FreakyUnraid

Well, it doesn't. Atleast not when flashing a backup. That is a bug, right? Zpool import and zpool status did not return anything. Forgot the exact phrasing, but zpool not found or something like that. And I always forget diagnostics... so I don't have that to share. I am going to wait a week or so before trying to update again.

Think I found a bug? The USB Creator tool gives the name "BOOT" to the flash drive. After manually changing this to "UNRAID" everything is working again with 7.2.6. Waiting up to 30 sec for device with label UNRAID to come online Checking /dev/sda1 ... Fsck.fat 4.2 (2021-01-31) Volume label 'UNRAID' stored in root directory and label 'BOOT' stored in boot sector and different. Auto-copying volume label from root directory to boot sector. After that booting continues and without issues.

I used the 7.2.6 backup to roll back. Flashed the flash drive with the USB creator tool. First boot: Error: "Waiting up to 30 sec for device with label UNRAID to come online" Flash drive was / seems fine. Windows sees the drive without issue. Flashing drive a for the second time...

Something went wrong upgrading to 7.3.4. Upgrade went fine? Rebooted the server. Was welcomed with the new onboarding wizard, which I skipped. Went to 'Main' tab to start the array; all array disks are present, but all ZFS pools show "none". Every missing disk is present under unassigned devices... What should I do?

Thanks for the quick reply. That's awesome. Okay, sorry for asking. I Also searched the forum here...

I am planning to move to internal boot. I have bought 2x Intel Optane M10 16GB drives for this purpose. A mirrored internal boot pool. I am watching The Uncast Show's video on this topic before upgrading and this raises some questions https://youtu.be/sruWM1PqjVo?t=495 SpaceInvader One is using a 2TB nvme drive and his screen is showing this: Each bootable pool contains two volumes a system boot pool volume used by unraid a storage pool for general data ... Boot reserved size minimum is 4GB, maximum is 931GB (50% of the smallest selected drive) Q1: What's the point of that storage pool for general data? Does that mean, when using Optane 16GB drives, I will get an 8GB storage pool under Pool Devices? Q2: Why is 50% of the smallest drive the maximum amount the boot partition can be? Why can't we use the entire drive size for this? Just like with USB drives, right? Maybe I am missing something here...

Well, I am absolutely not an expert but also not a complete noob so I'll stop arguing about the logic behind this. Fact is, it causes issues. I believe Unraid has to do a better job of explaining these settings to users. I followed Spaceinvader One's video's when setting up my server. This is the setting for TV Shows usershare. My current settings just seem to break Mover. Mover doesn't even move on to another drive. It keeps trying to move the same (.partial) file to this drive for a couple of hours now... And the log gets filled with "(28) No space left on device". Mover seems to not be able to handle this situation? This doesn't seem user friendly at all... I have to find this out by going in to the logs and see what's going on. No notification. I then have to manually, in console (also not user friendly), stop mover and now go use a third party plugin to empty out my drive before I can continue Mover. I really would like to hear what the settings for a TV Shows usershare should be, because to my recollection all youtube video's recommend the same split level settings? I kind of understand the problem you describe, but I can also give some counter arguments to this with my noob level of knowledge, but I'll just shut up now. Well that was basically my thinking with doing this. Most files are max ~25GB big. So with setting it to 50GB I would have enough space left and if a file goes over this I would move things around manually to balance things out a bit more. Currently this "free space" setting only seems to be for my cache drive? If the cache drive hits that limit, then it will start to directly write to the array... Like why is this setting also applied to the cache? Fill them up 100%? So no data can be changed at all in the future, like metadata? Performance of the drive, or at least for some data, down the drain? Don't think that is smart. And what about the new options for SSD pools? Isn't the advice to only fill SSD's to max 80% capacity? Like I said, I followed Spaceinvader One's guides. I have this set for the TV shows share and I have seen several other tutorials that recommend the same setting for such a share. Also, the explanation in the GUI doesn't mention anything like this at all. It just explains how it will manage the creation of certain directories, but not that you should keep an eye on the drives and manually move things around when using these settings... And again, this really seems like weird design. Mover can't even handle this split level situation? Don't worry, this is just my single ssd download cache. Appdata has it's own dedicated zfs pool. Because the mover tuner plugin went bad I have since then used the logic of just filling the download cache to ~90% and then manually hit mover. I have set notifications when the cache gets filled beyond 85% and then I go in and run mover over night manually before going to bed.

I understand split level somewhat , but I don't understand why this gets priority over the set amount of free space. If no free space is set and the drive is full then mover will move to another drive, correct? If I have a 2TB drive and want 100GB free space as a minimum, then mover should just see this as a 1.9TB drive and move on when moving a file goes over that threshold. Just like it would when a drive is full. Ignoring the set free space seems so illogical to me? I now end up with drives that only have 2GB of free space left... How does everyone else handle this? Is this some old logic left from the early days? I can't imagine you would want this bahaviour when you don't use normal HDD's, but only SSD's? With the current addition of pools etc. I mean, those don't really like it when you go over 80% or something, right? But I'm getting distracted here. Question I have now: How do I configure Unraid to get it to respect the set amount of free space? Is this even possible or do I have to use Unbalance every time manually? This is mostly *arr media files, so folders are not created by me manually.

@itimpi I have always used this setting: Automatically split only the top two directory levels as required Why does Split level not respect Minimum Free Space setting? I don't understand why this just ignores "keep this amount of space free". If the disk is full it goes on to the next disk. So why not "oh keep 50GB free, then disk is full now, let's move on". What's the logic here?

No, I stopped using that plugin after upgrading to Unraid 7.0.0. Mover was completely broken at that point. I tried it once again like a month later, but still had problems. I'm currently using mover manually when I see my cache drive is full.

