Wody

Since the data is the most important thing, so keeping that safe is the best thing to do. So, I'd turn off autostart for the array and any VMs and dockers that may still be on your old USB, then move the USB and drives over to the new system, start it up, and see if it runs. If everything seems okay, start the array. To verify everything is okay, I'd run a non-correcting parity check next (e.g. disable 'write corrections to parity' and start the check). Assuming everything goes okay, your data is now safe on the new system. Next i'd make a backup of the USB (click on Flash then flash backup). Also, make a screenshot of the drive assignments, and copy the keyfile from the USB, and I'd copy all that to the array as backup. Next, shut down the system, and use the USB creator to erase the USB and put a fresh copy of the latest unraid on there, and put the keyfile back. Start the system with the USB, to make sure it works. Now it depends on what you want to do, assuming you have new drives and don't want the old drives in the array, there's two ways to go about it. The first is to make a new array with the new drives, and then use the unassigned devices plugin to connect to the old drives and copy (not move) the data. This would leave the old drives and parity protected, so if something happens with the new drives, you can put the old array back. The other way, is to assign one of the new drives as parity, and the rest of the old data drives as data, and let it rebuild parity. and then replace an old drive or add a new drive as data, and then move over the data to the new drive, then once all the data is on the new drive(s), make a screenshot of the drives, do a new config, assign parity and the new drives and remove the old ones.

On the right and above of the array/pools is a blue icon, if you click that it switches between speed per second, and total number of reads/writes, which only change if something happens, as you found out. It can be confusing if you're expecting a big number of read/writes and see the speed being 0. So check if it says /s (for per second) or not and check the icon. A long time ago, it didn't matter what slots you used, but at some point the video cards became faster than the slot (like PCI) was capable of, so you got special slots for the video card like AGP that were faster. Then PCIe came, and it's supposed to be universal, so I don't know why graphics cards only PCIe slots even exist, but they do. In your case, it says it in the manual, so that's how you can know, but if it doesn't say anything, you should assume you can put whatever you want in there (as long as it follows the PCIe standards of course). So you can put both the graphics card and HBA in slot 3, or 1, or whatever, traditionally the slot used with a graphics card would be closest to the processor, which is slot 6, and in your case that would be the best one to use too, because it's an x16 slot, while HBAs typically are only x8 so slot 2, 3 or 4 would work best for that. Since slot 3 and 4 automatically switch between x16 and x8 x8, I'd put it in slot 2 for a more direct connection (but for HDDs it doesn't matter). Slot 1 is only an x4 and linked to the PCH (platform controller hub, or the chipset) so typically you wouldn't use that unless you have to, or for low-speed things like a card with extra ports (like 2.5gbit networking).

Moving the appdata and other files such as dockers and VMs to cache means they aren't stored on the array but in a pool, which means they don't have the parity-protection of the array, and that's what it is warning about. So this is expected behavior. You can always make a directory or share on the array for backups, and then copy the files from cache there when not in use (but rename the folders something else on the array, for example bappdata instead of appdata so it won't try to use them instead of the files on cache). Then when (not if) the cache drive fails, you can get a new one and put the files back on the new cache-drive (or temporarily copy/move them to /mnt/user). Don't forget to enable TRIM in the scheduler, and set autotrim on in the cache-settings, if it isn't yet. If TRIM isn't enabled, it'll lower how long the cache-drive lasts.

I'd start it up to verify that there really is nothing you want to keep. If there isn't, copy the key-file to a new location, and put a fresh copy of unraid on the USb and put the key-file back, then you can move everything to a new system or boot it up on the old if you want.

If you do get issues, please post diagnostics. Most of the time the issue isn't with the card, but because many older consuimer boards like the M270 only supports graphics cards, and not any other cards, which means you'd have to use the x1 slots for those. Seeing the manual, that is probably the case here. If it does accept other cards in the big slots, the E6 one is only x4, depending on what card you have, it may not accept that either, although most sata-cards are only x1 or x2. Also, if you're using the M.2 slots for a sata-card, you can't use sata and M.2 at the same time for many combinations, see the manual.

This happens when there is no internet-connection at the time of booting. And it is related to the Connect plugin. I think it tries to download it from the internet instead of getting it from a stored source.

