axipher

The "No space left on device" look to point to either a bad mount, or a disk or vdisk or image is full with no free space for the log file and that is causing errors. Have you checked where stuff is mounting too, your docker image, and permissions of the appdata folder?

Yeah, I would first try just changing the Plex Device to /dev/dri and seeing what GPU's you can select in Plex and then testing again. Most reports I read indicated that the HDMI dummy plug is mostly needed for the iGPU to be used and not so much discrete GPU's, but also, does not hurt to have the dummy plug. I'm not sure how it affects power consumption though, I know just installing my Arc card increased my power draw by a little over 15 W on my server so I'm assuming it's the card's idle power draw. I can't see Linux kernel 6.9-rc3 breaking anything Arc related over 6.9-rc1, but if you can easily roll-back to the same version as me, at least it might be easier to troubleshoot.

First off, I must point out that I'm not an expert on Plex, Unraid or Arc, but this is my experience. Also I can't remember if there are certain sub-title formats that will force CPU encoding, so make sure to try testing on something and turn off sub-titles just to make sure that isn't messing it up. Audio might get CPU transcoded still, but that is a tiny load and shouldn't affect a server, especially yours with that many threads. I'm currently Unraid 6.12.9 with this kernel from Thor2000: Linux 6.9.0-rc1-thor-Unraid+ x86_64. I believe it was from the 20240331 release: https://github.com/thor2002ro/unraid_kernel/releases/tag/20240331. I'm pretty sure it worked on 6.12.8 though with the 6.7.9 kernel as well from Thor2000. I had a 3-pack of HDMI dummy plugs from previous QuickSync stuff so I installed one in my ARC GPU from day one, so I have note tested without it installed. For my Plex config, I have a device setup as just /dev/dri. I believe I tried going direct to /dev/dri/renderD128 or /dev/dri/renderD129 but it sometimes wouldn't work so I just put in /dev/dri and then selected the card in Plex. But since my system has a Intel i7-9700k, I have two Intel GPU's installed: - Card0 is the 9700k iGPU and is linked to Render128 - Card1 is the Arc 750 DG2 and is linked to Render129 And inside Plex itself, under transcode settings, make sure to select the DG2, since I can select either Intel GPU there. Then a restart of Plex docker is likely required: Both show up in GPU Stats for me and while Transcoding, I see the proper DG2 get used:

I'm no expert on all the other things Plex server does in the background, but if it starts pretty much on an hour mark, it is more than likely some automated task or clean-up happening. In Settings at the bottom, there is a Console Menu that lets you see the active logs of the server and you might be able to search through that to see what is using up your CPU in Plex.

Do you have any Scheduled Tasks in Plex? They have a section in server settings and a set time that you can compare to when you typically see CPU usage peg 1 core 100% This could be audio normalizing, video thumbnail generation, intro/credit detection, local metadata updates, database backups, etc. Some of those settings might be on libraries themselves.

I was more so just posting to note that what he has now does work with a different kernel, in case someone else might have wanted to see monitoring for an Arc GPU. My switchover from the Coffee Lake iGPU to the Arc GPU was for a few reasons: My library is starting to have more AV1 content in it The Coffee Lake iGPU had no AV1 decode, so and AV1 videos on my server would switch to software decode which at times cannot maintain even 1.0x transcode speed on a 1080p video with only 8 threads. This is especially a problem with 3 or more streams or if something else is using CPU on the server or audio transcoding is required. ARC is able to transcode faster than the iGPU on the same source files allowing for quicker skipping through media playback The 20W if extra idle power and bursts of power draw during transcoding is okay for me in my home server rack Arc GPU on sale was cheaper than a new motherboard and CPU or the similar Nvidia option I switched to the Thor2000 Custom Kernel because the current 6.12.8 Unraid Official Kernel doesn't support Arc for Plex Transcoding. Ideally staying on the Unraid included 6.1 kernel would have been best, but Arc wouldn't work without changing kernels. From my research, it seemed like the only main things I would lose from the custom kernel was being to use the Nvidia-plugin and an Nvidia GPU, and ZFS, both of which I'm not using at all in this Unraid box. This is also my secondary Unraid that runs Plex, redundant PiHole and some other redundant dockers, the main box with the media files is still on official kernel to avoid any chance of data corruption.