For all shares I have set the following: Minimum free space = 50GB Allocation Method = High-Water Problem: Mover fills array disks to 100%. Drives are filled with only 2GB left... The files that are moved are way less than 50GB, so I don't understand why mover keeps writing when a drive hits that 50GB threshold. I don't want mover to fill my disks to 100%. I want to leave some free space. And I choose 50GB because Minimum Free Space also applies to the cache drive... I want to be able to fill my 1TB cache drive

Okay, perhaps my understanding of writing corrections to parity is not 100% correct. This reinforces my argument for adding more information / context for the user. But even so, I don't see a reason to immediately write corrections to parity without giving the user the ability to first check what was found by parity check.

Writing corrections should be OFF by default. Check: A process of examining or verifying something to ensure it is correct, consistent, or functioning as expected. The system says: Parity is valid. The system does a check if that assumption is still valid. A parity check stresses your hardware like no other operation. Cables, HBA/Sata controllers etc. If those somehow fail during the check, the "write corrections to parity" will ruin your parity. There should always be a check first. A second run should be used to write correction AFTER user intervention. This gives the user the ability to check the logs, see what happened and if there could be a hardware explanation. If so, fix that and do another check before you alter / ruin your parity. "You can always do another correcting parity check if the first one was because of hardware failure"; Yes, but you assume no HDD will die during that second process. If that happens, you can say bye bye to your data. In all settings there is an explanation of what a certain settings does. On the Main Tab there is no explanation under "Check" or "Parity Valid", just that it will start a Parity-Check. It also does not explain what "write corrections to parity" precisely mean and what the possible dangers of this setting is. My suggestion would be to add more information here. And possible move the question to write corrections to parity to a pop up window that will specifically ask you (and explain) if you want to write corrections to parity. If it is the first parity check since / parity is valid it should suggest the user to just do a check, without writing corrections.

I would like an option to disable the mover schedule. A user shouldn't be forced to pick a schedule. Choose a mover schedule ranging from hourly, daily, weekly and monthly. The interval determines how fast the mover will activated, it runs in the background. Setting a fixed schedule does not work for me. My downloads and work are to irregular for a fixed schedule. You might say 'just use the mover tuner plugin'. Right, I don't think this basic feature should be a plugin. Second of all I am running Unraid 7.1.2 and unfortunately the mover tuner plugin has been unreliable since 7.0.0 and I don't want to deal with that anymore. And although Mover "runs in the background" it will take up a lot of my cpu power and it will take a long time when my 1TB cache is almost full. So I don't want to have it run when I'm downloading / seeding or doing other stuff. I have set up notifications that will notify me when the cache is 80% full or higher. That way I know when I need to run mover; "cool, I'll stop my work tonight, stop downloads and the *arr apps (because of their "scan library task") tonight and let mover do it's thing". Somewhere in the morning it's done and I can resume everything again. That way I am in control. Now I had set it to "monthly" and this morning I was trying to do things and just losing my mind why my server was so slow. 1st of the month... mover is still running...

Yes, well, there is no warning before running mover so I didn't know that before. If I had known it would take mover 1 month to complete... I just went with what I had learned previously with changing cache drives. Move appdata to array. Change drive and put everything back. Before trying different things I would like to know how to stop the moving process. mover stop mover: not running While I see in the Main tab read/write on both the array and btrfs pool at the same time and at the same speed, telling me that mover is still doing it's thing. Only thing I can think off is that it's finishing it's 'cache' of files it had already lined up for processing? Anyhow, it's been 3 full hours and it's still going and I have no idea how to monitor what's going on or how to safely stop it, aside from stopping the array or rebooting the server. EDIT (before posting) and for anyone getting into this same situation; I had a chat with chatgpt and checked everything afterwards with Gemini to make sure Told it what I did and what was happening right now and if we could try the least invasive options first to fix this. It wanted me to look which processes were busy with files on my btrfs pool. lsof | grep "/mnt/cache_raid" This returned a lot of the same UID 122 messages, but also a couple of 'find' messages lsof: no pwd entry for UID 122 find 3707572 root 6r DIR 0,44 206 856278 /mnt/cache_raid/nextcloud/appdata_ocwgts04pj90/preview/8/8/a/c/7/2/8/371816 chatgpt suggested next to kill the process with "kill -9 3707572", but after googling "-9" (terminate process immediately) and being afraid of data getting corrupted, I opted for for "kill 3707572". After running that it took around 15 seconds for the Main tab showing me that read/write stopped completely. Looking at syslog straight after that it showed this: Jan 21 15:17:48 Server root: ionice -c 2 -n 0 nice -n 0 /usr/local/sbin/mover.old stop Could this perhaps have something to do with the mover tuning plugin and Unraid 7? I'm seeing mover and mover.old script. After the upgrade to Unraid 7 mover wouldn't move. I googled and removed the plugin after reading that I had still an old version installed. I thought Unraid would warn users for incompatible plugins before upgrading? Anyway, after removing that plugin and rebooting the server Mover was working again. But I wonder why mover.old is now the cause of this all... Or maybe was I wrong for killing the process first and should I have just ran "mover stop"... The mover part has been fixed, so now I need an option to move all the small files to another drive. @JorgeB Do you mean with Midnight Commander I should use Krusader? Just move files from one window to another, into the right disk/folder? Chatgpt suggested rsync rsync -av --ignore-existing /mnt/cache_raid/<sharename> /mnt/user/<sharename> I was using a specific disk in my array, so maybe change out /mnt/user for /mnt/disk ? And perhaps I should just use an unassigned drive for this, so the array won't slow things down doing parity calculations? Given the fact there are so many small files. I will be deleting that preview folder for sure and just redo that when it's all back up. Sorry for the long post.