[HOWTO] Instal ZFS-Plugin & use ZFS on OMV

  • thanks for detail reporting @wolffstarr


    I can reproduce your problem regarding creating pool
    whenever device alias is set to `By Id` or `mount point` is not empty
    pool is not created successfully
    this is definetely bug and I will investigate further what's wrong


    Device Alias set to "By ID" (which should be the default, by the way)

    By the way, can you let me know why `By ID` should be default ?





    The configuration object 'conf.system.filesystem.mountpoint' is not unique. An object with the property 'fsname' and the value 'tank' already exists.

    I can also reproduce this problem
    it seems that zfs plugin try to create new fstab entry in omv since mount point is changed
    but collision occur since old data exists
    it can be solved editing manually <fstab> with `vim /etc/openmediavault/config.xml `
    this is becuase that editing mount point via command-line is not accepted by plugin design or you should manually edit config.xml


    for snapshot problem
    I didn't test it, sorry I will test it next post
    but I suspect time-out issue since it is ok when the number of snapshot are small

    OMV3 on Proxmox
    Intel E3-1245 v5 | 32GB ECC RAM | 4x3TB RAID10 HDD
    omv-zfs | omv-nginx | omv-letsencrypt | omv-openvpn
    Click link for more details

    2 Mal editiert, zuletzt von luxflow ()

  • By the way, can you let me know why `By ID` should be default ?


    If you identify the disks by /dev/id zfs integrates them independent from the physical port. The system continues work if you interchange the disks of /dev/sda and /dev/sdb as an example.


    It's a best practise and often recommended to set up zfs pools by-id.


    By the way, is it a possible to integrate an existing pool by-id? If I remember right, this wasn't possible with the old omv3 zfs plugin.


    I hope to find some time this weekend. I want to build up a completely fresh system. At the moment my omv is installed on bare metal. In future I want to use proxmox or esxi with pci passthrough for my dvb tuner card (digital devices cine s2) and my onboard sas/sata controller (lsi3008). I am still unsure which hypervisor would be best to use.


    Even though the question is offtopic: Which hypervisor do you use?


    Greetings Hoppel

    ----------------------------------------------------------------------------------
    openmediavault 6 | proxmox kernel | zfs | docker | kvm
    supermicro x11ssh-ctf | xeon E3-1240L-v5 | 64gb ecc | 8x10tb wd red | digital devices max s8
    ---------------------------------------------------------------------------------------------------------------------------------------

    14 Mal editiert, zuletzt von hoppel118 ()

  • If you identify the disks by /dev/id zfs integrates them independent from the physical port. The system continues work if you interchange the disks of /dev/sda and /dev/sdb as an example.


    It's a best practise and often recommended to set up zfs pools by-id.

    Thanks for letting me why it is best practice



    I hope to find some time this weekend. I want to build up a completely fresh system. At the moment my omv is installed on bare metal. In future I want to use proxmox or esxi with pci passthrough for my dvb tuner card (digital devices cine s2) and my onboard sas/sata controller (lsi3008). I am still unsure which hypervisor would be best to use.


    As I know proxmox use kvm, although I personally didn't use proxmox, I use kvm currently
    And I used esxi shortly
    Esxi is not good at graphic card pci pass through and also support officially specific device
    (But I think your hardware is supported since server mainboard but i'm not sure about it)


    You can install omv as host or can install omv as guest but I prefer to install omv as host
    Why not try guide I wrote?
    [GUIDE] How to install KVM + WebvirtMgr

    OMV3 on Proxmox
    Intel E3-1245 v5 | 32GB ECC RAM | 4x3TB RAID10 HDD
    omv-zfs | omv-nginx | omv-letsencrypt | omv-openvpn
    Click link for more details

  • Hoppel118 got it mostly right, my personal reason for OMV to use By ID as default is that nearly every single time I reboot my NAS, the disks come up in different orders - when I rebooted yesterday, my 5TB was on /dev/sdk, and after the reboot it was on /dev/sdc. No hardware changes were made, it just... reordered the disks. That might be (probably is) a problem with my system, but it's not unheard of either.


