(6.12.3) Using Intel iGPU eventually crashes Unraid - i915 related?

[SinoBreizh]

Hello,

 

I've recently susbscribed to Plex Pass and installed Jellyfin to compare options and make up my mind on which I want to keep as my media server. I've had no issues using hardware transcoding in both Plex Pass and Jellyfin. Additionally, I've installed Intel GPU Top and the GPU statistics plugin to confirm hardware transcodingg was indeed being used, through the "intel_gpu_top" command.

 

As soon as I did this, I've experienced crashes that I'd notice because Plex and Jellyfin clients couldn't connect to my server. I'd hop on to the Unraid GUI, and it too would be unable to connect (the login screen doesn't appear). Hopping on the server itself connected to a monitor through HDMI on the iGPU, the image would either be frozen, or filled with pink and green artifacts. With no way to control the server, it being seemingly bricked, a hard reset is necessary each time.

 

I have read up on possible causes, and I've come to this thread, and my symptoms seem to be similar. Indeed, I did not have to be actively playing media to cause a crash: I regularly watch shows flawlessly in the evening and wake up with the system bricked in the morning; sometimes the system hangs minutes after it boots, without having the time to play any media back.

 

I've tried removing the additional command from Plex and Jellyfin assigning the iGPU to them, as well as removed the "Intel GPU Top" and "GPU statistics" plugins from my setup. This appears to have, for now, resolved the issues. I cannot comment with certainty as I'd have to leave the system on for longer to confirm, but I didn't want to wait until then to post.

 

TL;DR: Using the Intel iGPU eventually always crashes my system. Obviously doing without hardware transcoding is far from ideal, and negates the point of having an Intel build to begin with. So what are my options at this point? I've scoured the thread I linked, and I couldn't really find a solution, or at least one I understand and have been able to deploy.

 

Quick info: CPU is an Intel i3 - 10100 running on a Z490-A Pro mb, with 48GB of RAM ( 2 x 8 + 2 x 16). Plex + Jellyfin appadata run off a healthy NVMe-cache.

 

I've attached system diagnostics as well as a syslog (mirrored to flash). To avoid wasting your time, in the syslog, example crashes occured at 21h48 (9:48 pm) and 22h02 (10:02 pm) on August 12th.

 

Thanks in advance!

tower-diagnostics-20230813-0958.zip syslog

[strike]

Did you try what's mentioned in the 6.12.0 release notes?

 

Crashes related to i915 driver
We are aware that some 11th gen Intel Rocket Lake systems are experiencing crashes related to the i915 iGPU. If your Rocket Lake system crashes under Unraid 6.12.0, open a web terminal and type this, then reboot:

echo "options i915 enable_dc=0" >> /boot/config/modprobe.d/i915.conf

Setting this option may result in higher power use but it may resolve this issue for these GPUs.

 

[SinoBreizh]

No, I haven't seen this before; I'll try it now!

 

Is this a permanent change, or a command I'd have to type in after every reboot?

[strike]

it's a permanent change as it gets written to the config file on the flash drive

[SinoBreizh]

Understood. I've run the command and rebooted, and rebinded the iGPU to Plex and Jellyfin, as well as reinstalled Intel GPU Top to monitor. I'll report back in a bit once I see how it goes.

 

On a side note, does Unraid have an issue tracker of some sort? A place where I can follow a specific issue to see if it's resolved, rather than reading every single release note?

[strike]

2 minutes ago, SinoBreizh said:

On a side note, does Unraid have an issue tracker of some sort? A place where I can follow a specific issue to see if it's resolved, rather than reading every single release note?

You can follow the bug report discussions here: https://forums.unraid.net/bug-reports/ But release notes is a good idea to read anyway.

[SinoBreizh]

Hello again @strike

 

I've encountered another crash with my server. The config file was correctly created, yet the crash still occurred. The Unraid UI failed to connect, and I had no video output on HDMI for the server as of today August 17th 10h45 (10:45 am).

 

I checked the syslog mirrored to flash, and I can't find any error that would explain the crash. I'm posting the syslog and new diagnostics as an attachment in case anyone here might be able to help me figure this out.

syslog tower-diagnostics-20230817-1124.zip

Edited August 17, 2023 by SinoBreizh

[strike]

I can't see anything obvious in you last syslog either. I have no clue what the issue might be. But have you tried running in safe mode? 

[SinoBreizh]

I might have an idea; there's a memory leak while running a specific game server through wine in a docker container, and I've had to run a command provided by ich777 to remedy it: but he did warn that he didn't like this solution, and might cause instability. That's the only variable I can think of. I've never encountered a crash without an explanation in the logs before otherwise.

 

I'll turn this fix off, and see if it still crashes.

[Mwidders]

 I have this exact same issue as well.  Has anyone figured out the solution?  Do I need to disable my onboard video and just use my NVIDIA card instead?  As mentioned it does seem kind of stupid to do so, but nothing else has worked.  If I disable one card in the bios and not the other it is perfectly stable, but why should I have to? Any help would be great!