Faceman

I love it when a solution reminds us to check some of the simpler things first. Definitely not an embarrassing error or oversight, it's not something most people would think to check for.

out of memory errors all through that syslog, check your container memory usage. Particularly chromium, it's showing up a lot in that syslog.

You can use WSL to mount an unraid disk on windows, because WSL is Linux. There is some info from microsoft here

I tried running Plex software transcoding on a 6700K once, which was boosting slightly higher than stock so somewhat equivalent to the 7700k and it really struggled with even the simplest 4K 24 SDR 8bit video. in terms of CPU benchmark numbers, for good CPU transcoding of a single modern high bitrate 4K video you need double the performance of the 7700k, which would be something in the realm of a I5-12400 or higher, but even then CPU encoding is not a good idea at all, it looks good visually but it's slow, high latency, and incredible inefficient. On TV built in apps the trigger for forcing a transcode is often WiFi being unreliable (~60mbit might not be enough, even 100mbit hardwired Ethernet can struggle with some files and most TVs don't have gigabit) or subtitle formats needing to be burned into the video, even for forced subtitles in many cases. Especially PGS (most common bluray subtitles) and ASS/SSA (advanced subtitles with vector graphics and animations). Generally, even on higher end modern TVs the built in apps are deeply flawed in some way or another so an external player, ideally one with faster Wifi or Gigabit Ethernet, is recommended for a much more reliable experience.

Gemini isn't a person and is just repeating words that make sense with zero fact checking, what it is saying is based on a much older Intel comment that warned all 65w chips may be affected. That has been further clarified to a definitive list of chips starting at the 14600K and up (14600 non-K is unaffected). Evidence of chips outside of the 14600K and up are not significantly higher than baseline failure. The Raptor lake flaw does not apply to the 14100 as it uses the 12th gen IA power delivery silicon which has shown no firm evidence of degradation above the baseline for the 12100 and 13100.

these days its generally better to use DHCP reservation in the router than setting static IPs in most cases, especially static within DHCP which leads to conflicts depending on the router.

Your Motherboard bios is very old (7 versions out of date) and is missing the stability and longevity modifications made by intel in an attempt to "save" the faulty Raptor Lake CPUs, so I'd be wary of that as using raptor lake CPUs especially K models without those microcode updates will permanently damage the CPU over time. The 13/14700k and 13/14900k being the most commonly affected models in that range. Before doing any other troubleshooting I would update the bios, but if you are affected by that issue the damage may already be done and the CPU will need to be replaced under warranty.

Just wanted to chime in and say that a simple memtest is not enough to identify a memory issue, you need to run it for 12-24 hours on a continuous loop to catch deeper instabilities. I doubt its related to the raptor lake instability issues as the 14100 is not a true raptor lake part, it's a refreshed alder lake core, effectively a slightly clocked up 12100/13100, which as far as anyone can tell were unaffected. I'd be doing some testing with the bios starting at default settings, all gaming options off, xmp off, power limits back to default intel spec etc. Some gaming motherboards have "enhancements" enabled out of the box that could push some cpus too far. Then I'd reseat the cpu and memory, remount the cooler etc. usual hardware checks.

Don't know who, their moderation actions are anonymous so you cant know who removed a post unless they specifically tell you. Not attacking or pointing the finger, the mods have been doing a good job overall, but I think that page should have a clearly defined rules list like most of the other related subreddits and mod actions should be given explanations. The problem for the OP of this thread is that they arent given a reason, so they don't know how to reframe the question or if something needs to be censored or asked elsewhere. If asking an application specific question isn't allowed it should be clearly stated, but I don't think that makes sense. if you're running a specific app on unraid, the unraid subreddit should be a valid place to ask for help as long as it is within the other rules (which at the moment are not clearly defined)

How many disks and what type of ZFS pool? You cant use the extra space in a ZFS pool until every disk has the same capacity, you cant freely mix and match drive sizes like you can in the main array.

I'd try uninstalling the Folder View plugin first, when it's playing up it can take over that part of the UI, that version is also abandoned so no bugfixes.

Ancient 2023.12.19 version of Mover tuning doesn't work with 7+, you will need to remove that, get everything in order with the stock mover then you can install the updated mover tuning plugin available in community applications.

I believe the WD60EDAZ is also SMR, so it would have been a bottleneck to write to in some cases too, nowhere near as slow as the 5TB you removed but still measurably slower than the higher end drives you have in other slots. The next slowest are also obvious on the disk speed chart you posted, so target those as the next upgrades/removals.

One of the new mods there is going a bit far, I've been noticing it since they were added earlier this year.

Not really an expert, but your logs have a lot of errors for: Disk sdm (WDC_WD20EARS-00MVWB0_WD-WMAZA47938) throwing a lot of errors usually related to power/spinup issues or a controller/cable problem. a ZFS disk/pool with many errors not mounting, this is likely the main culprit here, appdata,system and domains are looking for "cache" but your only cache disk is showing as a new device. Nvidia driver needs to be changed to legacy mode as your GPU isn't supported in the new drivers. One of your "disks" is a hardware raid array of some flavour, which isn't an error but i wouldn't recommend it. I'm sure someone will help troubleshoot this for you.