    As for the mountpoint thing, I mostly mentioned it because I've shifted the mountpoint via ZFS command line tools (zfs set mountpoint=/foo/bar poolname) on OMV 2.x with the 2.1.0 plugin, and never got the error message. I remember there was a problem getting ZFS to mount for a while, it's in this thread somewhere, so some change to fix that after I made the change might've taken care of it. I do note that I have no entries for ZFS in /etc/fstab on my OMV 2.2.5 install; I'll check and see if they're in there on the 3.0.32 in a little bit.


    Looking at the entries under "mntent" in /etc/openmediavault/config.xml, the only difference I see between the two is "passno" is set to 2 and "hidden" to 0 on ext4 filesystems, wheras ZFS entries are set to "passno" 0 and "hidden" 1 on 2.2.5.

  • Followup, /etc/fstab doesn't have entries for ZFS, but mtab (on both 2.2.5 and 3.0.32) does; also the config.xml entries are virtually identical between the two versions, so I doubt that's got anything to do with it. Might just be that I got it fixed after I changed the mount point last time.

  • I will set default to `by-ID` next release


    I tested omv 2.x, yeah it works as you said, but it is not working correctly (although functionally it is ok)
    `<mntent>` entries in `/etc/openmediavault/config.xml`has following structure
    `<fsname>` is pool name, `<dir>` is mountpoint, `<type>` should be zfs
    After changing mountpoint, you will see 2 `<mntent>` entries in `/etc/openmediavault/config.xml`
    One is old one having old mountpoint and this is garbage data, the other is new one
    Due to `fsname` can be duplicated in OMV 2.x, error dosen't occurs
    But in OMV 3.x, due to `faname` cannot be duplicated, error occurs


    so currently correct way to change mountpoint is
    first, change mountpoint with commandline
    second, change `<mntent>/<dir>` in `/etc/openmediavault/config.xml` to correspoding mountpoint


    Anyway, It seems little cumbersome to change mountpoint, so I will handle this next release

    OMV3 on Proxmox
    Intel E3-1245 v5 | 32GB ECC RAM | 4x3TB RAID10 HDD
    omv-zfs | omv-nginx | omv-letsencrypt | omv-openvpn
    Click link for more details

  • Thanks, good to hear it. Please let me know if there's anything else you want me to test and post here when there's a new release and I'll re-check it.


    If you've got any idea why the timeouts happen with snapshots for the File System section, I'd love to know what's causing it. At first I thought it was because OMV was being told by the ZFS plugin that snapshots were actually filesystems, but that doesn't seem to be the case.

  • next release (3.0.3) will include ..


    1. Snapshot issue is partial solved, it is much more faster than previous implement
    But due to many web UI elements, it is very slow to load when snapshot is large
    This time, web UI is bottle neck.
    It takes 3 seconds per 1000 snapshots
    and I tested up to 4K snapshots
    I think it is enough, isn't it?


    2. set Id by default


    3. bug in addPool is fixed


    not include


    1. due to OMV core, mountpoint changing via commandline problem cannot be easily solved..

    OMV3 on Proxmox
    Intel E3-1245 v5 | 32GB ECC RAM | 4x3TB RAID10 HDD
    omv-zfs | omv-nginx | omv-letsencrypt | omv-openvpn
    Click link for more details

    Einmal editiert, zuletzt von luxflow ()

  • You can install omv as host or can install omv as guest but I prefer to install omv as host
    Why not try guide I wrote?
    [GUIDE] How to install KVM + WebvirtMgr


    If I install omv as a guest it's easier for me to test different versions of omv.


    next release (3.0.3) will include ..


    Great, well done! Do you know when 3.0.3 will get released?


