OpenSUSE Tumbleweed or Leap?

Hi!

Regarding ext4 vs btrfs:

There were some problems with btrfs when it was introduced in openSUSE Leap, which I think was more than 10 years ago (was it already called “Leap” then?). So I continued using ext4 for openSUSE and SLES until recently…

But now, I am also evaluating Tumbleweed as a successor for Leap, and for Tumbleweed I would always prefer btrfs for the root filesystem at least, because btrfs offers the ability to do a rollback to an earlier fs snapshot if something gets wrong with one of the naturally more frequent and bigger updates.

I can’t remember reading about serious problems with btrfs in SUSE OSses in the last years.

Regards
Michael

Check the description of both and see which one is a better match to your use case.

https://en.opensuse.org/Portal:Leap
https://en.opensuse.org/Portal:Tumbleweed

Tumbleweed

I used it for many years.

I get one major problem in the past due to an update. I had to wait for a fix !

Today I get a new major problem and the support was very bad.
see 1229803 – i can't open a kde session for my default user

My only solution is to go to an old version with btrfs snapshot and to install again Tumbleweed.

My advice to day is you must install Leap 15.6 and later when it will mature the openSUSE “Slowroll”.

btrfs snapshot tech saves my life after an update.

I advise you to use btrfs for your / partition.

Its very simple to use previous snapshot.

When starting openSUSE choose the previous snapshot.
open a kde session. Check all is right.
Then in a text console type “sudo snapper rollback”.
Then restart openSUSE.
Don’t update till there is a fix.

XFS or ext4 for /home ?

XFS is faster than ext4.

You can’t shrink an xfs partition.

I’ve been running openSUSE Tumbleweed for quite some time now on a Lenovo desktop that also has an Nvidia video card. I dare to recommend it, it works practically without problems. For example, in my experience, Debian SID has to be handled much more carefully, I don’t think an average user has any problems using openSUSE Tumbleweed.

At https://en.opensuse.org/openSUSE:Migrate_Leap_to_Tumbleweed they show how migration can be done. There are tutorials online, search for “leap to tumbleweed”.

I once tried migrating myself. I was not successful. However, I learned a lesson: the less your original Leap 15.6 has been “frankensteined” (i.e., couple of extra repos to bend a quite conservative Leap to your will) the easier going you shall have migrating to tumbleweed later.

If you expect that migrating might be a thing for you in the future, why not practise?

Maybe take your time and first install Leap 15.6 and have a brief look at it. Maybe also install the codecs via the packman repository, even if it’s just for the sake of it. If you have the time and the disk space make backups after each step. Then migrate to tumbleweed and see for youself what it might entail in the future.

PS: Embedded here is also a BtrFS vs Ext4 discussion I guess?

BtrFS is one of the more modern file systems out there with a bunch of extra features which were implemented because folks felt they somehow belong to the realm of the file system (btrfs snapshot, send and receive for subvolumes, snapper list and rollback for snapshots, etc).

Again, maybe take your time and try to get a brief overview what BtrFS is all about and what choosing it can offer to you. Then you can make an informed decision if it is worth the up-front effort right now.

I have been using BtrFS for years now and I would recommend becoming friends with it, anyway. Meanwhile, there is also a 3rd party backup option: https://github.com/digint/btrbk I’d recommend, too.

Hope it helps! – Regards, M.

Tumbleweed versus Leap

As other have written, Tumbleweed is a better choice if you need access to the very latest stuff. For example if you want to use the very latest digikam photo editor, or perhaps if you a developer and require a desktop with the latest libraries or most recent python, or what have you. On the other hand Leap will have less churn and the applications will be older and possibly more stable.

I moved from Leap to Tumbleweed when I started to do more Linux based development, I need access to the latest stuff.

Tumbleweed kernel and nvidia
With nvidia, as other have written, there is the Tumbleweed issue of the uncoordinated release timing of the TW default-kernel and proprietary nvidia driver. For example, at the moment the proprietary driver is has not yet been updated for the current TW kernel 6.10, so updating or installing the driver does not work. Somewhat unusually, this has been the case for some weeks (there are unofficial DIY solutions). Normally, the delay is only a few days, and usually only affects major kernel bumps, such as 6.9 to 6.10 (6.9.8 to 6.9.9 is usually pretty risk free).

With TW there are perhaps three main approaches to working around kernel proprietary nvidia-driver mismatches:

  1. Use btrfs as your root filesystem, on a mismatch, rollback and stay on the current TW version, wait and do no distribution-updates until the driver is updated.
  2. Use anything you like as a root filesystem and just hang back if you see a TW release that contains a kernel with a major kernel jump, such as 6.9 to 6.10. Wait for others to jump in the water first and report back. Stop distribution-updates until any report problems are fixed (you might be still be able to selectively update things such as the browser).
  3. Lock the kernel, and possibly the driver. This has the great advantage that it does not hold up distribution-updates. As in 2, just wait and see before allowing the kernel package to update.

I used to do 2, which worked pretty well. I now use 3, which I like very much because I can choose to stick with a stable kernel/driver combo until I have spare time to consider what’s in the update. For 3, I would recommend the nvidia “hard-way” installer over the “easy-way”. The hard-way installer is actually pretty easy and includes options that might help if you’ve locked the kernel (and something like the gcc version has rolled on).

