ZFS not working after fresh install

  • After a fresh install of OMV6 (amd64) on an i5 based PC and installation of ZFS Plugin, I get this error:


    Code
    modprobe: FATAL: Module zfs not found in directory /lib/modules/5.16.0-0.bpo.4-amd64

    What am I missing?

    OMV6 i5-based PC

    OMV6 on Raspberry Pi4

    OMV5 on ProLiant N54L (AMD CPU)

  • After a fresh install of OMV6 (amd64) on an i5 based PC and installation of ZFS Plugin, I get this error:


    Code
    modprobe: FATAL: Module zfs not found in directory /lib/modules/5.16.0-0.bpo.4-amd64

    What am I missing?

    you missed install first promox kernell

  • Thank you. That worked.

    Would it be a good idea to install Proxmox Kernel by default when installing ZFS Plugin?

    It seems that everyone has the same problem and then asking the same question...

    OMV6 i5-based PC

    OMV6 on Raspberry Pi4

    OMV5 on ProLiant N54L (AMD CPU)

    • Offizieller Beitrag

    Thank you. That worked.

    Would it be a good idea to install Proxmox Kernel by default when installing ZFS Plugin?

    It seems that everyone has the same problem and then asking the same question...

    That would go against giving the NAS administrator freedom to configure it according to their needs.

    As for the information provided on this, there is a doc on the wiki for the ZFS plugin. It's still in writing, but you can already see that it talks about the proxmox kernel. https://wiki.omv-extras.org/doku.php?id=docs_in_draft:zfs

  • I see. Freedom is good but it just doesnt work without Proxmox. So there is no option.

    Its a bit like saying let the people install Portainer and freely decide if they want Docker.

    OMV6 i5-based PC

    OMV6 on Raspberry Pi4

    OMV5 on ProLiant N54L (AMD CPU)

    • Offizieller Beitrag

    I see. Freedom is good but it just doesnt work without Proxmox. So there is no option.

    Its a bit like saying let the people install Portainer and freely decide if they want Docker.

    Actually it does work without the proxmox kernel. The proxmox kernel just makes it easy. But if you don't want to use it, it's not necessary. The system must compile the linux kernel on every update in this case.

    The first few months of using ZFS I didn't install the proxmox kernel and didn't have excessive problems.

  • I see. Then it must be system related. But on this (new) server (i5-6th gen- 8GB Ram, 4 SSDs) ZFS did not work without installing Proxmox (see first post)...And I have another older ZFS server with different hardware (AMD CPU, 2GB Ram, 4x HDs), where it also didnt work without Proxmox.

    What hardware are you using?

    OMV6 i5-based PC

    OMV6 on Raspberry Pi4

    OMV5 on ProLiant N54L (AMD CPU)

    • Offizieller Beitrag

    Intel Pentium Gold G6400

    16GB Ram

    Motherboard Gigabyte B460M DS3H V2

    5x4TB WD Red - RaidZ1


    I wouldn't think too much about this topic. At first I asked myself the same questions as you. Since I installed the proxmox kernel I have not cared about ZFS at all for over a year, maybe two years.

    The only precaution is not to install a kernel that is too recent, which may cause instabilities. In the plugin there are usually several to choose from.

    • Offizieller Beitrag

    Would it be a good idea to install Proxmox Kernel by default when installing ZFS Plugin?

    No. The zfs plugin works on all architectures and the proxmox kernel is only amd64.

    Freedom is good but it just doesnt work without Proxmox.

    It does work. It is just very hard to get the module to compile right with just apt dependencies (only real option for the plugin). It is very fixable from the command line if you insist on using the Debian kernel. I've built the zfs modules on an RPi as a concept and it works (has a lot of issues though).

    But on this (new) server (i5-6th gen- 8GB Ram, 4 SSDs) ZFS did not work without installing Proxmox (see first post)...And I have another older ZFS server with different hardware (AMD CPU, 2GB Ram, 4x HDs), where it also didnt work without Proxmox.

    If zfs-dkms worked better, this would be an issue. But just running the proxmox kernel on those system fixes the issue. I run the proxmox kernel on my systems even when not using zfs.


    I would think about using btrfs since support in OMV is so much better.

    omv 7.0-32 sandworm | 64 bit | 6.5 proxmox kernel

    plugins :: omvextrasorg 7.0 | kvm 7.0.9 | compose 7.0.9 | cputemp 7.0 | mergerfs 7.0.3


    omv-extras.org plugins source code and issue tracker - github


    Please try ctrl-shift-R and read this before posting a question.

