MetricTonto

Thank you: I've started the thread here: https://forums.unraid.net/topic/199814-loss-of-pool-devices-device-assignments-after-update-to-732-if-possibly-unrelated/

After updating from Unraid 7.3.1 to 7.3.2 and rebooting, my Pool Devices' device assignments vanished: the pools were there but empty, all devices showing up as Unassigned. The Array was unaffected. Everything else seems normal. The pools were: A dual BTRFS SSD cache pool. A four HDD RaidZ. A single laptop-grade HDD. Reverting to 7.3.1 didn't help (in a bit of a panic there, as I'm a novice). Checking things while guided by ChatGPT 😑 revealed that the /boot/config/pools/ config files had apparently been reset in some manner: no drive assignment info in them. Luckily everything else seemed intact and just assigning the devices back sufficed. The boot USB drive's current files reflect this (its pool config files show the drives now). I enclose both the diagnostic files generated inmediately after noticing the issue, and the pool config files from a backup previous to updating. nas-diagnostics-20260709_2227.zip Pool config files before incident.zip

The upgrade seemingly ate my Pool Devices' device assignments (a dual BTRFS SSD cache pool, a four HDD RaidZ, and a HDD for music streaming). Reverting to 7.3.1 didn't help. Luckily everything was intact and just assigning them back sufficed. Most scary for a non-poweruser (I had to recourse to ChatGPT, which is something I hate if I don't know enough to catch any glaringly bad advice). I keep backups of the boot USB drive and appdata. I'll try to see what could have happened: I don't have anything out of the ordinary, apps and plugins-wise, but maybe there was some weird interaction there.

You are right. I was grasping at straws. (Possibly unrelated, but who knows: the WebGUI, both locally and remotely, started to show enormous lag. I've rebooted and things seem normal now)

I see. Thanks for taking the time to investigate and highlight that. I guess I could produce some tests (get some TV episode file, duplicate it to a quantity enough to pass for a multi-season collection, try again, and monitor the process). (Previous balancing adventures used the Unbalanced tool. I'll limit myself to it fornow)

Thank you for your assistance. The destination folder was created before the move operation (the origin one no longer exists, so, I've recreated it). The ones in the destination hardly hold one or two files instead of the complete seasons they ought to. (I don't use the cache pool for staging, so, typically file transfers go directly to the Array) nas-diagnostics-20260627-2243.zip

To manually balance my array's drives, I used Dynamix File Explorer's move tool to group some TV series folders, in a share that straddles two disks, into the least filled one. The UI let me choose the destination folder (a newly created one), and once launched I could see the progress in the File Explorer and the drives' activity in the main window, its speed being as expected. But when checking the result I saw that it had hardly created some subdirectories and moved a few files. 95% of the material is just gone. The only thing I can think that could be an issue is that I had Krusader open in another webbrowser tab (I tipically have it so). Could that be a factor? I'm on Unraid 7.3.1. (Is there any procedure to have Unraid check the Array's drives and shares for inconsistencies that might let me try to rescue that material? It's not that crucial and could have it around in some old HDDs, but…)

This motherboard's BIOS at the very least allows for setting it to require a four seconds button press for it to act.

OK, so, it was the no-name PC keyboard's power down button 😑 (I never use such, and I used to have a Mac keyboard there instead). I wonder if I can disable that in the BIOS or somewhere else.

That's worrisome 🤨. The PC case is set in such a way that the button is difficult to reach. The cradle has one of those buttons meant to trigger a Windows HDD cloner app, though. I wonder if I pressed that one without meaning to and it produced a similar reaction. I'll have to test that. In any case, thank you for giving it a look 🙂.

