testdasi

You didn't ask which CPU is the best CPU. You asked which CPU is the best CPU for you. Your post had quite a bit of focus on power consumption, which makes the i5 a better fit. Media server (e.g. Plex) hardware transcoding is cheaper on the i5 due to the iGPU. Doing hardware transcoding on the AMD build would cost more to get the right GPU. No mentioning of VM uses, which lessens the advantage of the Ryzen higher core count. In short, I was focusing more on fit instead of being a fanboy (despite running a TR4 server myself). If you haven't noticed, I picked the AMD build for your question 3 i.e. it's a better long-term choice. Whichever build you select, I would recommend you to get a Gigabyte motherboard. If you ever try to pass through a GPU to a VM, you will thank Gigabyte's "Initial Display Output" BIOS settings. It doesn't guarantee you can pass through the GPU (nothing does!) but it surely makes life a hell lot easier. And one last thing I noticed, you picked the 860 Pro in one of the builds. It's a waste of money to get the "Pro" version for home uses.

My answers: For similar price, I would pick the i5. Your low power build isn't really that low compared to the i5. You are overthinking it when you start to consider idle power consumption because it is generally very low - low enough that if you are concerned about it, you are better off saving money for more critical items, like food and water. The AMD build will consume more power, just from the fact that it requires a dedicated GPU (which by the way, you have not included in the build cost) The AMD build 3x8TB. HDD failure follows statistical patterns so the fewer drives you have, the lower the chance that you have a failed drive. There is no "best", there is only "suitable". You need to consider how much data will be on the cache pool. And generally, you should go for higher capacity SATA rather than lower capacity NVMe. The real life diff in performance between SATA and NVMe is pretty negligible outside of pretty niche workloads. Define "reliable"? Available space doesn't need a calculator to be honest. Remove your parity drive then add up the rest. You can do that in your head. See above point about AMD build needs dedicated GPU That is correct.

What do you mean "update bios of card"? Or do you meant GPU drivers? Note that new AMD drivers have been shown in some cases to cause issues with pass through so you might want to try an earlier version to see if it helps. Try something that came out in 2018. You need to access the VM through VNC to uninstall the current driver and then do a clean install of the older driver.

That's why I don't recommend using NTFS with Linux. I had corrupted data about 3 years ago and stopped using NTFS on Linux completely since then. Don't know how UD NTFS is done but presumably ntfsfix can be used to fix ntfs filesystem but it's not the same as chkdsk. Since NTFS is a Windows file system, I would prefer to use Windows to fix errors to safeguard my data.

100W = 0.1kW = 0.1kWh / hour = 2.4kWh / day = 876kWh / year * $0.108 / kWh = $94.61 $70.96 for 75W That is obviously assuming your GPU is running max out all of the time but without overclock so the cost figure is closer to a worst-case-scenario than anything. Based on the reviews I can find, the 1650 idle power consumption average is a hair above 8W so I would round it to 10W just for simplicity. So if you estimate to be gaming for about let's say 2 hours a day then your cost estimate is closer to: $94.61 * 2 / 24 + $94.61 / 100 * 10 * 22 / 24 = $7.88 + $8.67 = $16.56 for the 100W Not much. A donut a day.

I had some corrupted data about 3 years ago.

Edited my original post. I always format a drive immediately after preclear without even thinking about it.

Your post has quite a bit of ambiguity which makes it a bit hard for me to understand the problem. Your screenshot only has 1 drive. Copying requires 2 drives - source and destination so is your screenshot i.e. "1TBWD" the source or the destination? Why no screenshot of the other one? Did it disconnect? The "folder where I transferred the files" - you meant the destination or the source? "free space and used space have remained the same" - you meant the free space / used space of destination after the transfer? How much data did you transfer? Are you sure the transfer completed successfully? "folder is still empty" - which folder? A subfolder under \\[server]\share-name\ is empty? or the \\[server]\share-name\ itself is empty? What is the output of this command from the console? ls /mnt/disks/1TBWD/ By the way, it's not recommended to use NTFS with Unassigned Devices on an on-going basis. NTFS + Linux don't play well together. If you are going to use the drives through Unraid UD, you are better off using a native Unraid file system e.g. xfs or btrfs. I personally use btrfs for my external drives that I access through Unraid UD because, as a last resort, I can still open them in Windows with the open-source (i.e. free) WinBTRFS software.

How is your NVMe passed through to the VM? via PCIe method or via the sdx method? If you are using the sdx method for the NVMe, have you even attempted the PCIe method? As I mentioned previously, anything other than the PCIe method will to some extent lag under heavy host IO. In addition, your CPU core assignment is not the way to do it. By sharing core between torrent + plex + VM, what do you expect to happen when 1 of the 3 is hammering the CPU? Core 8 and 0 are a pair. They both should not be used by Plex and torrent. For best VM performance, you want to set up CPU isolation to isolate the VM cores for VM use only. You probably need to watch more SpaceInvader One videos on Youtube.

Are you getting that error from Windows (cuz I'm pretty sure that's a Windows error)? Try deleting the file from the console.

Can you attach the screenshot to the forum post instead of using link please.

I used pipework docker (details below). There is the proper Limetech-recommended method using docker native functionalities which I found too troublesome to even try since pipework just worked and took me less than 5 minutes to set up. Nevertheless, I have now completely transitioned away from custom IP and into using +VPN dockers and VPN proxy dockers. There isn't any docker that I want to be behind VPN which doesn't have either of those 2 options available. Pipework docker.