    Please put your OMV system details in your signature.
    Please don't PM for support... Too many PMs!

  • I would think about using btrfs since support in OMV is so much better.

    The BTRFS support in OMV is better but I think ZFS is a much better products to use.

    - ZFS allow creating snapshots of any folder while BTRFS needs you to define your folder structure prior to taking snapshots - thats almost impossible to do

    - ZFS offers RAID as well and it works. I had a disk break and the resilvering was as easy as sticking a new disk in and pressing enter

    - ZFS is very fast as it allows to integrate 2 types of SSD caches

    - ZFS can compress on the fly

    ...

    OMV6 i5-based PC

    OMV6 on Raspberry Pi4

    OMV5 on ProLiant N54L (AMD CPU)

    • Offizieller Beitrag

    - ZFS offers RAID as well and it works. I

    So does btrfs. The btrfs 5/6 issue isn't any worse than mdadm raid 5/6. zero issues with btrfs raid 0/1/10.

    - ZFS can compress on the fly

    btrfs can as well. zfs can dedupe which btrfs can't but dedupe is very expensive with zfs.


    ZFS allow creating snapshots of any folder while BTRFS needs you to define your folder structure prior to taking snapshots - thats almost impossible to do

    why not take a snap of the parent folder?


    I mean you can come up with all of the reasons zfs is better but since I am the only one maintaining the zfs plugin and I don't really use zfs, it is never going to be great or be as integrated as btrfs. While I will try to keep maintaining it, I would love to not to.

    omv 7.0-32 sandworm | 64 bit | 6.5 proxmox kernel

    plugins :: omvextrasorg 7.0 | kvm 7.0.9 | compose 7.0.9 | cputemp 7.0 | mergerfs 7.0.3


    omv-extras.org plugins source code and issue tracker - github


    Please try ctrl-shift-R and read this before posting a question.

    Please put your OMV system details in your signature.
    Please don't PM for support... Too many PMs!

  • The btrfs 5/6 issue isn't any worse than mdadm raid 5/6. zero issues with btrfs raid 0/1/10.

    yes but needs a 2nd package. ZFS is integrated - easy to maintain


    btrfs can as well. zfs can dedupe which btrfs can't but dedupe is very expensive with zfs.

    And can you switch compression on/off with BTRFS?


    why not take a snap of the parent folder?

    Because that defeats the purpose of snapshot. Then its a full backup.

    If I install an update of a folder I dont want a snapshot of the entire disk. I just want to be able to roll back 1 folder.

    While I will try to keep maintaining it, I would love to not to.

    Haha. I'd really appreciate if you keep it. We could even add a few more features :)

    OMV6 i5-based PC

    OMV6 on Raspberry Pi4

    OMV5 on ProLiant N54L (AMD CPU)

    • Offizieller Beitrag

    yes but needs a 2nd package. ZFS is integrated - easy to maintain

    What? zfs needs many more packages. btrfs is native to the kernel and only needs the util package. With the proxmox kernel, the zfs module is builtin but there are still a bunch of packages needed.

    And can you switch compression on/off with BTRFS?

    Yes.

    Because that defeats the purpose of snapshot. Then its a full backup.

    If I install an update of a folder I dont want a snapshot of the entire disk. I just want to be able to roll back 1 folder.

    I'm not sure why the parent folder has to be the full disk. Snapshot on OMV are done at the shared folder level.

    We could even add a few more features

    I keep hearing that but no one wants to help and it gives me more to maintain. Good deal for me.

    omv 7.0-32 sandworm | 64 bit | 6.5 proxmox kernel

    plugins :: omvextrasorg 7.0 | kvm 7.0.9 | compose 7.0.9 | cputemp 7.0 | mergerfs 7.0.3


    omv-extras.org plugins source code and issue tracker - github


    Please try ctrl-shift-R and read this before posting a question.

    Please put your OMV system details in your signature.
    Please don't PM for support... Too many PMs!

  • What? zfs needs many more packages. btrfs is native to the kernel and only needs the util package. With the proxmox kernel, the zfs module is builtin but there are still a bunch of packages needed.

    ZFS is 1 package and includes file system and fully integrated RAID manager - BTRFS is only a filesystem and requires a completely separate package for RAID management on top of it. The more layers you have, the more can go wrong (exponentially), especially when updates come through.


