b3rs3rk

It likely won’t matter much. If they install the Nvidia driver or some third party package does it will likely have the nvidia-smi tool and that’s all my plugin requires.

Yeah I’m guessing they changed something with JavaScript processing based on the output you guys are showing. I don’t have a 6.9 beta machine to test with at the moment and I’m not going to run my UNRAID box on a beta version if I can help it. It’s too stable to mess with. I will see about getting a dev instance spun up. Unfortunately lime tech doesn’t give developers a good way to do this without forcing us to buy multiple licenses of UNRAID and I haven’t gotten around to that.

While the intention has always been to support multiple GPU vendors, we only officially support Nvidia GPUs with the UNRAID Nvidia plugin installed at this time.

I would have to add it. It is currently hard coded in the page JavaScript.

Two posts up I show a potential way to squelch those logs. Did you attempt it?

My suggestion is to update your BIOS, change to Legacy boot if currently UEFI (if possible), or switch the physical slot your GPU is in and see if the problem goes away. The issue is your build parameters, not my plugin. The only other option would be to figure out a way to squelch that log. Maybe a combination of using this script and creating a custom .conf file to remove the offending log messages. If anyone wants to try it, create a file (02-barmappingblock.conf) in /etc/rsyslog.d/ that contains: :msg,contains,"mapping multiple BARs" stop :msg,contains,"which spans more than PCI Bus" stop Then do a: /etc/rc.d/rc.rsyslogd restart > /dev/null If it works, then copy that .conf file to the location the script suggestions, and then add it to your startup scripts when UnRAID boots. It's not a 'fix' per se, but a kernel/driver issue can't be fixed by myself or the linuxserver folks.

It's mostly developed how I think it should work, but I have no way to test any of it. I'm trying to get remote access to a friend's UnRAID with an iGPU so that I can test pushing the code to it and see how the Intel stuff functions. I'm not a professional developer, so diagnosing this stuff based on user reporting isn't my forte. Just kind of hard to do.

Well nevermind then. I saw some old mailing list traffic about Haswell/Ivy Bridge systems having improper MCH mappings that cause issues like yours. But something else may be causing it.

Cool. Let me think on the best way to handle this in code. EDIT: Okay, so I'm going to change the GPU identifier when the nvidia-smi statistics command is run to use the GPU guid instead of the ID number. GUID should never change and I probably should have used that in the first place. Hindsight is 20/20 and all. Probably going to cause a little pain for updaters since the old GPUID will be invalid.

Okay, I’m on it now. Try starting your VM and then pasting me the result of: nvidia-smi -L Then shutdown the VM and run that command again and paste me this second result. My guess is when you pass the GPU through nvidia-smi can no longer see it. Then GPUID 1 becomes zero.

I'm happy to help as this seems like a bug, but I'm afraid I don't understand your post.

What CPU do you have in your system?

You can disable that statistic to prevent displaying it but that’s about it. Either the hardware or the driver doesn’t support the polling of the power draw. It stands to reason that buying cheaper cards can mean lesser monitoring features.

I will look at adding this when I'm finished doing the Intel integration. Shouldn't be difficult and now that individual statistics are selectable users can feel free to disable it if they don't want to display it.

I cannot reproduce the errors. Regrettably, even if I could reproduce them I'm not sure my code is in any way capable of fixing them. I'm only running a utility that comes with the Unraid-Nvidia build. In turn, they are only including a utility that comes with the NVIDIA driver. If it were compiled code that I authored it would be a different story. But I'm only receiving the output of that command to populate the widget data. The impact from running my code should be minimal to say the least. Uninstall my plugin and from the UnRAID console run: watch -n 2 nvidia-smi -q -x If you are still getting the errors in syslog, it begs the question how my code can be at fault.

If you cat the contents of /boot/config/plugins/gpustat/gpustat.cfg does it show your selected DISP settings at the end equaling 1?

I really need to write the page code to leave fields out that end up N/A. Unfortunately, what you see is what you get. I can see where there are other fields that contain the data the plugin wants that I can add logic to get instead but it won't be everything. I'll add it to the list.

There might be. Read the first post of this thread and send me the information requested at the bottom of it. It's possible your information is displayed in other fields of the XML but I won't know for sure without looking at it.

Bandwidth is important but so is delay. Being a few blocks away doesn’t mean anything if it is two different carriers with suboptimal peering on the backbone. How much delay between the two endpoints?

For posterity's sake, could you paste me the output of nvidia-smi -q -x -i 0

In which log are you seeing this? /var/log/syslog?

I'm considering doing something like the NIC widget that allows you to change the polled interface to a different one except for multiple GPUs. In the near term, I'd like to create a more detailed status page similar to the full display Disk Location has that displays all GPU information regardless of vendor or quantity.

Did you install the Unraid-Nvidia plugin, install an Nvidia build through the plugin settings, then reboot your UnRAID? Without doing this, you don't have the nvidia-smi utility which will cause that error.

Minimum version is 6.7.1. This app shouldn't even show up for you in Community Applications.

@dlandon I removed the preclear plugin and clicked the reload button in UD. No change. Went ahead and powered off the drive. Waited a minute. Reloaded UD again and verified it was gone. Then powered on the drive again. When it came back into UD it was still the wrong name. I cleared cache and reloaded a few more times and it finally showed up correctly.