_cjd_

Should be fine. I moved without issue the other way (chasing power savings).

Older SATA specs were slower. Heck, 10k rpm SCSI disks were slower. Even now we see incremental improvements in speed across the platter each generation.

HDD speed may limit you slightly with 250Mb internet service. I saturate RAID10 SATA ssd on 10Gb Ethernet (internet is far slower) but just... Close enough I'm not trying to figure out how to use m.2 for SSD cache. Single drive write is obviously slower but still beyond 1Gb speeds. Here's the thing though - you can always add cache later. 24 disks is going to be power hungry and expensive - why so many? Up against your concern over internet costs that strikes me as odd, but maybe that's just me.

Intel igpu may cover your transcoding needs, and the lowest power consumption builds are also Intel. AMD won't do quicksync if I remember right.

Also yes, you could swap it out and then use it for VMs. I have an nvme I use for docker/VM and a separate set of SSDs as write cache (also primary storage for files needing fast access, backed up to the main array)

Oof, yeah that's a lot and certainly all over the place on time. I usually see 22hr for a 16tb array where daytime use slows things down, but also a total of 5 drives in the array (mobo or asm1166 card connected). Time machine is the killer. No need for me to stop docker (no VMs any more). It's probably not even parity drives specifically, if you're saturating the card/slot. Are your VMs/docker running off the array? Curious what you have there with sufficient io to slow parity check.

If you are using an HBA card, actual drive speed should be the bottleneck. You can verify performance with the DiskSpeed application if you're curious. Rather than asking vague theory type questions: what's your current parity check time / drive size / speed results? Are you asking because you think something is wrong?

A couple thoughts - look into MTBF and lifespan numbers for the SSDs you choose. And consider whether you'd ever want to set those up in mirror for redundancy. Consider network speed against cache drive speed; standard SATA SSD may be adequate (I can saturate a RAID10 4 drive sata ssd array on 10gbe... without quite saturating the network - probably a useful thing). You may not need NVME there. Be EXCEEDINGLY careful buying drives from Amazon. They may not actually have much in the way of warranty. I've largely stuck to B&H, though I'd also consider new from ServerPartDeals based on feedback here. I've used Newegg as well, but with care. 16TB seems to be the sweet spot on capacity vs cost in the upper range, but only by a little. I'm also the nervous type and with drives that big I've gone dual parity.

Nothing wrong with onboard SATA and a basic asm1166 based adapter card or so. I have 10 SATA devices including 5x 16tb on an m.2 1166 adapter (other 5 to onboard) and parity checks consistent with onboard SATA or lsi HBA speeds (22-24h depending on use during the day... Time machine doesn't care, I hate it for it's constant chatter). I've tried all 3 setups, this gets me the best power efficiency.

What's the power supply in your computer? I'm making a couple guesses here but... outside chance it has a problem which is quietly putting voltage to ground/earth - as soon as you unplug the battery backup from the wall, you lose ground. (the assumption here is how you're cutting power to test) You might be able to test this theory by instead shutting off the breaker for that circuit rather than unplugging. I am only familiar with US wiring code in case that matters. If the battery is not plugged right into the wall already, try that too.

nginx needs to proxy to an externally reachable IP - furthermore, it'll need to be a unique port. it's just a proxy, not a router. In your setup, those services should be 192.168.5.11:PORT in either host or bridge mode for the nginx settings. Is that not working? I have esphome set up in host mode and it's working via nginx. I will note - I'm using macvlan for docker and most services get a lan/vlan assignment and a dedicated unique IP. Most services are only proxied so I can map :443 to the random port stuff runs on (esphome is :6052 for example).

Epic seems the path to me - more affordable than Threadripper, gobs of available lanes. I haven't looked hard but nothing with that many lanes from Intel in a similarly affordable package. I've seen some good deals on eBay from seemingly reputable places. Power efficiency may not do as well as a consumer board. If you only need a little more, something like the SUPERMICRO MBD-X13SAE-F-O may do the trick. Won't be cheaper up front but might do low power idle better. That's speculative though.

Another option is to use the User Scripts add on, set up the rsunc there and kick it off. It'll run till it's done and give logs etc. no terminal to worry about etc.

I've never seen it this cheap which is barely more than the low end x570 gamer oriented boards were not that long ago ... but certainly there are options, you just have to weigh needs against what is available. Also remember to count pcie lanes, not just slots. There's a limit on consumer/desktop oriented parts. rj45 10gb was not really helpful for me as I wanted SFP+, but also I have no need of a video card. In the end, I also amortize price when I'm budgeting stuff like this - lots of storage isn't cheap either, but it probably has a longer life of use than my the average gaming setup. It's just the prioritization and balancing game, and for every person it plays out differently. I'd probably pick parts slightly differently today too, based on other things learned since I initially put this server together.

ASRock rack x570d4u - only add ones for me are m.2 sata adapter (likely will change to pcie and mirror nvme some day) and x710-d2 SFP+ 10gb network card (DAC to switch). I ran an HBA for a long while but changed to improve power efficiency.

MTBF and such are how I've chosen - usually one size in a drive family has the best numbers: specific size varies by family.

Logitech media server here as well, whole house full of squeezebox devices but software player awesome too. Doesn't touch your music files either, just indexes. Some great plugins.

Unraid lives in memory and loads off the USB stick. No need for that 980. Check around to be 100% on the drives, but as long as they're CMR and have a good long warranty that should be good. IronWolf and IronWolf Pro are different, but I haven't researched recently enough to remember why my memory says "only the Pro NAS drives, or just go Enterprise". I only have 16tb UltraStar or IronWolf Pros at the moment. Seasonic is my go-to. Nice long warranties and an even longer track record of excellence.

