[Plugin] Linuxserver.io - Unraid Nvidia


Recommended Posts

I'm starting to suspect this is a Plex issue. No Emby users have reported an issue, and three similar reports recently.

Are any of you or all of you using the decoding patch for Plex?

Sent from my Mi A1 using Tapatalk

Link to comment
tonight my plex app crashed caused my unraid server to need a reboot upon doing so i had a problem loading plex and emby up, I looked in the logs to see 
 kernel: nvidia-uvm: Loaded the UVM driver in 8 mode, major device number 245kernel: NVRM: RmInitAdapter failed! (0x31:0xffff:834)kernel: NVRM: rm_init_adapter failed for device bearing minor number 0iceberg kernel: NVRM: RmInitAdapter failed! (0x31:0xffff:834)kernel: NVRM: rm_init_adapter failed for device bearing minor number 0rc.docker: PlexMediaServer: Error response from daemon: OCI runtime create failed: container_linux.go:344: starting container process caused "process_linux.go:424: container init caused \"process_linux.go:407: running prestart hook 0 caused \\\"error running hook: exit status 1, stdout: , stderr: exec command: [/usr/local/bin/nvidia-container-cli --load-kmods --debug=/var/log/nvidia-container-runtime-hook.log --ldcache=/etc/ld.so.cache configure --ldconfig=@/sbin/ldconfig --device=GPU-cc63faa0-8033-52ac-dad4-79279b371033 --compute --compat32 --graphics --utility --video --display --pid=31670 /var/lib/docker/btrfs/subvolumes/4787b541f516ce4d01faa8f10f4dfed05c53589f0b299ae78d883bc14cdc346d]\\\\nnvidia-container-cli: device error: unknown device id: GPU-cc63faa0-8033-52ac-dad4-79279b371033\\\\n\\\"\"": unknownrc.docker: Error: failed to start containers: PlexMediaServerkernel: NVRM: RmInitAdapter failed! (0x31:0xffff:834)kernel: NVRM: rm_init_adapter failed for device bearing minor number 0kernel: NVRM: RmInitAdapter failed! (0x31:0xffff:834)kernel: NVRM: rm_init_adapter failed for device bearing minor number 0rc.docker: EmbyServer: Error response from daemon: OCI runtime create failed: container_linux.go:344: starting container process caused "process_linux.go:424: container init caused \"process_linux.go:407: running prestart hook 0 caused \\\"error running hook: exit status 1, stdout: , stderr: exec command: [/usr/local/bin/nvidia-container-cli --load-kmods --debug=/var/log/nvidia-container-runtime-hook.log --ldcache=/etc/ld.so.cache configure --ldconfig=@/sbin/ldconfig --device=GPU-cc63faa0-8033-52ac-dad4-79279b371033 --compute --compat32 --graphics --utility --video --display --pid=32044 /var/lib/docker/btrfs/subvolumes/6ea3c2c0c5428be0afad1d23b1f459f821776f1dcf923641c55aae2b4acdd312]\\\\nnvidia-container-cli: device error: unknown device id: GPU-cc63faa0-8033-52ac-dad4-79279b371033\\\\n\\\"\"": unknownrc.docker: Error: failed to start containers: EmbyServer

and when i went in to my nvidia plugin my p2000 does not show up but it is listed in iommu group 28 under system devices, I'm on ver 6.7

Turn off all autostarting VMs if possible and reboot Unraid.

Sent from my Mi A1 using Tapatalk

Link to comment
18 minutes ago, CHBMB said:

I'm starting to suspect this is a Plex issue. No Emby users have reported an issue, and three similar reports recently.

Are any of you or all of you using the decoding patch for Plex?

Sent from my Mi A1 using Tapatalk
 

Not any more, because of this. I had a Plex meltdown last week that caused filesystem corruption. The server would lock up and require a hard reboot within minutes of starting the Plex docker. After several repeats of this, I started in maintenance, ran check and repair, nuked docker image and reinstalled everything minus the decode patch, and it's been fine.

 

TL;DR: +1 for this being a Plex issue.