If you buy a new board, it will have been tested for operation. So when you get the board (you only need the plain board, no CPU, no ram etc), take a photo or write down the IPMI password (there will be a sticker saying BMC and PWD), get a powersupply and connect the 24-pin to it, put a network cable in the dedicated ipmi port to your switch/router, apply power, you should get a LED that eventually starts blinking, which is the heartbeat LED for the BMC. It usually takes about 30-45 seconds to start up. For first use, it's easiest to use IPMIview which you can get here. (There's also an android app but that depends on how your network is set up but you can try that too). IPMIview has a search-function you can use, and it should come up with the IP address of the board. Use a browser to connect to that, and enter ADMIN as username, and the password from the sticker. If there was no password listed, you have an older board from before November 2019 and the password should be ADMIN. If they changed it, but the board boots, you can reset it, but let's assume it's the default password. Reason for wanting to connect the IPMI, is because even without processor and memory and other devices, is because it just plain feels weird to talk to a board that doesn't have anything attached and works anyway, it's not something a lot of people do, especially not with a new board. So instant celebrity status. But the real reason is practical. The main-screen will have information about what firmware and BIOS is installed and such. So you can copy/past that information to a text-file in notepad or something. It'll also show the MAC-addresses of the network-ports which can be handy. Next on the left it'll have Hardware information, if you click that, it'll show information about the system, and a whole lot of things you can open with arrow-icons, with information about BIOS, CPU, DIMMs, and powersupply. Most useful is to copy information about the CPU and all the DIMMs and add them to the information-text-file. The reason to copy all this information, is because you know the board booted with those devices. So if you have issues, and are able to get those same devices, and it still doesn't work, it means you're doing something wrong, or the devices are broken. Or if the devices do work on another board, you're doing something wrong or the board is broken. Since the information will be erased once you boot successfully, it's important to copy that. At this point, if you purchased or got a 'supermicro product key' for SFT-OOB-LIC you can also apply it under Maintenance->Activate license, and after that you can update the IPMI firmware from Maintenance->Firmware update, and the BIOS from Maintenance->BIOS update. Although you may want to wait with that until you've booted at least once. I have two of them, so I updated one after booting it, but the other I updated before putting a CPU on, so I know it can be done, but if something goes wrong, you won't know if it is the BIOS or the devices you connected, so its probably better to wait. After that, remove power, put on the CPU using the procedure in the manual (Intel has a video here but do not do what they did, use hand-tools). Also, because pin 1 is in a different direction than usual with processors, it helps to put the board, cooler and cpu next to each other so you can visualize how it all connects to each other before you find out your cooler is the wrong way around and now you have no cooling paste to replace the ruined stuff (the carriage comes with the cooler, make sure to select the right one if there are multiple). Make sure to follow the 1-2-3-4, and because of the weight, it helps to keep a finger or two on the cooler to keep it from tipping over. You can do 1 a little, then 2-3-4 also a little, and then 1 fully etc, but the order is very important to prevent damage. For the BIOS, you may have to change the boot-order, once you connected the USB, but that's about all. There is an USB-connector on the front of the board, so that's the easiest location. Otherwise, everything should start up, however, before moving all the drives and everything, disable autostart for the array, dockers and VMs, once moved, start the array, and then check if you have to edit anything for the dockers and VM's, especially regarding networking, because your board will have different MAC-addresses. Other interesting things you might not expect, the RAM-slots all face the same way, for older processors, memory faced differently so you'd always install memory with the labels facing away from the processor, one side was reversed. So if you automatically reverse the memory installing the second stick out of habit, and can't get it to go in, you know why. Edit: Forgot to say, like with all new memory, run a memtest! unraid has memtest86+ which should work okay. I have memtest86 so I used that, and found I purchased a bad stick, but I didn't test that one with memtest86+ to see if it also would fail.

