following sub-forum: https://forums.unraid.net/forum/47-docker-containers/ ,  if not the app store entry for the docker should link to its support thread"

 

Edited by Snubbers
Link to comment
1 hour ago, Snubbers said:

Or maybe

"Hi Gee1" this thread is for specific issues with the unraid NVidia Plugin, to request changes to Dockers, see the support thread for that docker (should be in the following sub-forum: https://forums.unraid.net/forum/47-docker-containers/ ,  if not the app store entry for the docker should link to its support thread"

 

Yes, but that is too much to write when in a hurry ;)

  • Haha 1
Link to comment
23 hours ago, Snubbers said:

the app store entry for the docker should link to its support thread"

Or even easier. If you have already installed the docker, just click on its icon on the Dashboard or Dockers page and select Support.

Link to comment
16 minutes ago, trurl said:

Or even easier. If you have already installed the docker, just click on its icon on the Dashboard or Dockers page and select Support.

Nice!

 

OT, my 1660 super keeps dropping off the bus again :( within a few hours of rebooting.
I've written a user script that is scheduled every 10 minutes to just run nvidia-smi and put a log entry to say if it's present or not to see exactly when it drops off the bus. Currently you only know if you have GPU Stats installed so when you login to the WEBUI it fires up nvidia-smi and that's when you realise, but since I login fairly infrequently I don't know if there is a pattern to the GPU disappearing or if it's truly random.


 

Link to comment
On 5/22/2020 at 9:13 PM, Gee1 said:

hey linuxserver.io can u pls compile ur TVH docker with "nvenc" support? Would be nice

Why not use the Emby or Jellyfin container with TVHeadend container and let Emby/Jellyfin do the encoding (they both have NVENC encoding built in).

And you got a nice UI too. ;)

Edited by ich777
Link to comment

i already use emby together with TVHeadend but i also want to use Kodi with HTSP Client.. ill trying to create my own docker with ffmpeg with hevc support + tvh with nvenc support in one container for 3 days now. But its hard to get this container running

Link to comment

Hi all, quick noob question...  Ive seen reference to an Unraid Nvidia Build but not seen anywhere to download it, or does this just simply refer to the Stock Unraid but with the Nvidia image loaded?  I assumed that on the driver plugin, the Bvidia build was the patch, and the Stock build to restore original settings, is this not the case?

 

Ive installed the Nvidia Build onto my Stock Unraid...

 

Cheers

Link to comment
57 minutes ago, DrUseless said:

Hi all, quick noob question...  Ive seen reference to an Unraid Nvidia Build but not seen anywhere to download it, or does this just simply refer to the Stock Unraid but with the Nvidia image loaded?  I assumed that on the driver plugin, the Bvidia build was the patch, and the Stock build to restore original settings, is this not the case?

 

Ive installed the Nvidia Build onto my Stock Unraid...

 

Cheers

You install the plugin, then choose the nvidia build to replace the stock unraid. There is no patching of the stock unraid. The stock unraid is there to easily go back to stock unraid.

  • Thanks 1
Link to comment
1 hour ago, saarg said:

You install the plugin, then choose the nvidia build to replace the stock unraid. There is no patching of the stock unraid. The stock unraid is there to easily go back to stock unraid.

Many thanks, I had assumed that but was informed that Unraid Nvidia Build was actually a complete alternative version of Unraid geared up towards Nvidia hardware.  Everything seems to be working as expected after following the rest of the guide.

 

Cheers for the clarification.

Link to comment
2 hours ago, DrUseless said:

I had assumed that but was informed that Unraid Nvidia Build was actually a complete alternative version of Unraid geared up towards Nvidia hardware.

That is what it is and that is what he said.

4 hours ago, saarg said:

the nvidia build to replace the stock unraid.

 

Link to comment
  • trurl locked this topic
Guest
This topic is now closed to further replies.