AgentXXL

One last comment: I rebooted my server prior to destroying the pool/dataset just so it was 'clean'. Alas I couldn't destroy the dataset as the path wasn't found. So then I tried this: [email protected]:~# umount -vR /mnt/animzfs umount: /mnt/animzfs (animzfs) unmounted And here's the kicker: issuing that umount command somehow remounts the pool and dataset and the full path and files/folders are now accessible. WTF?!? In any case, I'm not going to trust it as it stands, so now I'll delete the pool from unRAID and then clear the partitions on all disks with them under UD. Then I'll recreate the pool and dataset, and then start the lengthy restore process.

Changed Status to Closed

I took another stab at it and confirmed that renaming the dataset didn't fix it. I've even tried making a new dataset and moving the data to it, but with the way zfs works at the block level, a move results in a copy and then delete. Not as quick as moving folder to folder on a non-zfs filesystem. I'm now at the stage where I no longer have enough space to backup that pool, so it's time to start from scratch. I'll be destroying the pool and then creating a new one. Then I'll create a dataset called Media and start copying everything back to the pool from the backup disks. Thanks for the assistance regardless. I never ran into issues with datasets back in my FreeNAS days but sh*t happens. I'll mark the issue as 'CLOSED', with the resolution being to destroy, recreate and restore from backups.

OK, this is interesting. I tried using umount manually. Here's the result: [email protected]:/# umount -R /mnt/animzfs [email protected]:/# cd /mnt [email protected]:/mnt# ls addons/ animzfs/ disks/ remotes/ rootshare/ [email protected]:/mnt# cd animzfs [email protected]:/mnt/animzfs# ls Media/ [email protected]:/mnt/animzfs# cd Media [email protected]:/mnt/animzfs/Media# ls TV/ So after manually issuing the command, I got no error messages, but it didn't unmount - instead it now shows the TV folder. And the contents are fully accessible. I tried using the umount command again, and it now says this: [email protected]:/# umount -vR /mnt/animzfs/Media umount: /mnt/animzfs/Media (animzfs/Media) unmounted And now when I check, /mnt/animzfs/Media still exists, but the TV folder is again missing. Something's certainly fubar'ed with this zfs pool. I was thinking I should try removing any mountpoint with this: zfs set mountpoint=none animzfs Will that just remove the mountpoint without destroying the dataset? From what I've just read on the man page, it should just remove the mountpoint. I do have all that data backup up, but I'm still hoping I can get the pool to properly mount and not have to restore ~35TB of data.

I followed the SpaceinvaderOne video Setting up a Native ZFS Pool on Unraid. I added the commands I used to my unRAID journal, but since then the pool has been moved to /mnt/disks/animzfs and now since trying the upgrade to 6.12, it's at /mnt/animzfs. zpool create -m /mnt/zfs AnimZFS01 raidz1 ata-ST10000DM0004-2GR11L_ZJV037AG ata-ST10000DM0004-2GR11L_ZJV610GT ata-ST10000DM0004-2GR11L_ZJV67B9L ata-ST10000DM0004-2GR11L_ZJV6BW2S ata-ST10000DM0004-2GR11L_ZJV6BYCF ata-ST10000DM0004-2GR11L_ZJV6BZRG zfs create AnimZFS01/Media It won't let me rename the dataset: [email protected]:/# zfs rename animzfs/Media animzfs/MyMedia cannot unmount '/mnt/animzfs/Media': unmount failed animzfs.txt

Updated with the results in the bug report thread. And as expected, I got the same result. I would still like to compare my pool structure with someone who was successful so I'd still appreciate a look at the results for zfs get all.

As expected, the same result: [email protected]:~# zfs list no datasets available [email protected]:~# zpool status no pools available [email protected]:~# mkdir -p /mnt/animzfs [email protected]:~# zpool import -N -o autoexpand=on -d /dev/sdy1 -d /dev/sdw1 -d /dev/sdz1 -d /dev/sdu1 -d /dev/sdx1 -d /dev/sdv1 2464160279060078275 animzfs [email protected]:~# zfs set mountpoint=/mnt/animzfs animzfs [email protected]:~# zfs mount -o noatime animzfs [email protected]:~# zfs mount animzfs/Media cannot mount 'animzfs/Media': filesystem already mounted [email protected]:~# cd /mnt/animzfs/Media [email protected]:/mnt/animzfs/Media# ls -l total 0 As mentioned in RC2 release thread, I'd really like to see the output of zfs get all from someone who was able to successfully import their pool that was created using the old plugin. One more thing to bring up: I already have a share called `Media` that uses both the array and a cache pool. Since I'm using `Media` as the dataset name, is it possible that the pre-existing sharename is causing the conflict? The reason I used the same name is that I want media residing on the ZFS pool to be grouped along with the media on the array/cache pool. So for example, the array and cache pool have this structure: /mnt/user/Media/Movies /mnt/user/Media/TV And the ZFS pool is this: /mnt/animzfs/Media/TV Any other thoughts?

