Update, 43 to 44, stuck at 44% complete (in offline reboot)

I am trying to update from Fedora KDE Plasma from version 43 to 44. The offline reboot part of the update has been stuck at 44% complete for about 2 hours. What do I do now? Is there a way I can bump it out of its slumber?

This is on a ThinkPad P52 with a Quattro P2000 Mobile GPU. I did the three preliminary commands to prep the GPU drivers for the update, but no action was needed because the 580.142 driver was already installed.

(Prior to updating this computer I updated a ThinkPad E560 to 44, with no problems.)

Did you use the command line or KDE Discover to upgrade?

The ESC key should toggle between plymouth and text output.
Try switching to text output, take a photo and post it here. This may reveal some information where it“s stuck and why.

What commands would that be?
system-upgrade is going to install the 595.* drivers if you did not remove the f43 nvidia drivers.

F44:

$ dnf list xorg-x11-drv-nvidia --available
Available packages (available for reinstall, available for upgrade)
xorg-x11-drv-nvidia.x86_64 3:595.58.03-1.fc44 rpmfusion-nonfree
xorg-x11-drv-nvidia.x86_64 3:595.58.03-1.fc44 rpmfusion-nonfree-nvidia-driver

F43:

$ dnf list xorg-x11-drv-nvidia --available  --releasever 43
Available packages (available for reinstall, available for upgrade)
xorg-x11-drv-nvidia.x86_64 3:580.142-1.fc43 rpmfusion-nonfree-updates
xorg-x11-drv-nvidia.x86_64 3:580.142-1.fc43 rpmfusion-nonfree-nvidia-driver

MatH, I did the installation from the command line.

Mark, nice trick with the Esc key (thanks).

The commands I used were:

[Edit]: First, sudo dnf update

sudo dnf update -y # and reboot if you are not on the latest kernel
sudo dnf install akmod-nvidia # rhel/centos users can use kmod-nvidia instead
sudo dnf install xorg-x11-drv-nvidia-cuda #optional for cuda/nvdec/nvenc support

Followed by (for the driver #): modinfo -F version nvidia
All 4 from: https://rpmfusion.org/Howto/NVIDIA#Current_GeForce.2FQuadro.2FTesla

[Edit]:  But now I see that I grapped the wrong group of three. With update already done, the others should have been:
sudo dnf remove xorg-x11-drv-nvidia\* akmod-nvidia\*
sudo dnf install akmod-nvidia-580xx xorg-x11-drv-nvidia-580xx

(Fuzzy-headed after an all-nighter on a nasty printer/scanner problem.)

Yes and no.
While on f43 that driver is not in the rpmfusion repo for f43. You would need to add --releasever=44 so dnf could find the 580xx driver.
sudo dnf install akmod-nvidia-580xx xorg-x11-drv-nvidia-580xx-cuda --releasever=44 would be the appropriate command while still on f43.

1 Like

works also during boot.

it looks like the upgrade installed most of the packages.
You could try to login in another terminal with a root account ( if it’s unlocked == root account has a PW set )

Not sure if CTRL-c can have any impact here, probably not. It’s worth trying.

If you have root account unlocked, then login and reboot otherwise you’ll have to reset the system.
Then try to boot the system and see how far it gets. Otherwise add ā€˜3’ to the kernel cmd line at the grub stage to boot multiuser only.

then:

01/ save the journal for the failed system-upgrade
journalctl -b -1 --no-hostname --no-pager > journal_system-upgrade-failed
-b -1 is the previous session, if you rebooted more than once then change the index -1 accordingly to -2 / -3 / -4 etc. and make the file available
e.g. fpaste < journal_system-upgrade-failed

02/ check outpuf of dnf repolist that repositories are f44 and not f43
also hostnamectl might already report the Operating System as Fedora 44

03/ this should hopefully complete the upgrade online

sudo dnf distro-sync \
    --refresh \
    --releasever=44 \
    --allowerasing \
    --skip-broken \
    --setopt=protected_packages=

restart

04/ nvidia drivers

sudo dnf remove xorg-x11-drv-nvidia 
sudo dnf install akmod-nvidia-580xx  xorg-x11-drv-nvidia-580xx-cuda

wait a few minutes and restart

1 Like

Thank you, Mark – this is terrific help. I will give it try this evening when I am back home and will report back here.

[EDIT}: Ctl-C didn’t help. But on reboot I got the updated Plasma DE still under Fedora 43. I followed your instructions, got the 44 update, and rebooted.