Just as confirmation here, the multi-gpu-stat does work with multiple Intel GPU's, in my case, using the Thor2000 6.7.9 custom kernel for my Arc A750 GPU alongside my 9700k iGPU. Both have HDMI Dummy Plugs.

Just installed an Arc A750 in my server to upgrade from the 9700k iGPU. The server also happens to have a GTX 750 Ti that I was using for a VM that I had seen reports of the Thor2000 custom kernel not playing nicely with Nvidia-Driver-Kernel. I'm on Unraid 6.12.8 and used Thor2000 driver: https://github.com/thor2002ro/unraid_kernel/releases/tag/20240310 ( Linux 6.7.9-thor-Unraid+ x86_64) My process: Uninstalled Nvidia-GPU-Plugin Shutdown server Copied the bz* files Installed GPU with same model HDMI Dummy Plug in the iGPU Booted system Updated GPU-Stat to show new DG2 GPU Deleted all codecs from Plex Appdata folder Added /dev/dri/renderD128 as a new device Started Plex Docker Changed Transcode Device to DG2 Initial testing was great, hardware transcoding now of AV1 content to H264, 1080p content typically transcoding at ~11x the framerate which is amazing. GPU-Stat also show the usage properly. Second set of tests on 4k content worked great as well, including VP9 and AV1 SDR and HDR content. To note, I still have HDR tone mapping disabled in Plex since my Computer and main TV support HDR already so no need to tone-map. Third set tests on 8k content was where it all fell apart. Previously, my i7-9700k would struggle along to software decode 8k content then use the iGPU to hardware and very very very choppy. With the Arc A750, it shows it as hardware transcoding now, the audio plays back flawlessly and the seek bar no longer skips and the transcode buffer looks to fill up quickly, but all I get is a solid green video output. This applies to VP9 and AV1, SDR or HDR 8k videos. If I switch Plex Transcoder from the Arc A750 back to the iGPU, it will Software Decode AV1 8k and Hardware Encode to H264 with some hitches as the software decoder can't keep up, but it at least works on all my 8k files. It's just something with the Arc A750 that falls flat on its face when trying to Hardware Decode 8k Content, SDR or HDR.

Same, after the recent Plex Update, I went to watch a show and the episode said "Unavailable" when I first tried to watch it. I just restarted the Plex docker instance and everything worked fine. Must have been some weird bug on first launch after the update not properly setting up the volume maps or something.

Great news

When you hit Apply on the Docker edit screen, what does the resulting pop-up show for text? Could this be related to the empty fields in the Plex Docker config improperly being populated as ":" or something and causing the Docker to fail to create. I remember seeing something like that earlier in this thread and I think the solution was to either "Remove" the empty variables you aren't using, or just put in some dummy paths in there. Luckily form the Docker Tab and click "Add Container" at the bottom, you should have a drop-down list of your previously configured Dockers so you can re-select the Plex one and make changes as required.

I've had that too at local pools and sporting centers while on their Free WiFi. Seems to be them blocking out-going connections on certain ports, I've had luck adding some different port mappings and switching between them, but it's a pain and normally I just revert back to my LTE.

Since they mentioned that they are trying to pass through the Intel iGPU to a Windows 11 VM, wouldn't that cause issues? The VM trying to take over the GPU from Unraid. From what I recall of Unraid VM's and GPU's when I was trying a 2 Gamers 1 CPU setup, most consumer GPU's can't be used by a VM and anything else at the same time. If they don't have the Intel iGPU stubbed out, Unraid will load it and the Intel Driver, IntelTop and GPU statistics will recognise the iGPU, but as soon as the VM turns on and takes over the iGPU, Unraid will no longer be able to use it and if something else like a docker or another VM does try to use it, then that would cause the GPU to crash. If Unraid booted in GUI mode and the iGPU was the primary, then Unraid would be outputting to the iGPU until the VM starts, then potentially would freeze the GPU and from what I remember, on AMD GPU's at least, that would lockup the GPU. And if the GPU didn't properly support soft-reset, it would need a full shut-down at times to recover.