The 9305-24i uses SFF-6843 connectors also known as Mini SAS HD, your backplane uses Mini SAS which is SFF-8087. The SFF-6843 is newer, but it's basically the same connection, just in a different format, so you can get SFF-8643 to SFF-8087 cables, like you already selected. For SATA you do need forward or reverse cables, because SATA has cables crossed so receive connects to transmit on the other end, but for SAS it is straight through. That means if you have SAS at one end and SATA at the other, the data only flows in one direction, which is why you need the different cables depending on where it needs to go. The 6138 is great, will work fine, but I didn't like how it was a bit laggy with the unraid interface. I also only choose those because I could get them cheap. One of the things to keep in mind is the TDP which is how much heat the processor can generate. My board only supports 165W TDP processors, the X11SPi-TF supports 205W, so you could for example put a Platinum 8168 on there, or a slightly better than the 6138, the 6148 which is also really cheap. The only thing to keep in mind with processors, some have ES in the keywords or description, those are Engineering Samples. For some reason people really want those because they may have features that aren't in the released retail processors, but they won't run on most Supermicro boards without altering the BIOS. If it doesn't say, googling the S-Spec code usually tells you what kind it is, like SR3B5 is a normal model. The memory you selected is probably wrong, since it doesn't display any of the stuff that is important like manufacturer of the memory, or type. Supermicro really likes SK Hynix memory, or Micron but that might be hard to get (at least it is where I am). The 6138 supports speeds up to 2666Mhz, the 2nd generation (like 6248) supports up to 2933Mhz. The easiest way is to go to a boards page (like the X11SPi-TF page) and under Resources you'll find a Tested memory list. It'll list supermicro part-numbers, which aren't very useful, since they can show numbers for many part, however if you use the store-link or google those, it'll tell you exactly what type they are, for example MEM-DR432LC-ER26 is really Micron MTA36ASF4G72PZ-2G6E1, and if you google that, it'll show you places that charge about $30-$35 instead of $100 that Supermicro would if they weren't sold out. The other type they recommend is HMA84GR7AFR4N-VK which is what they tested one of my boards with, but if you find some memory, you can also go the other way around, for example, I'm using HMA84GR7CJR4N-VK, if I google Supermicro + the type it shows a link to the supermicro site where they sold this memory as MEM-DR432LC-ER26 (hey, it's the same part-number!), and there is a validity checker on the store-page for systems and motherboards which shows the memory is compatible with my (and your) boards, despite not being listed with the other method. So if you can get some memory cheap, this might be a better method.

The ASR-71605 is a RAID controller, not a HBA. This means if you configure the drives, they won't show up since the card is holding them hostage. They can be passed through, but you have to set it to either expose RAW, or HBA mode, and of course you have to make sure the latest firmware build 32118 is installed. I have an 8885 which has a newer build firmware, but the option should be somewhere on the second page of controller options. Currently I don't have the controller in a computer, but if you can't find it say so and I'll install it to give better instructions.

They only tested until 2016, see the compatibility document here. They tested and found compatible MG04 drives.

This doesn't help when the system is off, but there is a plug-in for that called Disk Location in the apps where you can set the layout and assign drives to it and it'll show up like that in the dashboard. So if something happens with a drive, it'll show the location there.

I'm a great fan of Supermicro motherboards, at one point the development system and recommended was an X10SL7-F with a Xeon e3 1230 v3. I got that and ran with it for years, even got an e3 1285 v4 for it, which has a GPU it can't use (chipset doesn't support it) but the cache makes it run fast. Then I got an X10DRH-CT which has 10gb on board, but then windows 11 came, and I got some really cheap Xeon gold 6138 processors, which I used on a X11SPL-F. These boards all have a BMC or Baseboard Management Controller, which means you can turn them on remotely, check logs, control them remotely, and it has onboard video so no videocard required which is very useful (no need to buy a monitor, keyboard or mouse) and it also has hardware information. Which told me the board was tested with 6248 processors, which I am currently running, with 192GB ram. That X11 also has 7 PCIe x8 slots (one runs at x4), so if you want a videocard for transcoding, you can just put it in, and I'm currently using it with 25gbe networking-cards. With 10gbit onboard, the equivalent with slightly less PCIe slots would be the X11SPi-TF (thats I not L). These are older boards, but there's great modern boards too. If I wanted to buy something new completely instead of getting used processors, I'd buy the X13SEI-F or X13SEI-TF for 10gbit. As for storage, the X10SL7-F was great because it has 8 SAS and 6 SATA connectors, so you'd only need a single HBA. I like reliability, so if something fails, I want to be able to replace that part, and keep going, so I'd run a 9305-24i, but with hard-drives, any combination that has enough slots will work. So for example a 9211-8i with a RES2SV240 expander, or a 9201-16i HBA with a 9211-8i (or 9207i or any other LSI/Avago/Broadcom card). It's more a question of power-usage and heat than what works or not. And cost of course. Speaking of cost, even though those are really cheap solutions, they cost about the same these days as more modern variations, like a 9300-8i with a AEC-82885T expander. That last one is really nice because it can be powered by PCIe or with a Molex plug so you can stick it anywhere and have it work. Oh, and for some cards there are cards with less connections, like the 9305-16i but then you see the sad empty places on the board, and those cards work fine but I hate seeing that, so I won't mention those.

