[Plugin] Linuxserver.io - Unraid Nvidia


Recommended Posts

17 minutes ago, reggierat said:

Then i deleted the container and added it back and the info was missing from the template

If you just added a new container then you would only have the default template which likely was missing some of your custom settings.

 

If you had selected your previous template from the dropdown at the top of the Add Container page it would have had all those settings. That template is stored on flash for you to reuse anytime.

 

Even simpler, and the way I usually make this recommendation, is to just go to the Apps page and use the Previous Apps feature. It will reuse those templates for you.

Link to comment
4 minutes ago, reggierat said:

root@localhost:# /usr/local/emhttp/plugins/dynamix.docker.manager/scripts/docker run -d --name='emby' --net='bridge' -e TZ="Australia/Sydney" -e HOST_OS="Unraid" -e 'NVIDIA_VISIBLE_DEVICES'='GPU-2c8cd41b-4bf5-73dc-1eff-96d27d64113b' -e 'PUID'='99' -e 'PGID'='100' -p '8096:8096/tcp' -p '8920:8920/tcp' -p '1900:1900/udp' -p '7359:7359/udp' -v '/mnt/user/Movies/':'/movies':'rw' -v '/mnt/user/TV Shows/':'/tv':'rw' -v '/mnt/user/Music/':'/music':'rw' -v '/mnt/user/appdata/emby':'/config':'rw' --runtime=nvidia 'linuxserver/emby' 

9cd52be25d822572dc4e5b54e16033f144f0afcddf59d438768d1e9224966f29

 

 

Still hitting the CPU

ffmpeg-transcode-069afe13-e1b1-438b-b6db-b01a2e627c6e_1.txt 24.41 kB · 0 downloads

You tried a different media file to play?

Link to comment
3 minutes ago, trurl said:

If you just added a new container then you would only have the default template which likely was missing some of your custom settings.

 

If you had selected your previous template from the dropdown at the top of the Add Container page it would had all those settings. That template is stored on flash for you to reuse anytime.

 

Even simpler, and the way I usually make this recommendation, is to just go to the Apps page and use the Previous Apps feature. It will reuse those templates for you.

What i mean is, the notes in the top of the template with instructions to add the extra parameter were missing

snip.JPG

Link to comment
On 12/9/2019 at 8:52 AM, CHBMB said:

Can you post the output of

 


nvidia-smi --query-gpu=gpu_name,gpu_bus_id,gpu_uuid --format=csv,noheader

for me please?

 

On 12/10/2019 at 11:25 PM, Eggman1414 said:

Sorry for the late reply (been sick).

Here is the output:

Quadro P2000, 00000000:0A:00.0, GPU-c6a0d9d0-ddbf-0fa6-56f4-dd385535d1b1

@CHBMB Any reason why it doesn't show? Even upgraded to 6.8 build and still does not show.

Link to comment
38 minutes ago, Eggman1414 said:

 

@CHBMB Any reason why it doesn't show? Even upgraded to 6.8 build and still does not show.

The only reason I can think of, is during the boot process I parse the cards and write the data to some tmp files, which in turn are parsed when you load the Unraid Nvidia page.  So my guess is that during boot, it's a race condition and we're parsing the card before it's ready.

 

Can you try running this.

 

rm /tmp/nvidia/nvidia

nvidia-smi --query-gpu=gpu_name,gpu_bus_id,gpu_uuid --format=csv,noheader | sed -e s/00000000://g | sed 's/\,\ /\n/g' > /tmp/nvidia/nvidia

then reload the Unraid Nvidia page

Link to comment
4 hours ago, CHBMB said:

The only reason I can think of, is during the boot process I parse the cards and write the data to some tmp files, which in turn are parsed when you load the Unraid Nvidia page.  So my guess is that during boot, it's a race condition and we're parsing the card before it's ready.

 

Can you try running this.