This is the lost wisdom that Folding@Home and BOINC have been dealing with for ages. This was even required back on Haswell days on Windows if you were running a dedicated GPU and wanted to use the iGPU for transcoding video or OBS for some cheap streaming setup. I also just had this more recently on my i7-9700k Unraid Server. I have also have a 1050 Ti installed for Folding@Home and wanted to use the iGPU for Plex HW transcoding so I had to pick up an HDMI dummy plug for the motherboard iGPU HDMI slot. The one I picked up is https://www.amazon.ca/gp/product/B07C4TWZRM but most of them you find should work, just check reviews as always and focus on the non 5-star reviews to look for major problems with them.

There's a check box under Settings > Library for Run scanner tasks at a lower priority that might help you case, I havne't tried it myself though. I have my server set to just add media at 2 AM and do all the scan tasks then since no one is ever streaming at that time. If I really want to watch a show right away I just downloaded, I either use VLC locally, or Add it to Library and wait the 10 or so minutes for just that one show to do its full scan. Not sure on the upgrade from n100 to i3-12100 though, not sure how different the decode/encode engines are in those. But if you are maxing out all 4 CPU cores for the other parts of the scan (chapters, metadata, intro and outro detection, all audio tasks a CPU bound, some HDR I think is still CPU bound), then the upgrade might help, but I don't think it will completely get rid of the problem.

I kept seeing this in my inbox and and didn't think someone was still running Haswell iGPU like me on my kids gaming PC. I can confirm that my Haswell i5-4570s would have its Intel Graphics 4600 show up as Xeon as well before I installed the Intel driver in Windows.

Here are my Unraid settings below. Just make sure when you copy/paste the App Password from your 2FA section of your Google Account that it isn't adding an extra space or tab character at the start or end of it.

Others with the EAE timeout had luck with deleting codecs and either mapping the tmp folder, or fixing its permissions:

I haven't explored other email providers, but since Gmail is free, I set up an email address just for my server to use so that I can easily filter emails on my main email that are sent from the server's email address.

Ah, okay, thanks for the clarification, at least I have the new plug-in (along with your NUT) for now in case I forget to update before 6.12 official drops.

I'm using SimonF's version of the GPU Stats plugin, but for some reason can't get Multi-GPU to work. Does it support showing the stats for noth an Nvidia 750 Ti and Intel iGPU UDH 630? I'm on Unraid 6.11.5 and have the Intel GPU TOP and the Nvidia Driver plug-ins installed. Not sure if I'm missing something here, or it's not supported on 6.11.5 or this combination of GPU's. I can configure the GPU Stats plug-in for either Intel or Nvidia and they will both work for single GPU stats display on the Dashboard, just doesn't show the Multi-GPU checkboxes for me.

Not sure why this started appearing in my Fix Common Problems, but all my Dockers have the "up-to-date" with green checkmark so am I okay to just ignore this "problem" in Fix Common Problems? If so, why doesn't the Fix Common Problems check it any of the Dockers say "Not Available" before worrying users that their servers have problems and prompting them to install extra plugins?

You definitely want your main Plex Appdata folder on an SSD for the best browsing experience. My current Plex library is using ~8 GB in its Appdata folder: - Movies: 2248 - TV Shows / Episodes: 444 / 12269 - Music Artists / Tracks: 31 / 1537 The Transcode directory should only have the current active streams and only for the amount of time you have set for transcode buffer, so unless you regularly have 20+ streams that your server is transcoding, you don't have to worry too much. I think mine with 4 streams once hit about 7 GB of usage, but that is purely me trying to remember something from years ago. With that said, and with Unraid supporting multiple pools now, I personally have my transcode directory mapped to an old Intel 660p SSD on it's own pool just called 'cache_temp' so that I'm not wastefully writing to my good redundant SSD cache pool. You could mount it to the linux temp file system at /dev/shm, but that is in your memory and if Plex does do something weird and doesn't delete previous transcode streams, then you eat your memory fast and cause other issues. The large Plex usage in my experience is mostly attributed to a few things: 1) Video Preview Thumbnails (Individual Library Advanced Settings > Advanced > Enable video preview thumbnails): This takes up a ton of space per video 2) Chapter thumbnails (Server Settings > Library > Generate chapter thumbnails): Takes up a good amount of space, less then video preview thumbnails, but still a good extra chunk 3) Lots of TV shows: Each episode gets meta data and with more scene changes, typically 3 or 4 episodes of TV will take up more space then a single movie

I believe it was related to the Data Breach Plex had earlier this year that required everyone to change their Plex password and had to reclaim servers in a lot of cases.