    I will start my tests after that.


    Greetings Hoppel

    ----------------------------------------------------------------------------------
    openmediavault 6 | proxmox kernel | zfs | docker | kvm
    supermicro x11ssh-ctf | xeon E3-1240L-v5 | 64gb ecc | 8x10tb wd red | digital devices max s8
    ---------------------------------------------------------------------------------------------------------------------------------------

  • Released, this plugin works without any modification when openmediavault >= 3.0.34


    https://github.com/OpenMediaVa…lt-zfs/releases/tag/3.0.3


    By the way, there are many bugs in omv-zfs (acually many bugs have exists in OMV 2.x + added while porting)
    Huge rework, refactoring should done for omv-zfs
    I'm going to be busy, so I may not cope with new issues ..

    OMV3 on Proxmox
    Intel E3-1245 v5 | 32GB ECC RAM | 4x3TB RAID10 HDD
    omv-zfs | omv-nginx | omv-letsencrypt | omv-openvpn
    Click link for more details

  • Ok, thank you for this information. At the moment I am also very busy, but I am going to make some tests and report them back here, if I have the time. Before I can do this tests, I have to setup proxmox. ;)


    Greetings Hoppel

    ----------------------------------------------------------------------------------
    openmediavault 6 | proxmox kernel | zfs | docker | kvm
    supermicro x11ssh-ctf | xeon E3-1240L-v5 | 64gb ecc | 8x10tb wd red | digital devices max s8
    ---------------------------------------------------------------------------------------------------------------------------------------

    Einmal editiert, zuletzt von hoppel118 ()

    • Offizieller Beitrag

    3.0.3 is in the zfs-testing repo.

    omv 7.0.4-2 sandworm | 64 bit | 6.5 proxmox kernel

    plugins :: omvextrasorg 7.0 | kvm 7.0.10 | compose 7.1.2 | k8s 7.0-6 | 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!

  • Released, this plugin works without any modification when openmediavault >= 3.0.34


    https://github.com/OpenMediaVa…lt-zfs/releases/tag/3.0.3


    By the way, there are many bugs in omv-zfs (acually many bugs have exists in OMV 2.x + added while porting)
    Huge rework, refactoring should done for omv-zfs
    I'm going to be busy, so I may not cope with new issues ..


    Thanks Luxflow, sorry I didn't reply to this sooner. We've been getting pounded at work leading up to end of fiscal year, and I've just now had a chance to check back in. I'm running burn-in testing on some new disks on my OMV test system right now, but once that's done I'll get the test disks back in place and create a bunch of snapshots to test it out.


    Thanks!

  • Released, this plugin works without any modification when openmediavault >= 3.0.34


    https://github.com/OpenMediaVa…lt-zfs/releases/tag/3.0.3


    By the way, there are many bugs in omv-zfs (acually many bugs have exists in OMV 2.x + added while porting)
    Huge rework, refactoring should done for omv-zfs
    I'm going to be busy, so I may not cope with new issues ..


    Okay, got the drives burned in and just used them. Had a devil of a time getting things going, but all of it was my fault - if this had been a clean install it would've been perfectly fine; I forgot I had the tank mountpoint thing to clear out and a couple of other things.


    Anyhow, I was successfully able to create a mirror pool named "dozer" with two of the drives, mountpoint of /media/dozer, the new default by-id, and ASHIFT=12 as these are 4k sector drives. (I'd like to note that it was nice actually getting errors this time when creation failed; before I was left wondering if it was broken or just taking a while.) I was then able to easily expand it with another mirror VDEV (so now I'm at two mirror vdevs of two drives each for the pool), edited the pool to turn on compression, created a filesystem, edited the filesystem to turn OFF compression, did a couple of write/read tests, and had no issues.


