https://rancher.com/ logo
Join Slack
Powered by
This message was deleted.
# harvester
a
This message was deleted.
👍 1
w
Also - when I tried to run the installer on this machine it got stuck when I went to load the Nics - I'm assuming thats because of the missing driver, it would crash and go back to the first page of the install... Not 100% sure how to fix the installer image to check this! Will raise on the bug-tracker when I get a chance - I needed to get the cluster upto full heath so smashed on with the fix.
b
The driver isn't in the linux kernel?
w
Yep - the harvester image I think is missing this.
b
I mean, does it exist upstream?
Does the nic work with like... Fedora?
w
Not sure - I worked out how to compile it. I suspect it works with some distros and drivers are readily available from the hardware vendor - just not provided in the harvester image meaning you have some fun getting it installed.
I tried a fresh install when things were going pear shaped but the installer just crashed back to the install first page when it got to nics... luckily only issue was this NIC not working so rather than re-install I booted the existing install and patched it (see the repo/issue for info).
b
I don't know how much you trust the AI....
gemma-4-31b
Today at 7:31 AM
To answer your question accurately, it is necessary to distinguish between the upstream Linux kernel (Mainline) and out-of-tree (Vendor) drivers.
The short answer is: Yes, the Realtek RTL8125 is supported in the Linux kernel, but the "patching" process has been a point of contention between the community and Realtek.
1. The Mainline Driver: 
r8169
The Linux kernel does not use a driver specifically named 
r8125
. Instead, the 
r8169
driver is a unified driver designed to support a wide range of Realtek Gigabit Ethernet controllers, including the RTL8111, RTL8168, and the RTL8125.
• Integration: The RTL8125 was integrated into the 
r8169
driver in the mainline kernel.
• The "Patching" Context: For several years, users reported stability issues, "link down" events, and poor performance with the mainline 
r8169
driver on RTL8125 hardware. This led to a fragmented ecosystem where users were encouraged to install a proprietary "patched" driver from Realtek's website.
2. The Out-of-Tree Driver: 
r8125
Realtek provides a standalone driver (often referred to as the 
r8125
driver) that is not part of the official Linux kernel source tree.
So to be clear, I don't work for Suse.
But that's some bullshit on Realtek's end. Their driver is proprietary and from some the sound of it can't be bundled/distributed. legally in a lot of markets. Broadcom and nvidia use to pull this crap all the time.
The "real" way to fix this is to get it working in the linux kernel, which all distros share. The upstream r8169 driver.
Once that happens you pull in the patches to the linux kernel that Harvester uses.
Suse likely can't (legally) just yeet the third party driver into the distro/installer.
Canonnical gets away with crap like that with Ubuntu because of the laws of the country they're in (South Africa) which are very much in their favor.
It wouldn't surprise me if they end up closing this because this is bug for kernel.org not Harvester. The Harvester side would be a backport once it gets fixed there.
But that's just like... my opinion. 🙃
w
Well - I've packaged it on our repository so its pretty easy to download from us if anyone needs it 😉
b
Yeah totally!
w
Had to do the same for the QLOGIC driver...
b
I'm just pointing out that's likely the ONLY paths to get it working.
IANAL - so they might be able to redistribute the proprietary drivers.
w
Just some extra bullshit to tackle during upgrades... its a shame as this board was really good value - though no true OOB (WOL and power on restore enabled - board supports ECC and 3xm2, 4xSATA). Have had 2 server grade boards fail in the last few months...
b
(which you might want to check you're ability to redistribute it via the repo) I'd hate to see you get hit with a lawsuit for trying to do a nice thing.
w
Founds some really nice NanoKVM's... Their licence states GPL-2.0 ??