Everything I've read says not to use consumer grade drives. Jump for enterprise or "pro" Nas drives. I'm super particular about power supplies, but the one you've chosen won't do you any favors for power efficiency.

Intel seems to have the edge on absolute minimum power draw builds... The question is whether that'll be the case for your build, especially with so many VMs. I don't have specs for latest Gen memorized but I'd also consider pcie lanes and available ports. That swayed my decision. I needed mATX when I had to choose so ended up with an amd 5600x - mean idle is still ~45w with drives spun down, no VMs, and a USB device drawing a few W (power meter reader). Home assistant docker adds a few on its own. Only 32gb ECC so even two windows VMs at once becomes a challenge. Used to run HA as a VM, more stable as docker for me. IPMI also adds to my power draw. Multiple VMs may prefer the AMD but not sure how the Intel core strategy shapes up with heavier virtualization.

Looks like it's unable to open the output device. It seems none is specified. I'd start there. Haven't used this docker myself but use it on the desktop and it doesn't start unless all the settings work - assuming that holds true here...

Intel with integrated GPU certainly seems like the right place to start - I don't run Plex but recall reading that it's the place to start in this scenario. With the right parts I think Intel has the edge on ultimate power efficiency at idle; whether your system will idle is a different question. I have a different set of requirements so haven't researched in this direction enough to point you at anything more specifically. My main server started life on an AMD 965 Black, spent a little while on an older i3 board (which is now my backup Unraid server) and landed on an AMD 5600x + X570d4u; I was looking for IPMI, ECC support, and adequate pcie lanes and sata ports...

So... it's not JUST a "you" problem - you're definitely not alone in the struggle trying to do something in a totally different way than you're used to doing it. Doesn't change the frustration in the moment whatsoever, but it can help when you take that deep breath to re-orient and plan next steps. I did a (multi level) directory copy with a simple drag & drop in Krusader. I had to tell it I wanted Copy. Timestamp was correct on each folder as well as individual files. rsync is amazing when you have it sorted properly. That's a "when" with a bit of a learning curve, and internet "just do this" don't always help you understand the specifics which may actually matter. I couldn't automate my Unraid backup without it; the backup server is powered off most of the time, and its ability to not just transfer every bit over the network when not required is quite nice or it'd be days of backup (hard drives being the throughput limitation). I don't know what krusader does under the covers, to be honest. mv vs cp vs rsync and they all work differently. You definitely do NOT want to be moving files. But rsync is the right tool for your particular needs if you're working in terminal. Comfort with command-line and reading man(ual) pages is worth doing, regardless - even in Windows. As a server, stability is as much down to hardware choices and setup as anything. Unraid builds in some tricks Windows doesn't offer without significantly more setup work (if it's even an option), without a doubt. If you'd gone with another *nix NAS option you'd probably be dealing with similar struggles. Windows is better at glossing over the details for a lot of things the "average" user is likely to do, but it also limits quite a bit of what you can do as well.

First things first: I think Krusader will do what you want it to. I just installed it (Krusader) to do a test, because the timestamp change you mentioned recalled an observation I had with rsync. It maintains timestamp correctly, copying from an Unassigned NTFS mount into the array. HOWEVER - this is only reflected when it completes everything. While the copy is active, the timestamp is the ~now. For a large copy, it may reflect the wrong timestamp for quite some time. Second: Unraid is not the reason for your troubles getting files copied. An unfamiliar system is - it happens to be Unraid, but fundamentally this is a *nix challenge, and particularly a command-line thing. GUI usually bakes in assumptions - and you are familiar with those baked into the Windows File Explorer. Command line has none of those. It can be frustratingly precise. Setting up rsync for my backup Unraid I made a number of similar oversights getting the path exactly right. Correct and move on. You can also do things in smaller pieces, or at least do a test on your command over a smaller data-set to verify it's doing what you want before taking on the large copy.

I've been using Duplicati to make backups of critical data and it does provide similar "timeline" capabilities as Time Machine... not nearly as slick an interface but it gets the job done. At the moment all "critical" data also requires fast access so is on RAID 10 SSD, not the parity protected array, but there's nothing to prevent writing such data to a single array drive and using Duplicati to replicate to another array drive (that's one of the things I particularly like about the default Unraid array setup). I also have a backup Unraid server (scheduled rsync) and external backups of things as well. You're taking the time to verify results, which is something a lot of people forget to do. For the rest - keep calm, keep learning. Test, verify, and leave yourself an exit strategy. And do more work in terminal. One thing I learned about rsync is that it's weird (or at least, not obvious) about some things - I had to go back and add -H (linked files become actual in the target otherwise) and -S for sparse files as my backup was a couple TB larger than source... so even when it works, there's always something new to learn. My biggest panic so far was updating the encryption key on my array. My oversight was forgetting how add-on scripts work in Unraid - I thought I was setting up a non-destructive addition of a new key, but it was a replace. Even still, verified everything and all good... till I went to start the array. Tried everything I could think, no luck. Finally decided to risk a reboot and everything started up perfectly. No idea why it needed a reboot. When you start doing this kind of stuff, you *will* make mistakes as you learn. I've had to jiggle cables to fix disk speed issues and early on, bad memory causing all sorts of weird; when setting up my backup server, two hard drives up in (literal) smoke (fortunately NOT the brand new 16TB drives... I was using stuff I had lying around to test the automated rsync before buying new hardware) - probably the power supply (it worked fine with one spinner, but wouldn't even spin up two; was fine with 3 ssd). This is why you keep backups. Next time it'll be a NEW problem. I've also done more stuff without problems than stuff I've had problems doing.