WDC WD50NPZZ-00A is a 5TB 2.5" 5400RPM SMR drive with only 8MB of cache. It is just a very, very slow disk under extended loads and when closer to full it has to perform SMR writes, which can slow the the KB/s range. Go through each disk and check if they are SMR, they should be your targets for future upgrades.

Oh, and just a note, the meter on the dashboard is in GibiBytes, so if for example you set any of these variables to "8000000000" you will actually be setting 7.45GiB so keep that in mind if you need to see exact values in order to sleep soundly.

once two of the old 2TB disks are upgraded to 4TB through the normal disk replacement/upgrade procedure for each disk one at a time, you should have more than enough free space to run the Unbalance plugin to scatter the data from the remaining 2TB disks to the new ones, then you can run a disk removal procedure. Those steps are just as you have mentioned in the original post, but it is easier and safer to upgrade the existing disks first, less risks that way as we keep the parity valid and protecting you through the whole process. You can also upgrade a single disk, then unbalance to clear one old disk, then remove that old disk, then upgrade the second disk and repeat if you wanted to do reuse the old disks slightly faster in something else. There are two ways to remove disks from the array once they are empty and the data is confirmed safely on the new disks, the easy way is to go to tools>new config then just don't map the old disks when you re-assign the disks, but that will require a parity rebuild. the other method zeroes out the disks to ensure the parity stays valid. both are details in the link above.

There are many plugins for optimizing unraid for this sort of usage, but I cant see limetech doing much in this regard as it is only one of many niche use cases for their software. They want to keep it as simple as possible and let the community plugins tweak it for these specific cases and only add those plugin features when the majority of installs are relying on them, though i do think they could add some of the features of mover tuning like moving by age instead of doing a full cache flush on a schedule. I just think it makes more sense to use the cache like a "leaky bucket" this way. Funnily enough that Reddit thread has me as the top comment, but things have advanced quite a bit in the last few years. There is plugin called Cache Mover by Alturismo that can pull a series folder from the array to the cache when an episode is played. This can pre-load the cache with a series when a binge watch starts and pretty much does what you are asking. he also has the Simple Mover, which offers media specific mover tuning features. Then there is PlexCache, a set of scripts that can look at things like "up-next" and "watchlist" in plex and preload files to cache based on that. This works very well but takes a bit more setup. Then there is the Mover tuning plugin, that can be used to keep recently added media on cache until space is needed, minimizing the need to touch the array disks at all. The old Plex Preloader script I mentioned in that Reddit thread does still work, and if you have plenty of ram and not a lot of media it is pretty much a perfect solution, but it doesn't scale well to very large libraries and only solves the disk spinup issue.

You have to force it to HTTPS now, so you need to adjust the advanced settings in the template so that the webui link goes to the https address and port. That's fine, but selkies is awful for this and has massive latency unless you force a low quality MJPEG stream, perhaps we can give it hardware acceleration access to improve it?. I know selkies is supposed to be technically excellent, but in this case i just want to access a web browser on my local server for some specific use cases, I don't need a feature rich gaming and remote work kvm system for that.

Could this be the (weirdly common) IPV6 devices being public and completely open without you knowing issue that many routers have by default? Actually finding an open IPV6 service is quite hard as there are an incredible number of them, but there are so many devices in people homes just running completely open to WAN, no NAT no Firewall and that's just how it is with IPV6 on so many routers, and the more ISPs enable IPV6 by default, the more this will become an issue.

Thats where I've always done it.

if the pcie lanes are connected properly and signalling cleanly it should "just work", regardless of the adaptors or connectors in between.

The 13500 is technically an Alder Lake refresh though, not true Raptor Lake which was an updated architecture and had more cache, the device ID for the 13500T's version of the UHD770 is the same as the 12600K etc. which is 0x4680. I don't think there are actually any real-world differences in the iGPU architecture at all though, it's all UHD770. The 14600k for example has a different device ID for its version of the UHD770, 0xA780, that does show as "Raptor Lake-S GT1 [UHD Graphics 770]" despite having the same iGPU specs, same CUs, same Video Engines, Clock speeds etc... The 13600T and higher also have the newer Raptor Lake iGPU Device ID, but not the 13500, or even the 14500, they have the older device ID as they are both technically still Alder Lake refreshes, not Raptor Lake.

Looks like setting a minimum works just fine in Unraid 7.1.4 options zfs zfs_arc_min=8000000000 options zfs zfs_arc_max=16000000000 Correctly sets both the minimum and maximum and the system works within those bounds. I haven't tested setting the minimum and max the same yet as this server is my main production server and I cant reboot it willy-nilly, but I cant see why that wouldn't work according to the OpenZFS documentation and some threads on other ZFS forums