Unraid and lower power can go together fine. Gaming + low power is like trying to do an F1 race with a Fiat 500.

A docker or plugin creates those since I don't think they are native to Unraid. For example, my Calibre docker creates a Trash folder (but never uses it as far as I can tell).

What do you mean by "both passthrough? It's not possible to do both since PCIe pass through will make the NVMe unavailable for other method. When Plex transcode, is it doing in the VM or in docker? Are the docker core overlapping with your VM core? Beside the NVMe, what other disks / vdisks are used with the VM?

Pay very careful attention that You don't accidentally assign any disk into the parity slot You don't accidentally run preclear on a disk with data you want to keep. You don't click "format" on disk with data on it. (If you see "format" as an option to progress, something is wrong or you have done something wrong because it's never part of any of the steps I forgot that I always format my disk right after preclear so there is a step with Format on it). While it's actually rather hard to accidentally do those things, don't underestimate anyone's ability to do silly things.

Your best bet is to try custom kernel (at your own risk) with the navi patch - see quoted posts at end of this reply. There is no other fix for the AMD reset bug (which essentially means your GPU can only be powered on once per boot). You can buy a cheap low-end GPU (something in the $30 price range) and set that as Initial Display Output (it's in BIOS settings) so Unraid boots with it. Them dump your own vbios of the 5500XT. That will give you a fighting chance (but NOT GUARANTEED!) to be able to pass it through. With an AMD GPU, you are still likely to have to reboot the whole server to restart the VM due to that same reset bug. This option can be done together with the other option. The 3rd option is to go Nvidia but again no guaranteed. You may still have to get a cheap low-end GPU to dump your vbios, but the key diff is you probably don't need it to be reserved for Unraid to boot (e.g. you can borrow a GPU from a friend just to dump the vbios). I am currently running my 1070 as a single GPU after dumping its vbios using a GT710 as the low-end. The fix lies with AMD and I have not seen any indication that AMD is offering any perm fix (and hence no perm fix with Linux kernel - and then you have to take into account Unraid will never run on the latest bleeding edge kernel). So your next best bet is for a navi patch that doesn't cause regression (i.e. break other stuff) for other users. That patch is again not available at the moment. TL;DR: you will be waiting for quite some time. Some resources for both of you regarding custom kernel with navi patch. Try it at your own risk, as always. (the best option is to compile your own Unraid with the navi patch).

You can install the OpenVPN plugin that would put your entire server behind VPN. Otherwise, you have to do it on your router (e.g. have each docker with its own IP and control the VPN access from the router) and/or router VM. I used to have a VM that is connected to VPN and serves as a gateway for some dockers to route traffic through. I don't do that any more since I found it rather cumbersome. Generally the dockers that I need to be behind VPN already have a "+VPN" version available on the app store.

Then it's pretty easy. Connect 2x new HDD's to your server, as well as the old HDD. Install the preclear plugin from the app store and run it on the 2 new HDD's (which should show up on the Unassigned Devices list on the Main page). This step will take some time. If there's any reallocated sector / pending sector / other critical errors at the end of the preclear, start your RMA process instead of continuing (i.e. you have a faulty HDD). Note: the exception is CRC error which is usually cable related. Stop the array, add the 2x new HDD's to the array (disk 2 and disk 3). Start array. Install the Unbalance plugin from the app store. Go to Settings -> Unbalance to enable it and find the port it is using. Then point your browser to http://[ip of server]:[port] to access it. Use the Unbalance plugin to distribute your data from disk1 to disk 2+disk 3. The interface is very self-explanatory. Wait for it to finish. Verify that your disk 1 is now empty and disk 2 and disk 3 contain your data. Stop array. Tools -> New Config -> tick cache (so cache slot is MAINTAINED). Then go back to the Main page, it would now allow you to assign just the 2 new HDD's to the array. Then start array (now with only the 2 new HDD's) and reverify that your data is still there. (Optional) Your old HDD will now in Unassigned Devices list and you can actually run a preclear on it to wipe it so you can resell / recycle etc. Shut down server and disconnect the old HDD.

You may need to balance your btrfs pool. Read the FAQ which johnnie mentioned how to do it. That's one cause of spurious out of space error. Another possibility is that a RAID 10 with 4x1TB and 2x4TB to give 6TB available space is just not a viable option. Doing RAID with uneven drive sizes is generally not a good idea. Traditional RAID doesn't let you do that. I think your pursue of max speed with (presumably SATA SSD) is heavily misguided. What "high demand function" are you doing? A game drive does not at all offer any perceivable improvement going from SATA to NVMe (or as I have tested, even SATA to Optane 905p!)

Are you able to get both new HDDs and the old HDD connected at the same time?

I can see the mergerfs docker under Advanced View of the Docker tab on the GUI. It's greyed out like orphan but it isn't deleted by the orphan deletion script so I manually deleted it using the GUI.

How did you "put it as primary"? Have you tried if it would work as secondary?

You have to turn on Syslog mirror to flash (settings -> syslog server) to see pre-reboot log. Try clicking on the black icon with plus sign on the right hand side of the sdx text. The mount button may be enabled there. If it doesn't work then you might want to try checking the drive in another computer. Random power off could be an issue with your PSU.

It's the same as your using of Intel IGP with Plex docker but it's for Nvidia GPU instead. The problem is that Nvidia drivers are quite troublesome to compile with Unraid (just look at the Unraid Nvidia plugin topic + LT post above about VM pass through implication) while Intel IGP tends to sort of "just work" with kernel updates (with VM pass through implication not applicable / ignored).