I was checking the contents of a SSDs via my USB3 cradle (it never was trouble), working directly with the server's display and keyboard. Once appearing as an Unassigned drive, I attempted to mount it. Instantly, the GUI was replaced by the console and a flurry of lines, and the system shut down in maybe ten seconds. Fearing the worst, I booted and logged in, but I saw no warnings about any need to rebuild the parity data anywhere. Not restarting the Array just yet, I tried mounting the SSD's volume again, and this time it worked normally. After perusing its contents, I unmounted it and switched the cradle off. Then I started the Array. Again, no warnings of anything wrong with the parity data. The only notification was one of the SSD having showed an inordinate amount of CRC errors (although I seem to remember those not being real but some kind of artifact). I'm checking syslog-previous. As I'm a complete novice, I don't understand it, but I can see the references to the "SuperSpeed USB device" (the SSD), and what feels like an orderly shutdown process. I would be grateful if somebody could give it a look and opine on it. syslog-previous.txt

In the end I went that route: ZFS Master set to manual refresh, and having Dynamix Auto Fan Control to exclude all the concerned drives from reading their temperatures, while setting the minimum PWM values to ones that can keep them decent while they are active (max rpm, these days 🫠). I need to do that because this old mainboard's BIOS is too limited and won't let me set a constant value: it varies it in real time based on a mix of CPU and chipset temperature reads.

Thanks! Safe mode worked 🙂. I'll have to uninstall the most suspect plugins and reinstall and try them one by one. My guess is the fan control one is the culprit, as it has to interrogate the drives for their temperatures. If the Dashboard can't show them already while spun down, that suggests getting the info will reactivate them. The problem is that I can't rely on the BIOS's fan curves (too old a board, "what's a fan curve?" 😒) to keep the HDDs decent. We'll see.

I've tried setting it so that refreshing its data requires manual action, to no avail. I'll try uninstalling it. It might not be the culprit, though: there are also the Dynamix System Temperature and Dynamix Auto Fan Control plugins: I'm happening to need them, as my motherboard's BIOS doesn't give me any means to regulate fan speeds based on HDD temperatures, and we are in the middle of a heatwave. They might be the ones waking up the drives, every time they check their temps. I guess I could manipulate their settings so that they exclude all the concerned drives while having a floor value high enough to keep them cool while spun up.

As I'm having the same problems many do with trying to keep ZFS-formatted drives spun down (a RAIDZ in my case), I was wondering if a base Unraid installation with no additions (plugins, dockers, etc.) has been verified to be able to achieve that, at all. And, if it has, what are the required settings. (If the Dashboard, alone, is prone to waking up the drives to do its reporting, it would need some kind of adjustment to limit that)

In theory, while doing an incremental file backup it replaces any unchanged folder holding more than ten items inside with a symlink to the original in previous backups. See: https://www.urbackup.org/administration_manual.html#x1-750008.6 I'm having a severe problem with it and my Windows PC's backups, though. UrBackup's GUI shows them and their contents, and I can restore them or their items with no issues. BUT: Unraid's built-in file browser doesn't show them inside the share where the backup is stored (a RAID-Z one). The share, mounted on in my PC via SMB doesn't show them either. Kommander shows the symlinks but says they are broken. Via Bash, I can ls in that directory and see the symlink: "EIAS@". Doing ls -l EIAS gives me: lrwxrwxrwx 1 nobody users 64 Mar 17 23:55 EIAS -> /media/MY-PC-NAME/.directory_pool/DM/DMthUhIsJj17422521537920821 which seems valid. I've tried a few SMB recipes to make symlinks accessible, unsuccessfully. That Unraid's own file browser doesn't see them feels even more concerning. Does this happen to anybody else? Should we disable this default setting when dealing with backups of Windows data (doing so gives me the normal, expected behavior)? Or is it a ZFS issue? I'm on Unraid 7.0.1.

So, it seems it's actually normal 🤔: (Also, I'm realizing now that I didn't set the disconnected motherboard NIC to a different subnet before trying to make the Mellanox the primary one 😒. I'm not sure that would have solved the internet connection, but it certainly ought to have helped) So, case closed, I think. Thank you all 🙂.