 


rm /tmp/nvidia/nvidia

nvidia-smi --query-gpu=gpu_name,gpu_bus_id,gpu_uuid --format=csv,noheader | sed -e s/00000000://g | sed 's/\,\ /\n/g' > /tmp/nvidia/nvidia

then reload the Unraid Nvidia page

Now the 1080 doesn't show but the P2000 is now there.

image.png.4587578580842dc3f7d5c35e1c4cd9cb.png

 

Playing a file in Plex does show that HW transcoding is working too. Thank you for solving that little issue.

image.png.401d1777cf89ef203ad3ed9247e96f01.png

Link to comment
Now the 1080 doesn't show but the P2000 is now there.
image.png.4587578580842dc3f7d5c35e1c4cd9cb.png
 
Playing a file in Plex does show that HW transcoding is working too. Thank you for solving that little issue.
image.png.401d1777cf89ef203ad3ed9247e96f01.png
The 1080 won't show as you've passed it through to a VM haven't you?

Sent from my Mi A1 using Tapatalk

  • Haha 1
Link to comment

Thanks for all of the work on this plugin. My question is about HW decoding, which I understand isn't strictly the plugin, but hopefully it's ok to ask here.

 

This is my first update since adding a P2000 to my server. I expected to need to re-enable HW decoding after the update (manually or via a script), but when I run 'nvidia-smi dmon -s u' it looks like the decoder is still in use. So now I'm confused. Am I missing something?

Link to comment
45 minutes ago, shooga said:

Thanks for all of the work on this plugin. My question is about HW decoding, which I understand isn't strictly the plugin, but hopefully it's ok to ask here.

 

This is my first update since adding a P2000 to my server. I expected to need to re-enable HW decoding after the update (manually or via a script), but when I run 'nvidia-smi dmon -s u' it looks like the decoder is still in use. So now I'm confused. Am I missing something?

If you installed the unraid nvidia version and setup the Linux server plex container correctly (specifying the nvidia runtime and Your Gpu is) it should just work

Link to comment

Now i realise this is probably outside the bounds of what is supportable, but i have noticed since moving to unraid nvidia that the sleep plugin causes an unclean shutdown when attempting to put the server to sleep.

 

I have tested this with unraid nvidia 6.7.2 and 6.8, stock unraid the sleep plugin correctly puts the server into s3 sleep and i'm able to easily wake it.

 

Is this worth investigating or do i just give up on s3_sleep?

Link to comment
2 hours ago, reggierat said:

Now i realise this is probably outside the bounds of what is supportable, but i have noticed since moving to unraid nvidia that the sleep plugin causes an unclean shutdown when attempting to put the server to sleep.

 

I have tested this with unraid nvidia 6.7.2 and 6.8, stock unraid the sleep plugin correctly puts the server into s3 sleep and i'm able to easily wake it.

 

Is this worth investigating or do i just give up on s3_sleep?

 

The decision is down to you, but I doubt many of us use s3 sleep so you're probably going to be by yourself figuring it out.....

  • Like 1
Link to comment

Let me open by saying that the following is not in reference to any particular post here; it is simply a stream of consciousness holiday post!

 

You know, sometimes anyway, we are all “a few fries short of a Happy Meal!” Mostly though it is because we are ‘distracted’ by life, as @CHBMB pointed out to those who sometimes forget. The vast majority of the wonderful work done here is for solvo. So, even if it is only $5 (or a simple thank you), contribute something to those who give up their time for us, and try to remember they have lives too. The ghosts of Christmas past, present, and future would be so proud!

 

 

 

Well, this evening I upgraded my secondary servers, that were on RC releases, to 6.8 - stable; perfect, ZERO issues. Okay, all good. May as well do the media server too right? Ooops. So, here’s my point to this post (finally, right?): remember to read up a little and prepare for major changes to your system(s). I, for one, TOTALLY overlooked upgrading my Nvidia driver and then found myself scratching various body parts trying to figure out why Plex wouldn’t boot! Duh! Helps if you do all the prep work.

 

As has been suggested, you should always use the fantastic “Update Assistant” under Tools; which I always do. It has definitely saved me from grief in the past! And maybe, time and resources permitting, a future version Update Assistant might look to see if the Nvidia driver is installed and toss up a reminder to those of us that simply forgot, because this D@#$ed software just - flat - out - works. It is for that reason, and that reason alone, I initially overlooked pulling the Nvidia driver update! (I thought I ordered fries with my Happy Meal!?)

