ZFS Plugin: there are some changes every time I open "Pools" tab

petrovich666

Have you crossed checked the output of  zfs get all | grep 'omvzfsplugin:uuid' against the mntent entries in the config.xml file? There should be an entry per zfs filesystem with matching UUID values. Use:

# omv-confdbadm read conf.system.filesystem.mountpoint  | jq -r '.[]|select(.type=="zfs")'

(ref: https://docs.openmediavault.or…/tools/omv_confdbadm.html) Is the cross check correct?

If the cross check is OK, and you don't mind exporting your pools yet again, then the following step at the CLI will force the zfs plugin to set new values for omvzfsplugin:uuid which might cure your problem.

1. Recursively remove omvzfsplugin:uuid property from all pools

zfs inherit -r omvzfsplugin:uuid BackupPool
zfs inherit -r omvzfsplugin:uuid FastPool
zfs inherit -r omvzfsplugin:uuid MainPool

2. Export all pools

zpool export BackupPool
zpool export FastPool
zpool export MainPool

3. Import all pools

zpool import BackupPool
zpool import FastPool
zpool import MainPool

Now go to the zfs pool tab and apply the changes. Test if yellow configuration banner still appears every time the zfs pool tab is accessed.

petrovich666 There was a typo in my previous post.

Zitat von Krisbee

If the cross check is OK, and you don't mind exporting your pools yet again, then the following step at the CLI will force the zfs plugin to set new values for omvzfsplugin:uuid which might cure your problem.

That sentence should have started with "If Cross check is NOT OK ... ". Clearly that's the root cause as you've discovered. It's your choice, but if you edit config.xml directly you need to know what salt commands etc. to issue afterwards. Hence my suggestion to re-generate zfs plugin property via listed commands and apply change via the WebUI which should ensure a consistent system.

chente Thanks.

petrovich666 I can't remember if just rebooting after removing the plugin property from pool at the CLI would have been sufficient to regenerate plugin UUIDs on ZFS filesystems instead of trying to export a busy pool.

Now you can see my dislike of directly editing the config.xml. Maybe it's a simple error in the config file. Try comparing working pool entries with BackupPool entry in the config schema. It should be of this form:

<mntent>
        <uuid>8cf67f77-349e-4192-8b16-493de3e52bc5</uuid>
        <fsname>pool2</fsname>
        <dir>/pool2</dir>
        <type>zfs</type>
        <opts>rw,noatime,xattr,posixacl,casesensitive</opts>
        <freq>0</freq>
        <passno>0</passno>
        <hidden>1</hidden>
        <usagewarnthreshold>0</usagewarnthreshold>
        <comment/>
      </mntent>

Otherwise, I vaguely remember seeing a similar error on the forum elsewhere.

Hi !

I have similar problems

omv-confdbadm read conf.system.filesystem.mountpoint | jq -r '.[]|select(.type=="zfs")'

returns me the list of my pool with datasets where almost all datasets have two entries with different uuid.

I tried delete as recommended with

zfs inherit -r omvzfsplugin:uuid raid

but I got fatal errors from many other services. It doesn't work for me.

How safely to fix duplicates uuid for datasets in OMV config/database?

Shperrung You cannot fix your problem by simply using zfs inherit -r omvzfsplugin:uuid raid in isolation. In fact it make matters worse. Hopefully you're able to reboot your OMV system and see if your zfs pool is correctly imported. You should have started by deleting the duplicate entries from the config database.

Only when you can import your pool and a new set of uuid values have been applied can you proceed to compare zfs pool UUID values to those in the OMV config.xml file.

Then the safe way to delete duplicate entries in the OMV config.xml files is to (a) make a copy of the file, (b) use the appropriate omv-confdbadm command to delete all duplicated entries ¹ , and then (c) use the correct omv-salt command to deploy the changes ² .

Rather than attaching files as txt docs, please either post them between code tabs or place them on pastebin. I have looked at your config.xml extract. What is noticeable is many of the pairs of duplicated entries have different mount options and/or long lists of mount options that include duplicated options. Only a few of these entries appear correct, most do not. You compare the values of the "opts" field with the output of the mount command. It looks like you need to delete all the ZFS entries, before trying to generate a new set.

It would be useful to know how this problem arose. How was your pool created, inside or outside OMV? Which version of OMV are you using? What action/change to your pool or dataset(s) triggered this problem? Have you made multiple attempts to export and import your pool? Have you been trying to modify zfs filesystems mount options at the CLI?

As fas as I can tell, if you can import your pool and all zfs filesystems a have a valid non-zero omvzfsplugin:uuid property then you should delete all zfs filesystem entries from the config.xml file and then use the correct omv-salt command ² to get your OMV system in a consistent state. At that point you should be able to apply any outstanding config changes on the zfs page of WebUI.

1. https://docs.openmediavault.or…/tools/omv_confdbadm.html - see bottom of page.

2. See #7 above.

Krisbee , thank you for your recommendation! I solved my issue with "Apply changes" banner on "zfs" page of OMV gui.

The problem was in duplicated mountpoints for datasets in "omv-confdbadm" output. I compared two lists

omv-confdbadm read conf.system.filesystem.mountpoint | jq -r '.[]|select(.type=="zfs")' > omv-confdbadm.txt

and

zfs get all | grep 'omvzfsplugin:uuid' > omvzfsplugin-uuid.txt

All wrong mountpoints in omv-confdbadm which hasn't real dataset with the same uuid I deleted using this

omv-confdbadm delete --uuid fe5746bd-efa2-4a75-b6b4-f6618138744d conf.system.filesystem.mountpoint

I don't have ideas why it happened. I made clear OMV7 installation but I imported forcedly pool previously attached in OMV6. Also I did "zfs upgrade" but I don't remember before or after that issue has appeared. Here are two mountpoints from omv-confdbadm

Wrong:

{
"uuid": "fe5746bd-efa2-4a75-b6b4-f6618138744d",
"fsname": "raid/backups",
"dir": "/mnt/raid/backups",
"type": "zfs",
"opts": "rw,noatime,xattr,noacl,casesensitiverw,noatime,xattr,posixacl,casesensitive",
"freq": 0,
"passno": 0,
"hidden": true,
"usagewarnthreshold": 0,
"comment": ""
}

Right:

{
"uuid": "df15bc20-81d9-4d93-866e-5187bfd64774",
"fsname": "raid/backups",
"dir": "/mnt/raid/backups",
"type": "zfs",
"opts": "rw,noatime,xattr,posixacl,casesensitive",
"freq": 0,
"passno": 0,
"hidden": true,
"usagewarnthreshold": 0,
"comment": ""
}

Wrong mountpoint has 3 repeating "opts".

Zitat von Shperrung

I don't have ideas why it happened. I made clear OMV7 installation but I imported forcedly pool previously attached in OMV6. Also I did "zfs upgrade" but I don't remember before or after that issue has appeared. Here are two mountpoints from omv-confdbadm

Shperrung I'm glad you were able to resolve your problem. I'll have to test if a pool created in omv6 can be correctly imported into a clean install of OMV7 whether it had been properly exported or not.. I would have expected that to work

ryecoaaron  Krisbee

My pool is very old. It was created in Xigmanas 4 years ago.

My clean OMV7 installation was followed with problems in incompatibility of zfs-plugin installed on Debian kernel and PVE-kernel installed after.

I reinstalled twice OMV7 untill I guess that zfs-plugin must be installed after PVE-kernel is turned on.

Anyway, such problem is rare. We may close "ticket". Thank you!