Link to comment
 
Untitled.thumb.png.ecfba83db58edea15fa918a69905ca60.png
And are you using the Plex decode patch that somebody posted details of here. If you are remove it, as we don't support it and the person above you has confirmed it causes issues.

Sent from my Mi A1 using Tapatalk

Link to comment

No, I'm not using it

 

I have previously been hw transcoding with Intel Quicksync on my CPU. Is there a legacy issue with that?

 

May 19 15:59:33 Tower kernel: nvidia 0000:01:00.0: swiotlb buffer is full (sz: 4194304 bytes)
May 19 15:59:33 Tower kernel: nvidia 0000:01:00.0: swiotlb buffer is full

What the hell is swiotlb?

Edited by u13rr1
More info
Link to comment
3 hours ago, u13rr1 said:

No, I'm not using it

 

I have previously been hw transcoding with Intel Quicksync on my CPU. Is there a legacy issue with that?

 


May 19 15:59:33 Tower kernel: nvidia 0000:01:00.0: swiotlb buffer is full (sz: 4194304 bytes)
May 19 15:59:33 Tower kernel: nvidia 0000:01:00.0: swiotlb buffer is full

What the hell is swiotlb?

This might be a kernel issue. Have you tried an earlier release to see if it happens there?

It has to to with devices accessing memory.

Link to comment
8 hours ago, fr05ty said:

tonight my plex app crashed caused my unraid server to need a reboot upon doing so i had a problem loading plex and emby up, I looked in the logs to see 

and when i went in to my nvidia plugin my p2000 does not show up but it is listed in iommu group 28 under system devices, I'm on ver 6.7

 

Here's your error here, as we've mentioned several times in the thread, there's a bug in the forums that pastes extra (invisible characters) to your command.

 

unknown device id: GPU-cc63faa0-8033-52ac-dad4-79279b371033\\\\n\\\"\"": unknown

This bit looks very suspicious of that.

 

\\\\n\\\"\""

Try typing the GPU id in manually.

Link to comment

@CHBMB i had been testing the nvdec script as all of my family are outside my home network and mostly are transcode, the server was up and working (after 6.7 update) for just over a week with no drama before going belly up whilst transcoding an episode, now my p2000 does not show up in the unraid nvidia plugin page, i have disabled the vm's from unraid settings and removed the script and rebooted once and also powered off once, but under iommu i can see it listed, do you think trying 6.7.1rc1 install might be woth a shot?

Link to comment

Reposted with more details and images;

 

Spec:

CPU: i7-3770S

MB: ASRock Z77 Extreme11

RAM : 24GB

GPU: 1050Ti

Unraid Version: 6.7.0 Nvidia build - Not running any patch to bypass transcode limit

binhex-emby - latest (I have an Emby Premiere Subscription)

binhex-plexpass - latest (I have a Plex Pass)

No VMs running or installed

 

Emby isn't showing the GPU and isn't HW transcoding:

 

binhex-emby_transcode_conf.jpg.2a61decc8e080ed9cbd8fcc76f8a2d2a.jpg

 

nvidia-smi_idle.jpg.955a10551ce7adc51b14c5fef7c1147d.jpg

 

Plex isn't showing the GPU but is transcoding:

 

nvidia-smi_plex_transcode.jpg.e4af1ad6237baf14f51d62116261b964.jpg

nvidia-smi_plex_transcode_stat.jpg.578c630a59f50394ab8d61f2b188b5c8.jpgbinhex-plexpass_transcode_conf.jpg.703b476a688894987ccdc4c497c36a99.jpg

 

I get this message as soon as I run the either dockers, but it does appear to be transcoding for Plex. The last two line just repeat over and over again.

May 21 14:06:12 Tower kernel: nvidia-uvm: Loaded the UVM driver in 8 mode, major device number 245
May 21 14:06:12 Tower kernel: resource sanity check: requesting [mem 0x000c0000-0x000fffff], which spans more than PCI Bus 0000:00 [mem 0x000d4000-0x000d7fff window]
May 21 14:06:12 Tower kernel: caller _nv000934rm+0x1bf/0x1f0 [nvidia] mapping multiple BARs

 

Unraid Nvidia page:

Unraid_Nvidia_page.thumb.jpg.51e66bee22f3f363eddb71e8e80e4fd7.jpg 

 

binhex-emby docker config:

binhex-emby_conf.thumb.jpg.a1c9554a33d55ae3ea6f369d1365631c.jpg

 

binhex-plexpass docker config:

binhex-plexpass_conf.thumb.jpg.51b070c7dc2cf775f4e9c0dd97e8e493.jpg

 

 

Any ideas?

