Posts by molnart

    i forgot to mention i am on ZFS. so maybe this is somehow related to how the ZFS datasets are served to NFS?


    if you add it back now, do they work?

    no, they stop working again. i also tried removing the NFS shares and re-adding them (had to do it in the past a few times) but did not help until i removed the subtree_check param


    also realized showmount is not even working when executing locally on the OMV host.


    This is my /etc/exports file. Not sure where those last 3 lines are coming from

    What is installed at this point?

    just the 6.14 kernel. no proxmox-kernel or proxmox-headers meta package - because that installs the 6.17 kernel as far as i understood



    You download from github.

    Thanks, all fixed now

    thanks a lot for your help, i really appreciate it.


    so i did this:

    - sudo dpkg -P proxmox-kernel-6.17

    - run omv-upgrade to see how things are working, but still got the attempt to build dkms for 6.17

    - so i removed the nvidia drivers by running apt remove nvidia-kernel-dkms

    - ran omv-upgrade again, this removed a bunch of unused nvidia dependencies

    - rebooted

    - ran sudo apt install nvidia-driver but this has again attempted (and failed) to build the module for 6.17


    looking at dpkg -l there are still some remnants of the 6.17 kernel

    i assume these dependencies are the proxmox meta packages.

    - so i did a sudo apt remove proxmox-headers-6.17 followed by a sudo apt autoremove. this time there were no 6.17 packages listed by dpkg anymore.

    - rebooted again

    - sudo apt install nvidia-driver - this time modules only built for 6.14

    - rebooted

    - success


    so now everything seems to be OK, but i don't have the proxmox kernel meta packages installed. i wonder if this will will bite me back in the future.


    this is what the fix7to8upgrade script fixes

    where do i find this script? does not seem to be installed on the system


    now i should update by secondary server as well. this one is running bare-metal, so no easy rollback. but also no docker, zfs or nvidia drivers, so hopefully will go smoothly.

    Try: sudo omv-salt deploy run nginx phpfpm

    Thanks, that was it ! The web ui now works.


    But in the UI all the plugins look to be stuck at their 7.x versions


    Not sure if this is a bug or something related to my install. According to dpkg -l the plugin versions are correct:


    The kernel plugin allows installing the 6.14 proxmox kernel on omv8 which should allow compiling the nvidia module.

    yep, that is the one i am running, and its working fine, incl. the nvidia drivers.


    but i still have the 6.17 kernel somewhere lingering around, as every apt update tries to build the nvidia drivers for it, basically blocking me from installing or removing anything.


    the 6.17 kernel is not visible from the UI, but also at the same time 6.14 is NOT marked as the default (i assume 6.17 is then)


    for a moment it appeared (after trying to remove non-proxmox kernels) and when i tried to uninstall it from the UI i got this - basically looks like the unbuild dkms module is blocking the removal of kernel

    Headers don't cause modules to be built. kernels do.

    well, there where no excess kernels in the OMV menu except the one pve kernel booted.


    IF you have pve-headers or proxmox-headers installed

    i am quite positive i had those installed. also the installation procedure from the Kernels menu must have installed them.


    I am guessing you are removing the proxmox header meta package that the zfs plugin has an 'or' dependency on that I linked to above.

    I have removed the kernels from the web interface and then the headers by doing sudo apt remove proxmox-headers-6.5 proxmox-headers-6.8 proxmox-headers-6.11. But in the end it might have removed the meta packages as well, as now when i try to remove the 6.17 kernel (which looks like it does not play with the trixie nvidia drivers) it also tries to remove the meta packages



    i have also noticed that dpkg still lists a lot of kernel residues:


    the big question now is, how to move forward?


    i have a mostly functioning OMV 8 install - meaning ZFS works, docker works, shares work. but i cant get into the web ui because of error 502 and cannot build the nvidia drivers for kernel 6.17, while i also cannot remove it, because it wants to remove the proxmox meta packages too.