 

So, do your prep work and the upgrades should go smoothly. And try to remember to give credit where credit is due (it would take PAGES to acknowledge everyone). We should all try to remember to say ‘thanks’ once in a while to all those that make this product, and Community, as awesome as it is; particularly when posting to find an answer (hopefully, after we have tried to find it on our own).

 

In closing a hearty THANK YOU to EVERYONE who takes the time out of their lives to make these pieces all fit together and JUST WORK! If I had the money, you would all have a new Ferrari in your driveways Christmas morning; or a pony, or a kitty cat, or Christmas goose, or (fill in the blank). You all rock!

Happy Holidays everyone!

  • Like 1
Link to comment
1 hour ago, TechMed said:

s). I, for one, TOTALLY overlooked upgrading my Nvidia driver and then found myself scratching various body parts trying to figure out why Plex wouldn’t boot! Duh! Helps if you do all the prep work.


how do you update the nvidia driver? I tried to look through the first few posts of this thread and can’t find that info. 
 

I noticed that the nvidia site has a slightly newer set of drivers. 

Link to comment
3 minutes ago, tmchow said:


how do you update the nvidia driver? I tried to look through the first few posts of this thread and can’t find that info. 
 

I noticed that the nvidia site has a slightly newer set of drivers. 

 

You can't upgrade the driver.  It will be the driver version at the time LinuxServer compile wathever new release of Unraid they release (rc or stable).  Also, it's not the one from Nvidia.com, but available in the base OS of Unraid (it's slightly behind than the windows version).

Edited by Pducharme
Link to comment
6 minutes ago, Pducharme said:

 

You can't upgrade the driver.  It will be the driver version at the time LinuxServer compile wathever new release of Unraid they release (rc or stable).  Also, it's not the one from Nvidia.com, but available in the base OS of Unraid (it's slightly behind than the windows version).

https://slackbuilds.org/result/?search=nvidia&sv=14.2

 

It's actually slackbuilds that makes the drives for us from nvidia.

 

Linuxserver rely on them for drivers.

Link to comment
3 minutes ago, Dazog said:

https://slackbuilds.org/result/?search=nvidia&sv=14.2

 

It's actually slackbuilds that makes the drives for us from nvidia.

 

Linuxserver rely on them for drivers.

Yes, that is the source (i wasn't remembering the name :) ).  So, when Slackbuilds have a new one  AND   LimeTech have a new release of UNRAID, then, LinuxServer will compile a new UNRAID-Nvidia build.

Link to comment
11 hours ago, tmchow said:

This isn’t updating the driver by itself this is updating unraid which I know how to do :)  I was understanding that post I quoted to mean we could update the drivers independently (which others have corrected me saying you can’t)

Technically you can.

 

1.  Install all the required dependencies.

2.  Recompile the Linux kernel including any OOT drivers

3.  Install more dependencies, including compiling some Nvidia dependencies from source like Go, which takes an age.

4.  Compile the Nvidia drivers, there's a tiny bit of hackery here as there's no desktop environment by default in Unraid so it fails without (iirc it's something to do with the GTK toolkit, I can't remember how I fixed it in the end, I think I removed part of the Slackbuild to skip that bit as we don't need it)

5.  Compile the Nvidia docker runtime and libseccomp

6.  Change the docker daemon json config

7.  Pull apart Unraid to it's component files, install all the bits and bobs then put it back together.

 

Before we made this everyone was asking for the Nvidia drivers, problem is, that doesn't solve the problem of utilising them in the docker containers,   actually compiling the drivers and making Slackware packages is the easy bit of the whole thing, problem is if you install at boot time like many other slackware packages you get a kernel panic.  (Probably because it's too late in the boot process)  In fact, when I started this process, that was the approach I initially took, install everything at boot time.

 

To put it into perspective, we have scripted this whole process and using 24 threads on my server, with 32GB of RAM, it still takes ~ 30-40 minutes to get through the whole thing.

 

 

  • Like 3
  • Thanks 1
Link to comment
  • trurl locked this topic
Guest
This topic is now closed to further replies.