 

Thanks

Edited by _sb
Link to comment
2 hours ago, ramblinreck47 said:

@_sb Do you have your iGPU disabled? Do you have a monitor plugged in? If so, where?

No it isn't disabled, but I do have the 1050ti set as the main gfx card in the bios. No monitor plugged in to either the MB or the 1050ti

Edited by _sb
Link to comment

I've disabled the "iGPU Multi Monitor" setting and have the 1050ti as the primary card. From what I can ascertain this effectively disables the iGPU. Same message and results.

 

May 21 18:28:46 Tower kernel: nvidia-uvm: Loaded the UVM driver in 8 mode, major device number 245
May 21 18:28:46 Tower kernel: resource sanity check: requesting [mem 0x000c0000-0x000fffff], which spans more than PCI Bus 0000:00 [mem 0x000d4000-0x000d7fff window]
May 21 18:28:46 Tower kernel: caller _nv000934rm+0x1bf/0x1f0 [nvidia] mapping multiple BARs

 

Link to comment
29 minutes ago, _sb said:

I've disabled the "iGPU Multi Monitor" setting and have the 1050ti as the primary card. From what I can ascertain this effectively disables the iGPU. Same message and results.

 


May 21 18:28:46 Tower kernel: nvidia-uvm: Loaded the UVM driver in 8 mode, major device number 245
May 21 18:28:46 Tower kernel: resource sanity check: requesting [mem 0x000c0000-0x000fffff], which spans more than PCI Bus 0000:00 [mem 0x000d4000-0x000d7fff window]
May 21 18:28:46 Tower kernel: caller _nv000934rm+0x1bf/0x1f0 [nvidia] mapping multiple BARs

 

Hmmm. Guess that eliminates the iGPU possibly being the problem.

 

Do you now have your 1050TI plugged into a monitor? I’m thinking it’s not registering properly because it needs to have a monitor plugged into it or a dummy HDMI.

Link to comment
15 minutes ago, ramblinreck47 said:

Hmmm. Guess that eliminates the iGPU possibly being the problem.

 

Do you now have your 1050TI plugged into a monitor? I’m thinking it’s not registering properly because it needs to have a monitor plugged into it or a dummy HDMI.

 

Plugged a monitor in via hdmi and still the same result

Link to comment
1 hour ago, saarg said:

@_sb

 Can you try to set the iGPU as primary?

 

iGPU set as primary, multi monitor disabled and a monitor plugged in to the hdmi on the MB.

 

Plex still HW transcodes but with no errors. Emby doesn't HW transcode, but also now doesn't show any errors. Oddly the monitor plugged doesn't show any output on the screen, but it does know that something is connected.

 

So we're getting there slowly. Starting to wonder if this is an Emby issue

 

Link to comment
14 hours ago, saarg said:

So the lesson is to not use the gpu you want for transcoding.

Is that the lesson though? I haven’t heard of anyone having trouble using their primary GPU (the one they use to boot up their server) to pass through for transcoding. This one was weird. I’m thinking it was some sort of issue with his Docker container. Correct me if I’m off base.

Link to comment

no amount of restarts reboots disabling things could bring back my p2000, so my last ditch effort was to upgrade to 6.7.1rc1, I rebooted and I went to the unraid NVIDIA plugin page and my p2000 was listed there again, I will just have to wait until plex decides it wants to implement hw decoding, thanks for the help and great work on these builds

Link to comment

I'm not sure what you guys are doing but using the latest LISO Plex container with the latest NVDEC script and it works fine for me on UNRAID 6.7 w/ an Nvidia GTX970. 

 

Some of you might be using an outdated NVDEC patch script.

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