    I then started my snapshot generator. It's currently generating one snapshot a minute, so I'll let that run overnight and see how the handling goes for me - from my end, I'm less worried about the ZFS plugin giving me fits with large numbers of snapshots than I am with the file system screen (And for that matter, any dropdown that asks you to select a device for any reason) being able to load quickly. Not being able to select devices and/or create shared folders basically kills a ton of the usefulness for me.


    I'll go ahead and let it do it's thing, then once I've tested out tons of snapshots on mirrordevs, I will nuke those (I note that while you can delete snapshots through the interface, it's not an option for filesystems or pools on the plugin screen and it's greyed out on the File Systems screen), recreate them as a RAIDZ2, and redo the testing.


    If all goes swimmingly, I'll be putting the plugin into production once the Supermicro CSE-836TQ case I bought today arrives next week. Thanks for taking the time to get it this far Luxflow!

  • Thanks for hard testing @wolffstarr
    Glad to see there is no issue regarding snapshot


    By the way I found way to handle zfs renaming issue (not tested yet, but I think it will works)
    If it works, It will be released in omv-zfs 3.0.4

    OMV3 on Proxmox
    Intel E3-1245 v5 | 32GB ECC RAM | 4x3TB RAID10 HDD
    omv-zfs | omv-nginx | omv-letsencrypt | omv-openvpn
    Click link for more details

  • Thanks for hard testing @wolffstarr
    Glad to see there is no issue regarding snapshot


    By the way I found way to handle zfs renaming issue (not tested yet, but I think it will works)
    If it works, It will be released in omv-zfs 3.0.4


    Glad to be of help, and thanks for the work you've done; I was on the brink of switching to FreeNAS (despite REALLY not wanting to) because of the snapshot issue. My "live" datasets have about 2000 rolling snapshots; I tested just now with 4000 and yeah, it takes a while - about a third of it just the UI displaying it - but that's only in the ZFS screen, it's still less than 30 seconds, and the File System screen and any dropdowns are - if they're slowed at all - only an extra second or two to load.


    I assume by "naming thing" you mean the mountpoint change? I'll definitely test that as soon as it lands then, thanks!

  • Hello guys,


    I am on the way to virtualize my openmediavault3 with proxmox in a kvm.


    The main system is up and running:


    openmediavault 3.0.40
    openmediavault-omvextrasorg 3.3.14
    kernel 4.7.0-0.bpo.1-amd64?


    I need an up-to-date kernel from the backports, because of some hardware I want to passthrough to the kvm to work out-of-the-box.


    After activating the "ZFS repo" and the "ZFS Testing repo" I installed the available updates and activated/installed the plugin "openmediavault-zfs-3.0.3" by the command line:


    http://pastebin.com/ZmF3wpB5


    The process hangs some minutes and ends with the following messages, while building the modules for 4.7.0-0.bpo.1-amd64:



    So I started googling for this error and found the following known issue:


    http://askubuntu.com/questions…lling-zfs-on-ubuntu-15-10


    I restored the vm to the point before I installed the plugin "openmediavault-zfs". After booting the vm I installed "libc-dev" and "spl" with it's dependencies per apt-get and then installed "openmediavault-zfs" again. But this didn't help at all. It stopped with the same messages.



    Do you have any idea for this problem?



    Greetings Hoppel

    ----------------------------------------------------------------------------------
    openmediavault 6 | proxmox kernel | zfs | docker | kvm
    supermicro x11ssh-ctf | xeon E3-1240L-v5 | 64gb ecc | 8x10tb wd red | digital devices max s8
    ---------------------------------------------------------------------------------------------------------------------------------------

    • Offizieller Beitrag

    The zfs modules may not work with the 4.7 kernel yet. It has only been out a couple of days.

    omv 7.0.4-2 sandworm | 64 bit | 6.5 proxmox kernel

    plugins :: omvextrasorg 7.0 | kvm 7.0.10 | compose 7.1.2 | k8s 7.0-6 | 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!

Jetzt mitmachen!

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