Your remove command for Nvidia executed successfully. But the command with akmod stalls at the 3rd task, and all the rest, through the 24th task are failures with red text. The failure mode is a Status 403, and I have retried several times about a half-hour appart (in case the server is busy). The errors are like:

Status code: 403 for https://ftp.fi.muni.cz/pub/linux/rpmfusion/nonfree/fedora/development/44/Everything/x86_64/os/Packages/a/akmod-nvidia-580xx-580.142-1.fc44.x86_64.rpm (IP: 147.251.48.205) - https://ftp.fi.muni.cz/pub/linux/rpmfusion/nonfree/fedora/development/44/Everything/x86_64/os/Packages/a/akmod-nvidia-580xx-580.142-1.fc44.x86_64.rpm

looks like your rpmfusion repositories are outdated and still use the development path nonfree/fedora/development/44 should be nonfree/fedora/release/44
see also Making sure you're not a bot!

the latest rpmfusion repos are

$ rpm -qa rpmfusion\*-release
rpmfusion-nonfree-release-44-3.noarch
rpmfusion-free-release-44-3.noarch

I ended up with three curl errors in the process from those commands. So I followed as best I could what I found in this thread: Akmod-nvidia-580xx and similar packages not found - #9 by computersavvy

I am already on F44. Following that thread I end up with 580.142 installed instead of 580xx. What I did is:

$ sudo dnf remove *nvidia* --exclude nvidia-gpu-firmware
$ sudo dnf install akmod-nvidia-580xx /
xorg-x11-drv-nvidia-580xx-cuda --refresh

That completed with no errors. Then:

$ modinfo nvidia | grep ^version
version: 580.142 [EDIT: confirmed after a restart.]

$ repolist (only relevant ones shown here)
fedora Fedora 44 - x86_64
fedora-cisco-openh264 Fedora 44 openh264 (From Cisco) - x86_64
rpmfusion-free-rawhide RPM Fusion for Fedora Rawhide - Free
rpmfusion-nonfree RPM Fusion for Fedora 44 - Nonfree
rpmfusion-nonfree-nvidia-driver RPM Fusion for Fedora 44 - Nonfree - NVIDIA Driver
rpmfusion-nonfree-steam RPM Fusion for Fedora 44 - Nonfree - Steam
rpmfusion-nonfree-updates RPM Fusion for Fedora 44 - Nonfree - Updates

As I understand that thread, the above rpmfusion-nonfree repo has the 580xx driver, but it didn’t install.

The thread says rpmfusion-nonfree-nvidia-driver repo ā€œshould haveā€ the 580xx driver but does not. So do I need to remove that nvidia repo to avoid a conflict (and if so how do I do that)?

(Also, I do not have the rpmfusion-nonfree-nvidia-driver-debuginfo or the rpmfusion-nonfree-nvidia-driver-source repos, which the thread says can be a problem.)

It seems that either I have done something wrong, or there is a conflict in my repo list, or I am missing a needed repo.

you have installed akmod-nvidia-580xx, which currently installs driver version 580.142 from the 580.xx driver branch. Only the names of the rpm packages have changed from akmod-nvidia to akmod-nvidia-580xx!
see rpm -qa akmod-nvidia\* xorg-x11-drv-nvidia\*

On F44, akmod-nvidia follows the production branch currently 595.xx, which dropped support for the GPU in your system.

legacy drivers like nvidia-390xx, nvidia-470xx and the nvidia-580xx are available only on the rpmfusion-nonfree repositories. A few days ago nvidia released driver 580.159.03 with many MST fixes. Maybe you will be able to connect both displays to the TB3 dock again.

1 Like

Wonderful. Thanks ending my confusion, and for all your help here.

1 Like

To verify the packages installed (regardless of the driver version) use the command
dnf list --installed \*nvidia\* which should show the 580xx packages. Note though that the nvidia-modprobe and nvidia-persistenced packages are the 595 versions which is expected.

Thanks! Added to my notes.

[EDIT]: In the listing I got, the akmod line is for 580.142-2, and there are several xorg-11 lines for 580.142-1.

This actually a mismatch, and if so is it problematic and how might I do an update of the xorg packages?

Not a mismatch.
There was a minor bug in the akmod package version 580.142-1 that was fixed with a new build and resulted in the 580.142-2 version. The remaining packages were not rebuilt and remained at the 580.142-1 release version.

Again, the version numbers are correct, and as long as package names show the 580xx within their names every thing is good.

Well, here’s a curve ball if ever there was one!

My installation thinks it is version 44. $ lsb_release -d gives: ā€œFedora Linux 44 (KDE Plasma Desktop Edition).ā€

However, at bootup I just now happened to notice that Grub has no version 44 choice, only version 43. The choices are for versions ā€œ6.19.14-200.fc 43. x86_64ā€ and 6.19.10-200.

After discovering this mismatch I tried doing refreshes and updates for 44, 44-1 and 44-2, but the operating system is having none of that.

Try completing the upgrade from the command line.
sudo dnf distro-sync --refresh --releasever=44
Follow the instructions from before

You are perhaps having this issue

Check if you are in this case.

What gives also: dnf list --installed kernel\*

This is one of the worst nightmares when trying to upgrade… even worse than an update refusing to install :fearful:

Here is a screenshot from that command:

Have you checked in the issue I mentioned if you are in this case ?
Ie: what gives (as preformatted text please, using the </> button):

  • sudo ls -d ā€œ/boot/efi/$(cat /etc/machine-id)ā€`