I haven't tried again because it's pretty apparent it will give me the same result. Plus I've had other more important issues to work on, like replacing the air conditioner for the server room. As both of us expect that manually mounting with the same commands that unRAID uses will fail, I'd like to find out what's set differently so that a single mount command works. That's why I want to compare the output of `zfs get all` with someone who's been able to successfully import a pool created with the old plugin. I tried importing it under Ubuntu a few months ago and it had no problems with seeing the data. Why does unRAID do two mount commands? When I mounted it under Ubuntu it was a single command, and when I tried the same thing under unRAID on 6.12 with a single mount command, it would also see the TV folder and all sub-folder and files. Anyhow, I'm now ready to try it again under 6.12 RC2, but I'd first like to do that comparison to see if my pool is organized differently.

I have a ZFS pool (6 x 10TB RAIDZ1) that was created on 6.10.x using the old plugin. It has one dataset. Pool name animzfs, dataset Media. On 6.11.5, the pool mounts fine and the data is all accessible. I've tried upgrading to 6.12 RC2 three times now and my pool won't import properly. It creates the mountpoint and seems to mount the dataset, but no files/folders are visible when I browse to it at /mnt/animzfs/Media. There is an error during the mount that reports the pool/dataset is already mounted. What's odd is that the drive usage appears to be OK, but I can't see any of the folders or files. Permissions are listed as 99:100 for all folders/files when checked under 6.11.5. On 6.11.5 it's seen properly and the data is all visible - just a single 'TV' folder in the root of the dataset, with lots of subfolders and files. Permissions are listed as 99:100 for all folders/files when checked under 6.11.5. I've created a bug report and @JorgeB has thankfully been assisting me. The URL is here: I do have all the data on the pool backed up, so worst case is I do the upgrade to 6.12 RC2 and then destroy the old pool, re-create it and then restore my data from my backups. Alas that's about 35TB of data so it'll take a couple of days. I'd like to try and figure it out rather than destroy and re-create the pool, but I'm not making much headway. If someone using 6.12 RC2 has successfully imported a pool from the old plugin, could you please share the results of the command: zfs get all I'm trying to figure out if there's a way to re-organize the pool under 6.11.5 so that it will import properly under 6.12 RC2 or later. If anyone can supply me with the output of that command, hopefully it may reveal the issue that's preventing me from being able to see the data. Thanks in advance!

@JorgeB I just remembered that I had created a user script to set zfs_arc_max when using the plugin under 6.10.x/6.11.x. I deleted the user script and tried another reboot. There are no datasets or pools mounted before I start the array. Once I start the array, the animzfs pool is mounted, but the Media dataset is empty (even though drive space usage appears correct). Here's the relevant section from the syslog: Apr 7 15:10:32 AnimNAS emhttpd: mounting /mnt/animzfs Apr 7 15:10:32 AnimNAS emhttpd: shcmd (282): mkdir -p /mnt/animzfs Apr 7 15:10:32 AnimNAS emhttpd: shcmd (283): modprobe zfs Apr 7 15:10:32 AnimNAS emhttpd: shcmd (284): /usr/sbin/zpool import -N -o autoexpand=on -d /dev/sdu1 -d /dev/sdv1 -d /dev/sdw1 -d /dev/sdx1 -d /dev/sdy1 -d /dev/sdz1 2464160279060078275 animzfs Apr 7 15:10:42 AnimNAS emhttpd: shcmd (285): /usr/sbin/zfs set mountpoint=/mnt/animzfs animzfs Apr 7 15:10:43 AnimNAS emhttpd: shcmd (286): /usr/sbin/zfs mount -o noatime animzfs Apr 7 15:10:43 AnimNAS emhttpd: shcmd (287): /usr/sbin/zpool set autotrim=off animzfs Apr 7 15:10:43 AnimNAS emhttpd: shcmd (288): /usr/sbin/zfs set compression=off animzfs Apr 7 15:10:43 AnimNAS emhttpd: /mnt/animzfs root profile: raidz1 Apr 7 15:10:43 AnimNAS emhttpd: /mnt/animzfs root groups: 1 Apr 7 15:10:43 AnimNAS emhttpd: /mnt/animzfs root width: 6 Apr 7 15:10:43 AnimNAS emhttpd: /mnt/animzfs root ok: 6 Apr 7 15:10:43 AnimNAS emhttpd: /mnt/animzfs root new: 0 Apr 7 15:10:43 AnimNAS emhttpd: /mnt/animzfs root wrong:0 Apr 7 15:10:43 AnimNAS emhttpd: /mnt/animzfs root missing: 0 Apr 7 15:10:43 AnimNAS emhttpd: /mnt/animzfs root missing already: 0 Apr 7 15:10:43 AnimNAS emhttpd: shcmd (289): /usr/sbin/zfs mount animzfs/Media Apr 7 15:10:43 AnimNAS root: cannot mount 'animzfs/Media': filesystem already mounted Apr 7 15:10:43 AnimNAS emhttpd: shcmd (289): exit status: 1 When I got it to work manually, I only did 3 steps: mkdir, zpool import and mount animzfs/Media. The syslog shows ``zfs set mountpoint...`` and ``zfs mount -o noatime animzfs``, followed by the trim and compression settings as shown above. It looks like it's this 1st mount command (6th line down in above paste) which is doing a ``zfs mount -o noatime animzfs``. This is obviously not the same as the ``zfs mount animzfs/Media`` command that worked when tried manually. So it's something in the way unRAID is parsing the mountpoint. The 2nd ``zfs mount...`` command (3rd line from bottom) is the command that worked when tried manually, but fails now because of the previous mount command (6th line from top). Thoughts or suggestions? Thanks again!