You are right: I somehow got DNS tunnel vision and missed the bigger picture. As for your other suggestions: I had disabled the switch's VLAN already, to no avail. I have distinct static IPs for each device, and also I have their MACs bound to those IPs in the router's DHCP so that they stay stable when testing them with Unraid's default dynamic IP settings. The router sees all of them correctly. 192.168.1.1 ISP's router. 192.168.1.2 switch. 192.168.1.20 Windows PC 192.168.1.254 NAS (When both the NAS' 1G and 10G NICs were active, the one acting as eth1 would use 253) My guess is everything is right with the inexpensive Chinese managed switch, as the PC has had no trouble at all, and the NAS, when bonded, reaches the internet without issues. The thing to note is that, both when the 10G NIC is bonded and un-bonded, the switch and the router see the same MAC and static IP address. So, I don't know what could be going on. I'm trying to find documentation about the network settings' behavior as a whole, but my Google-fu is lacking. I'll keep on trying things and reporting the results. If not getting anywhere, I'll post screenies and config files. Thank you for your assistance 🙂.

I have the usual Unraid NAS setup of a motherboard's built-in 1G Ethernet port and a 10G SFP+ NIC. For a while I used Ethernet to connect to my ISP-provider's internet router/switch while having a point-to-point 10G link to my PC's own SFP+ card. A little ago I got a cheap Chinese switch with a pair of SFP+ ports and four Ethernet ones, one of those connected to the ISP router/switch in turn, so that the NAS and PC can both talk to each other at SFP+ 10G speeds and access internet and my WIFI at 1G ones. The NAS's Ethernet port is now empty. My issue is: THIS WORKS: If I let Unraid reset to its default networking settings (by deleting the relevant files and rebooting), it bonds the 10G NIC to the unplugged 1G one and everything works, static DNS included. Having the unplugged one be the primary sounds a bit silly, though, and also leads to this question: as the 10G NIC's settings get hidden while in its bonded state: should I de-bond it to change its MTU settings and re-bond? Or should I set them in the eth0's "Desired MTU" field? What happens at the MAC level? Etc. THIS DOESN'T WORK: if I do what seems tidy, which is set Interface Rules so that the 10G NIC becomes eth0 and the unplugged 1G one is eth1, no bonding, for some reason the 10G NIC can't access manually entered static DNS addresses (such as 1.1.1.1 or 8.8.8.8: pings fail). It can access everything in the LAN and transfer data, though. My PC has no trouble accessing DNS servers, so, I'm assuming it's not a Chinese switch issue. (I've seen suggestions such as disabling the Ethernet port at the BIOS level, but one might want to keep it alive for diagnostics and stuff) Given all this, my question is, then: what is the preferred way to configure such a two NICs scenario? And, if it's the one Unraid defaults to, how should one proceed in order to set the bonded NIC card's specific settings? In short: is this normal?😅😓

Sadly, I can't do that in an effective manner: the old motherboard's BIOS only allows for modes that either regulate fan speeds depending on temperature reads (which wouldn't deal correctly with my diverse storage units' types and situations: very different thermal ranges and airflow situations) or just let the fans go 100% uncontrolled. No fixed percentage settings available. I think I've managed to work it out, though. The solution for setting a constant RPM seems to be: Set the Maximum RPM Value to the desired speed. Set the min/max storage temperature range to something that will never be reached, say, 80º to 90º C.

Can System AutoFan be used to achieve plain constant fan speeds by just setting each PWM Controller's Minimum PWM value field and emptying all the rest of the fields? Or could there still be some interactions with the drives' temperatures? (As I only have two PWM headers being shared by five case fans, the drives and cards having very diverse temp. ranges and cooling needs that don't lend themselves to depend on drives' temps. only, and the case is in a storage room where noise isn't too much of a concern, I rather set fixed speeds that work for all situations)