    I'm not sure why the parent folder has to be the full disk. Snapshot on OMV are done at the shared folder level.

    With BTRFS snapshots can be taken on the BTRFS Root or BTRFS Subvolumes but they have to be defined before hand.

    If I decide I need a snapshot on an existing folder deep inside the file structure and have not defined that specific folder as a BTRFS Subvolume before hand, I must take a snapshot of Root, which is just wrong. If I need to roll back that single folder later, everything under Root will rolled-back, which would be a disaster.

    ZFS allows for flexible snapshots anytime with one command.




    root on which a snapshot can be taken.

    As the filesyste ost of the time dont know how your file system structure will develop.

    OMV6 i5-based PC

    OMV6 on Raspberry Pi4

    OMV5 on ProLiant N54L (AMD CPU)

    • Offizieller Beitrag

    ZFS is 1 package and includes file system and fully integrated RAID manager - BTRFS is only a filesystem and requires a completely separate package for RAID management on top of it. The more layers you have, the more can go wrong (exponentially), especially when updates come through.

    We must be talking about different packages. I am referring to debian packages and zfs requires at least 7 packages. But zfs goes wrong just on startup more than any filesystem I have ever seen because it has to be imported on boot every time. It has 8 systemd services to make it "work" with Linux. It is very alien as a filesystem. I didn't mean to turn this into a waste of time. just a simple suggestion. If it doesn't work for you, don't use it.

    If I decide I need a snapshot on an existing folder deep inside the file structure and have not defined that specific folder as a BTRFS Subvolume before hand, I must take a snapshot of Root, which is just wrong. If I need to roll back that single folder later, everything under Root will rolled-back, which would be a disaster.

    ZFS allows for flexible snapshots anytime with one command.

    It is amazing that I have used Linux for 25 years and never needed this ability.

    omv 7.0-32 sandworm | 64 bit | 6.5 proxmox kernel

    plugins :: omvextrasorg 7.0 | kvm 7.0.9 | compose 7.0.9 | cputemp 7.0 | mergerfs 7.0.3


    omv-extras.org plugins source code and issue tracker - github


    Please try ctrl-shift-R and read this before posting a question.

    Please put your OMV system details in your signature.
    Please don't PM for support... Too many PMs!

  • It is amazing that I have used Linux for 25 years and never needed this ability.So lets say you have subfolders for each

    So lets say you create folders for each Docker Application and want to test a new version of one application. You take your Root snapshot and 2 weeks later there were 1000s of changes to folders under Root but you realise the Docker App update was not stable.

    How do you roll back only the unstable Docker app?

    OMV6 i5-based PC

    OMV6 on Raspberry Pi4

    OMV5 on ProLiant N54L (AMD CPU)

    • Offizieller Beitrag

    How do you roll back only the unstable Docker app?

    Can't remember the last time I had to do that. I test in a non-production location if I am that worried about an update. If I did need to do that, I would restore the necessary files from backup.

    omv 7.0-32 sandworm | 64 bit | 6.5 proxmox kernel

    plugins :: omvextrasorg 7.0 | kvm 7.0.9 | compose 7.0.9 | cputemp 7.0 | mergerfs 7.0.3


    omv-extras.org plugins source code and issue tracker - github


    Please try ctrl-shift-R and read this before posting a question.

    Please put your OMV system details in your signature.
    Please don't PM for support... Too many PMs!

  • o does btrfs. The btrfs 5/6 issue isn't any worse than mdadm raid 5/6. zero issues with btrfs raid 0/1/10.

    So I tried the BTRFS RAID. Sorry to bring this up, but the File Systems Plugin probably has a little display bug:

    My tested Setup:

    2x SSD 250GB

    1x SSD 500GB


    In the Plugin I add all 3 SSDs into one BTRFS filesystem in RAID1. What I would expect is a BTRFS device with 250Gb available (and mirrored to the other 250SSD and the 500SSD). I would have thought that the RAID1 device can only be as large as the smallest disk in the RAID.


    However, the File System plugin shows me 500GB as "available".


    When I do a df -h it shows exactly what I expected:

    /dev/sda as "size" 500G and "available" 250G

    OMV6 i5-based PC

    OMV6 on Raspberry Pi4

    OMV5 on ProLiant N54L (AMD CPU)

Jetzt mitmachen!

Sie haben noch kein Benutzerkonto auf unserer Seite? Registrieren Sie sich kostenlos und nehmen Sie an unserer Community teil!