gubbgnutten

SATA 3 as in the motherboard port labeled as such? That has absolutely zero connection with md3, which is always disk3 in unRAID. Disk numbering in unRAID is based on how they were assigned manually when setup, not how they happen to be connected.

Can't see any pictures, Microsoft just asks me to log in. If you use the console, execute "diagnostics". It should save a diagnostics zip on the flash drive. Attach that one to a post.

Seems like you are panicking and trying things randomly. You are probably making things worse... /dev/sdX is not guaranteed to be consistent between boots. Did you identify the drive based on model/serial number or just sdd? If md3 is your parity drive, assignments are messed up. And parity does not have a file system, so if that is the case it shouldn't be "repaired" by xfs_repair. Also, do you understand the implications of working against /dev/sdd?

But in the case were one sets up a new array, and all data disks as well as the parity-to-be have been pre-cleared, the exact contents are already known - zeros! No file systems on the data disks yet, formatting is later in the process.

Sounds like you have a fully functional setup at the moment, do you have any reason to mess with that? The main reason to run OpenVPN on a full computer instead of a consumer router would be speed. A full computer will likely be able to encrypt/decrypt more bytes per time unit than the router. But do you have a need for faster VPN connections? And does your internet connection even support faster connections, or is it saturated already? No idea what speeds you get from the N7000, what your server would be capable of, or the specs of your internet connection... Another potential reason would be the ability to update the OpenVPN server software independent of the router's firmware releases, should some horrible security hole be discovered. And yes, you would have to setup port forwarding.

What do you mean by "duplicate"? And have you emptied the share?

If "Minimum free space" for a share is set to 3000000000 it means at least 3TB of free space is required to allow writing to the share. Did you mean 3GB? Write that, include the unit.

In the diagnostics, shares V---o, s----m and d-----s were also set to "prefer". Well, a-----a too. But I assume that is appdata which usually should be prefer. If you check the settings for the downloads share, which field contains 3000000000?

I think you are right. It used to be that I looked for the combination ReiserFS+nearly full disk as a potential problem, but nowadays I cringe just seeing ReiserFS disks present. Simply too many reported cases of symptoms magically disappearing after migrating to another FS...

From syslog in the first post, yes.

Safest option would be to set up a new share and then set only one specific disk to be included. That would keep files on one disk without exposing disk shares and risk mixing them with user shares. As noted, any individual file is only stored on one disk. Multiple files in the same folder may however, depending on settings, be written to different disks.

Since you can SSH to the box you can try to grab diagnostics from the command line by executing diagnostics The diagnostics zip should then be saved to your flash drive.

Was more thinking of the downloads share. Posting the diagnostics zip would be a more convenient way to supply logs and settings, see Need help? Read me first!.

Sounds like you have use cache set to "prefer" instead of "yes".

Well, the phrasings "SLI GTX 970" and "one GPU for Linux OS and then the other GPU for Windows" kind of implies two cards, or maybe it is just me... I would also assume "my KVM switch" means a traditional KVM switch, and not anything VM related. If the two cards are separated and passed through to two different VMs, I don't see any reason why it wouldn't work. Then again, I haven't really played with that yet.

I wouldn't use the disks in an array in the meantime, there are simply way too many potential situations that can cause writes to them... If you need access to the data while maintaining the hope of rebuilding one of the failed drives, make sure any access to the disks is 100% read-only.

The diagnostics zip, see Need help? Read me first!.

Five seconds is absolutely noticeable, just long enough for someone to start worrying that something might have crashed. Absolutely. I used to plan split levels for shares to keep stuff together and ended up having to check available space all the time... Allowing files to be split as required is way more convenient! I'm usually happy with High Water and just rearranging a few files once or twice a year. Way less hassle for me, and typically fewer disks spinning at any one time.

One reason would be binge-watching. Having the full TV show on one single disk ensures that the next episode is not on a spun down disk.

Check the contents of /boot/config/ssh, do you have zero bytes keys there?

Try using another web browser when adding the cache drive next time.

Sounds like OpenDNS. Famous for extending their DNS services with crap that breaks stuff. Edit: You might need to reboot or wait for cached results to expire.

There are plenty of users complaining about non-existing addresses resolving to that particular IP, so I would suspect either some crappy security software gone awry, or malware. Unlikely to be due to the ISP since the Windows 7 machine can resolve things properly. Compare the DNS settings of the 10 and 7.

There is a bad DNS server configured somewhere. Either on your Windows 10 machine, your router or your ISP. What results do you get with ping on your Windows 7 machine?

Hmm. Don't like that at all... Why tower.Home? If you run ping i.dont.exist does it also resolve to 173.45.161.113? Is something appended to it? Edit: Don't have Windows 10 anywhere, so I can't be more specific, but I would verify the DNS settings (server, suffix, everything) and additionally make sure there are no toolbars, crap security software or other malware interfering with the settings.