Ok, gave it another go and this seems to work: [email protected]:~# zfs list no datasets available [email protected]:~# cd /mnt [email protected]:/mnt# ls addons/ disks/ remotes/ rootshare/ [email protected]:/mnt# mkdir -p /mnt/animzfs [email protected]:/mnt# zpool import -N -o autoexpand=on -d /dev/sdy1 -d /dev/sdw1 -d /dev/sdz1 -d /dev/sdu1 -d /dev/sdx1 -d /dev/sdv1 2464160279060078275 animzfs [email protected]:/mnt# zfs list NAME USED AVAIL REFER MOUNTPOINT animzfs 33.8T 9.67T 204K /mnt/animzfs animzfs/Media 33.8T 9.67T 33.8T /mnt/animzfs/Media [email protected]:/mnt# zfs mount animzfs/Media [email protected]:/mnt# ls addons/ animzfs/ disks/ remotes/ rootshare/ [email protected]:/mnt# cd animzfs [email protected]:/mnt/animzfs# ls Media/ [email protected]:/mnt/animzfs# cd Media [email protected]:/mnt/animzfs/Media# ls TV/ The TV folder is now visible, and I can play media from it. Alas I then start the array and all 6 disks in the ZFS pool now show as unmountable and the pool contents are no longer accessible. After a reboot the pool is still created, but no messages - just filesystem set to auto. zfs list reports no datasets. When I start the unRAID array, the pool looks like it mounts OK, but again I now have an empty folder under animzfs/Media. Checking through syslog, I get this: Apr 7 14:02:31 AnimNAS emhttpd: mounting /mnt/animzfs Apr 7 14:02:31 AnimNAS emhttpd: shcmd (174): mkdir -p /mnt/animzfs Apr 7 14:02:31 AnimNAS emhttpd: /sbin/btrfs filesystem show /dev/sdu1 2>&1 Apr 7 14:02:31 AnimNAS emhttpd: ERROR: no btrfs on /dev/sdu1 Apr 7 14:02:31 AnimNAS emhttpd: shcmd (175): modprobe zfs Apr 7 14:02:31 AnimNAS emhttpd: /usr/sbin/zpool import -d /dev/sdu1 2>&1 Apr 7 14:02:32 AnimNAS emhttpd: pool: animzfs Apr 7 14:02:32 AnimNAS emhttpd: id: 2464160279060078275 Apr 7 14:02:32 AnimNAS emhttpd: shcmd (176): /usr/sbin/zpool import -N -o autoexpand=on -d /dev/sdu1 -d /dev/sdv1 -d /dev/sdw1 -d /dev/sdx1 -d /dev/sdy1 -d /dev/sdz1 2464160279060078275 animzfs Apr 7 14:02:43 AnimNAS emhttpd: /usr/sbin/zpool status -LP animzfs 2>&1 Apr 7 14:02:43 AnimNAS emhttpd: pool: animzfs Apr 7 14:02:43 AnimNAS emhttpd: state: ONLINE Apr 7 14:02:43 AnimNAS emhttpd: scan: scrub repaired 0B in 1 days 13:34:19 with 0 errors on Wed Apr 5 09:29:43 2023 Apr 7 14:02:43 AnimNAS emhttpd: config: Apr 7 14:02:43 AnimNAS emhttpd: NAME STATE READ WRITE CKSUM Apr 7 14:02:43 AnimNAS emhttpd: animzfs ONLINE 0 0 0 Apr 7 14:02:43 AnimNAS emhttpd: raidz1-0 ONLINE 0 0 0 Apr 7 14:02:43 AnimNAS emhttpd: /dev/sdu1 ONLINE 0 0 0 Apr 7 14:02:43 AnimNAS emhttpd: /dev/sdv1 ONLINE 0 0 0 Apr 7 14:02:43 AnimNAS emhttpd: /dev/sdw1 ONLINE 0 0 0 Apr 7 14:02:43 AnimNAS emhttpd: /dev/sdx1 ONLINE 0 0 0 Apr 7 14:02:43 AnimNAS emhttpd: /dev/sdy1 ONLINE 0 0 0 Apr 7 14:02:43 AnimNAS emhttpd: /dev/sdz1 ONLINE 0 0 0 Apr 7 14:02:43 AnimNAS emhttpd: errors: No known data errors Apr 7 14:02:43 AnimNAS emhttpd: shcmd (177): /usr/sbin/zfs set mountpoint=/mnt/animzfs animzfs Apr 7 14:02:44 AnimNAS emhttpd: shcmd (178): /usr/sbin/zfs mount -o noatime animzfs Apr 7 14:02:44 AnimNAS emhttpd: shcmd (179): /usr/sbin/zpool set autotrim=off animzfs Apr 7 14:02:44 AnimNAS emhttpd: shcmd (180): /usr/sbin/zfs set compression=off animzfs Apr 7 14:02:44 AnimNAS emhttpd: /mnt/animzfs root profile: raidz1 Apr 7 14:02:44 AnimNAS emhttpd: /mnt/animzfs root groups: 1 Apr 7 14:02:44 AnimNAS emhttpd: /mnt/animzfs root width: 6 Apr 7 14:02:44 AnimNAS emhttpd: /mnt/animzfs root ok: 6 Apr 7 14:02:44 AnimNAS emhttpd: /mnt/animzfs root new: 0 Apr 7 14:02:44 AnimNAS emhttpd: /mnt/animzfs root wrong:0 Apr 7 14:02:44 AnimNAS emhttpd: /mnt/animzfs root missing: 0 Apr 7 14:02:44 AnimNAS emhttpd: /mnt/animzfs root missing already: 0 Apr 7 14:02:44 AnimNAS emhttpd: shcmd (181): /usr/sbin/zfs mount animzfs/Media Apr 7 14:02:44 AnimNAS root: cannot mount 'animzfs/Media': filesystem already mounted Apr 7 14:02:44 AnimNAS emhttpd: shcmd (181): exit status: 1 Any more suggestions? I assume the BTRFS items on lines 3-4 are part of unRAID determining the filesystem, with the modprobe zfs identifying it as a ZFS pool. And then on the 2nd to last line above, it tries to mount the pool and errors out stating the filesystem is already mounted. I'll likely have to rollback again as it's Friday and some of my Plex users will scream foul if they can't access my server, but I'll try and give you some time to respond before I do the rollback. Thanks again for the assistance!