Some do, if you use storcli it says which ones and the temperatures, how accurate they are I don't know though. In the terminal, assuming storcli is in the current directory, you would use the following to show everything: ./storcli64 /c0 show all And for just temperatures: ./storcli64 /c0 show all | grep temp Which for me shows a temperature of 79 celcius on a 9300-8i (really an IBM M1215) which is pretty high, so I'm not sure how accurate it is.

That's not good. It has port-multipliers, which means many problems, see here. Even if they work, they are extremely slow, they only connect with x1 or x2 at most. For what they cost, you could get a 9200-series or 9300-8i and a sas-expander, although it would use more electricity and generate more heat, it would run full-speed. As for memory, you need 4GB or more, so for a file-server 16GB is fine.

So 1gbit is about 125MB, and there is some overhead and such, so usually you'd reach about 120MB. Modern hard drives can easily reach that speed, let alone SSDs. So that means as soon as you start copying, the entire link can be consumed with that, and if you need it for something else like internet, there won't be space. So 10gbit is the way to go (although I would have gone with fiber, since those cards also get warm, it's less than copper, and uses only 1watt or less per port instead of up to 5watt.)

Could be, that's why it is interesting to see what shows up in the logs, and if the device gets detected properly, which is what would show in the diagnostics. I don't have a 9305 yet, so I don't know what temperature they reach, but generally sas-controllers do get hot and without proper airflow they can overheat and break.

It sounds like something has gone bad, try showing us diagnostics. If they are SATA drives, try connecting one or more to the motherboard to see if they show up that way to exclude bad disks, and unplug and put back the cables from the controller and the drives, or use a different cable to make sure the connection is working properly. A good cable will click when it is placed correctly, but some don't make a noise and it's difficult to tell when it's plugged correctly. Also make sure they are facing the right way, since you can put them on the wrong way.

For file-servers it doesn't really matter, although the unraid GUI responds faster with faster single-core performance usually. When you want VM's and dockers and transcoding and such, that's when you want more specific, but for file-servers almost anything goes. That will work fine, although the X99 chipset is not a server-chipset so if you want to go crazy and put a xeon e5 2680 or something on it later, it will work but may miss some features. If you're only using SATA/SAS harddrives and not SSDs, you can also consider a single 9305-24i, and if you need more ports, an expander like an Adaptec 82885t. Also the 9400-16i may be around the same price or available cheaper with or without expander.

Those x1 slots are only PCIe 2.0, which means a single lane gets about 500MB/s or about 4gbit, so at most you'd get a 2.5gb card which would work. But most 10gbit connections won't do speeds of 2.5 or 5 gbit so it depends on what it'll be connected to if it accepts that or not. So, I think it's time for an upgrade for your motherboard at least.

An adapter normally gets an 169 address when there is no connection, so there may be something wrong with the Eth0 cable, or the device on the other end (perhaps a bad port). Since you don't have bridging enabled, it uses that non-existing connection and so can't reach the internet, so you'll need to fix that.

According to the manual in section 2.4 about the expansion slots, only the x1 slots can be used, the other slots are only for graphic cards. From the pictures, it looks like the x1 slots are not open, so your TP-Link TX401 won't work, since that has a physical x4 connector and doesn't fit.

To remove a drive, see here (even if 6 is already removed, it doesn't change the steps) You have two pools, one called Cache and one called Pics (and the array). If you are running Unraid 7, you can use the mover things from one pool to another so the name doesn't matter. It's fine to have apps run anywhere too, but make sure to enable trim on the drive-settings and in the general settings if that hasn't been done.

It's pretty simple, so it doesn't really need a writeup, but for security: Have a backup of the data, make a screenshot of the unraid page showing the drives, just in case. Disable autostart on the array, dockers, and VM's. Shut the server down. Move the drives and USB, start it up. Go to network settings, make sure the right port is selected, since the new server will have different MAC numbers. If it looks good, start the array, if that works well, check the configuration of dockers or VMs and start them one at a time, if they work well, go to the next and so on. Finally, you can enable autostart again if it was enabled before.

From lspci, the card isn't detected at all, which isn't strange since Dell is proprietary hardware, so a non-Dell card may not work. The other way around is the same thing, you have to cover/disable pins to make a Dell card work on a non-Dell server. So in this case, the solution is buy one of the Dell HBA's.

You can also do it manually using the terminal or connecting through ssh or something, and then entering commands manually or using MC (midnight commander). For example, say you have a new 4TB drive, and 1 and 4 are the smallest it makes sense to replace those. So, replace disk 4, wait for the rebuild, then move the data from /mnt/disk1 to /mnt/disk4, stop the array, move the drives over except disk 1, do a new config, and assign the disks as they are now.