Some folk just don’t use the proprietary driver, I never tried that.

btrfs versus ext4

I think the rough edges have been knocked of btrfs, but it is quite complicated, and you should definitely be sure of how to do a rollback, how not to do recovery (avoid fsck I believe), and how to measure free space (df and du don’t tell the whole truth on btrfs). Perhaps also read up on the structure and exposed metadata. Although /home can live as a btrfs sub-volume, if you ever want to change distro’s or want to reinstall the OS, it might be best kept separate?

Personally, I’ve stuck with ext4 for root and home (each separately). It’s partially because it’s what I know, partially because it’s a lot simpler. I’m used to doing my own backups and recovery, btrfs would just add complexity to my own efforts.

When btrfs was less well refined I did kick the tires extensively in a virtual-box, I would recommend setting up a disposable experiment to practice with it. ( Link to my old Leap 42 btrfs notes that includes some explanation of roots structure, which has subsequently been simplified and improved in lapter version of Leap/TW.)

I’ve read that ext4 was more resistant to corruption due to power failure than xfs (but it’s been a while, so xfs may have improved).

1 Like

See working Nvidia output the easy way below. The issues on your system are caused by locking packages as pointed out also by other devs in your bugreport. G06 and G05 drivers work flawlessly the easy way with the actual Tumbleweed 6.10.x kernel-default.

G06 driver (the easy way) works flawlessly with kernel 6.10.x:

ich@rennsemmel:~> inxi -SG
System:
  Host: rennsemmel Kernel: 6.10.7-1-default arch: x86_64 bits: 64
  Desktop: KDE Plasma v: 6.1.4 Distro: openSUSE Tumbleweed 20240903
Graphics:
  Device-1: NVIDIA GA102 [GeForce RTX 3080 Ti] driver: nvidia v: 550.100
  Display: x11 server: X.Org v: 21.1.12 with: Xwayland v: 24.1.2 driver: X:
    loaded: nvidia unloaded: fbdev,modesetting,vesa gpu: nvidia,nvidia-nvswitch
    resolution: 2560x1440
  API: EGL v: 1.5 drivers: nvidia,swrast
    platforms: gbm,x11,surfaceless,device
  API: OpenGL v: 4.6.0 compat-v: 4.5 vendor: nvidia mesa v: 550.100
    renderer: NVIDIA GeForce RTX 3080 Ti/PCIe/SSE2
  API: Vulkan v: 1.3.290 drivers: N/A surfaces: xcb,xlib
ich@rennsemmel:~> zypper se -si *G06*
Repository-Daten werden geladen...
Installierte Pakete werden gelesen...

S  | Name                          | Type  | Version               | Arch   | Repository
---+-------------------------------+-------+-----------------------+--------+------------------------
i+ | nvidia-compute-G06            | Paket | 550.100-25.1          | x86_64 | nVidia Graphics Drivers
i+ | nvidia-compute-G06-32bit      | Paket | 550.100-25.1          | x86_64 | nVidia Graphics Drivers
i  | nvidia-compute-utils-G06      | Paket | 550.100-25.1          | x86_64 | nVidia Graphics Drivers
i+ | nvidia-driver-G06-kmp-default | Paket | 550.100_k6.9.7_1-25.1 | x86_64 | nVidia Graphics Drivers
i+ | nvidia-gl-G06                 | Paket | 550.100-25.1          | x86_64 | nVidia Graphics Drivers
i+ | nvidia-gl-G06-32bit           | Paket | 550.100-25.1          | x86_64 | nVidia Graphics Drivers
i  | nvidia-utils-G06              | Paket | 550.100-25.1          | x86_64 | nVidia Graphics Drivers
i+ | nvidia-video-G06              | Paket | 550.100-25.1          | x86_64 | nVidia Graphics Drivers
i+ | nvidia-video-G06-32bit        | Paket | 550.100-25.1          | x86_64 | nVidia Graphics Drivers

I recommend a traditional partitioning set-up. Guided partitioning, no LVM/encryption, no expanded Swap for suspend, Ext4 root, no separate home. On my set-up it’s a 512MB /boot/efi, 2GB /swap, and 900+ GB root / as ext4. Works fine today right now with Tumbleweed :slight_smile:

They do now or will do once all the mirrors update. But that’s not the full story. If you read comment #11 in the bug report, you will see this was not the case for a few weeks.

I should add I had been happily using the easy-way fro around two decades. It’s a nice solution providing you accept that one must be wary of blindly dup’ing (distribution-upgrade) when there is a major kernel version bump (or be familiar with rolling back btrfs - reported by many to be easy enough). While I have used the hard-way to help step around the recent delay, I may well return to the easy way now that things should be back to normal (the hard-way has more options, but the easy-way is very convenient, maybe I will just keep the hard-way in reserve).

There is absolutely no need for rollback or for btrfs. “If” there is a real problem with the Nvidia driver, simply boot with the prior kernel. That is why multiversion kernel exists. Way easier, no locking required and you still have an up to date system.

1 Like

Oh yes, very good suggestion. And a strategy I’ve employed in the past. If you already have a working driver, this is the way to go. Apologies for forgetting it.

If the delay is going to be more than a few days, be careful to keep the working kernel, because further dup’s may remove it. That can be done by editing the multiversion.kernels spec in /etc/zypp/zypp.conf (I think the default is latest,latest-1,running). I used to keep a “favorite” fallback kernel and driver by setting multiversion.kernels, but since changing to locking, I’d forgotten about that one. Spoilt for choice.

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.