    i kinda consider this half-success and am reluctant to go back to OMV7 and start from scratch, given i see little chance to get somewhere under 2-3 hours (twice that much if i have to do it again with all the kernels installed).


    do i have any paths to "fix" the OMV 8 install?

    i am not expecting the upgrade script to handle this, i am trying to upgrade somehow with manual interventions, but so far they are failing. i have the proxmox kernel installed and i ve been running it for the past years. 6.11.11-2-pve is the only kernel i have installed. yet, i cannot remove zfs-dkms, because it wants to remove omv-zfs as well. i could try to remove it with dpkg, maybe it will keep omv-zfs intact, but i am bit reluctant to do that.


    i tried to remove the dkms module via sudo dkms remove zfs/2.3.2 then update to kernel 6.14, but somehow all the old headers i have removed before crept back it, so its again rebuilding the kernel modules for 40 minutes, for 6.8, 6.11, 6.14 and 6.17. the rebuild in this case fails because there is already a newer module in the kernel:


    Code
    Error! Module version 2.3.2-2 for zfs.ko
    is not newer than what is already found in kernel 6.14.11-4-pve (2.3.4-pve1).
    You may override by specifying --force.
    
    Error! Module version 2.3.2-2 for spl.ko
    is not newer than what is already found in kernel 6.14.11-4-pve (2.3.4-pve1).
    You may override by specifying --force.


    Anyhow, tried to update, then reboot, then remove the zfs-dkms package, then reboot again. i had to manually import the zfs pools and unlock them, but i still cannot get into the OMV webui and getting that 502 error.


    there is also a remaining dkms error for nvidia drivers and kernel 6.17 but i don't think thats blocking OMV since i am booted into kernel 6.14

    i have spent at least 6 hours trying to update to OMV 8, with no luck so far.


    one thing that caused extra annoyance was the presence of kernel headers even after removing the kernels itself via the OMV Web UI. This has caused kernel modules to be rebuilt at basically every step, with each rebuild taking 20-30 minutes for the many headers installed - (and the modules get rebuilt even when the kernels are being removed with apt remove).


    the issue with the update is, that PVE kernel 6.17 which comes with OMV 8 is incompatibile with zfs-dkms 2.3.2

    i have tried manually installing kernel and headers 6.14, also trying to remove zfs-dkms post upgrade (since its still stuck at 2.3.2 after the update) but i still could not log in into web ui and the whole zfs service was down.

    if i try to remove zfs-dkms before the upgrade, it wants to remove the zfs plugin as well. also apt autoremove is not removing the packages marked as no longer required.

    I released a new version of omv-extras 7.0.6 that fixed the upgrade on my test system. I was able to reproduce the failure before that.

    i have run into the very same issue as the OP by running omv-release-upgrade. i too have zfs-dkms installed and building the kernel module for kernel 6.17 failed, so i could boot only the 6.11 bookworm kernel. for now i have just reverted to version 7 and did not try any fixes, mostly because i realized that i have too many kernels installed, that are just prolonging the process.


    also i am not sure if i should entirely remove dkms, since i need it (probably) for the nvidia drivers and gpu acceleration. anyhow my understanding from the this thread was that omv-extras 7.0.6 should handle the update fine even with zfs-dkms

    my expectation is, that OMV8 will be out shortly before the support for Debian 12 expires, which will be in June 2026. I would not expect a release before april. also it seems that since OMV 6 it moved to a kind of a "rolling release" model with new features delivered in minor versions, the major ones essentially "just" bumping the Debian release. given that no significant features seem to be around the corner i'd expect the same with OMV8

    i am thinking moving to a more standard REST API could help the OMV ecosystem and potentially bring in plugin developers.


    (as much as i love OMV, i feel their maintainers are getting irritated at every slight joke attempt or feature suggestion)

    I have recently noticed I am not getting any notifications about ZFS scrubs. When checking, i realized the `zed` service is failing. Trying to run zed manually, get the `zfs_unavail_pool` error, but my pools are available, as repoted by `zpool status` - see below.


    What could be causing this?