Thanks for that clarification. @Synd on the unRAID Discord server mentioned that it may be related to how the pool was created. We've noticed that the pool is imported using the sdx identification, whereas my pool was created under the old plugin using by-id identification. So far I haven't come across any reason why the pool won't import correctly, but I'm at the stage where I'm about ready to upgrade to 6.12 RC2, destroy the old pool, create a new pool and then restore the data from my backups. I'll hold off for a while just in case you or others can comment on the differences when using sdx vs by-id for creation of the pool. It would save a lot of time if I could get it to import and my data was accessible.

As mentioned I rolled back to 6.11.5 as some of my 'snowbirding' relatives were bugging me about access to their shows. I'll give it another try tomorrow. In the meantime, is there anything I should look at on the 6.11.5 version that might help pinpoint the cause? As mentioned also, the ZFS pool is being mounted and the dataset/subfolders/files are fully accessible before the array is started. I was under the impression that it didn't get mounted until the array started, so if that's what's supposed to happen, I need to figure out how it's being mounted before the array starts.

One more note: I did a rollback to 6.11.5 and interesting to note that under it, even with the array stopped the ZFS pool is mounted and the Media dataset contains the TV folder and all is accessible. Something is obviously messed up with the way the pool is mounted as I didn't think it would mount until I start the array.

Also note that zfs list still shows the Media dataset. [email protected]:/mnt/animzfs# zfs list NAME USED AVAIL REFER MOUNTPOINT animzfs 33.0T 10.5T 204K /mnt/animzfs animzfs/Media 33.0T 10.5T 33.0T /mnt/animzfs/Media