Copy code
# SPDX-License-Identifier: GPL-2.0-only
################################################################################
#
# r8125 is the Linux device driver released for Realtek 2.5 Gigabit Ethernet
# controllers with PCI-Express interface.
#
# Copyright(c) 2025 Realtek Semiconductor Corp. All rights reserved.
#
# This program is free software; you can redistribute it and/or modify it
# under the terms of the GNU General Public License as published by the Free
# Software Foundation; either version 2 of the License, or (at your option)
# any later version.
#
# This program is distributed in the hope that it will be useful, but WITHOUT
# ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
# FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
# more details.
#
# You should have received a copy of the GNU General Public License along with
# this program; if not, see <<http://www.gnu.org/licenses/>>.
#
# Author:
# Realtek NIC software team <nicfae@realtek.com>
# No. 2, Innovation Road II, Hsinchu Science Park, Hsinchu 300, Taiwan
#
################################################################################

################################################################################
#  This product is covered by one or more of the following patents:
#  US6,570,884, US6,115,776, and US6,327,625.
################################################################################
b
That's a good enough for me reason.
w
"You must provide the source code, keep copyright notices intact, and allow recipients to modify and redistribute it, prohibiting closed-source or proprietary distribution of derivative works" Which is what I have done 🙂
Anyway - hope things good with you - seems your often first to chirp up here 😉
b
I wonder if there's components that have submarine licenses.
I try. Create what we wish existed right?
w
Who knows - but it would be a week case if they got shirty - and if they did I can always take it down (and return the hardware as not fit for purpose ... 🙊 )
b
heh
I would say you could also just buy a $20 Intel card, but with the way these import taxes have been lately...
who knows.
w
ive got loads of kit floating about - but id rather make the board work without extra stuff added - this is a low power setup (well fast cpu - but low watts per thread)
only extra card added was a 2x10GiB SPF PCI card
b
Nice.
w
works like a charm now... starting to understand the setup a bit more now
b
I need to mess around with our ARM stuff more. The overhead has always been attractive.
w
running k8s ontop - now weve added anti-affinity on the cp nodes...
were using AMD Ryzen - this is a small scale on-site development stack for running our dev-ops and our own workloads... I'd definitely go ARM at scale - the only issue is no DC is actively supporting it - and co-locating is a pita... most big providers are still very much in bed with intel. You can get 96 core arm machine for peanuts here - but the issue is you need 3 or them minimum and you cant get the inter-node network speed - its capped, you really want them to have some phat bandwidth between them ....
we're already building all our images as cross/platform (arm+x86) as the dev team are all on macs, and were testing in local k8s (rancher-desktop), just sacrificed one arm studio mad as a slave worker to our cluster so it can build out the arm version of the images...
b
Almost all our new stuff has been AMD.
But we got 9 arm servers right before all the tariffs went crazy.
w
AMD is so cheap and its really good power wise - 65w tdp for 12/24 cores/threads is great power efficiently
9 arm servers... nice!
b
yeah
w
where you based?
b
I work at University of California Santa Barbara.
I work remotely, but still in the area-ish.
w
Wow - awesome - I'm in the UK - running a small web development company
b
Nice!
I don't do much dev work, I'm mostly on the ops side.
w
So your 7 hours behind? I do a lot of dev, qualified as a designer believe it or not! but felt I needed to understand what sits behind what I do to understand how things could be designed to work well...
b
Yeah, getting that broad knowledge is always helpful.
I have some friends that just moved out to Portugal that are devs/owners for a small FOSS company that does websites for inventory management.
How they scale up/down instances is crazy to me because it's not how I would have done it, but it works so much better than what I would have done.
But it's all because they know a ton about their whole stack.
w
Never ceases to amaze me how much effort is often make to over-complicate or obfuscate what is ultimately something simple ... once you get to the basic underlying process....
I find part of the fun though is keeping an open mind and not being afraid to try out new ideas - this is what I love about harvester/rancher - it does help that process if you want to quickly prototype.
b
Absolutely.
There's lots to complain about, but there's always lots to complain about.
And honestly I love that the devs hang out here with us all too.
It seems like it's often getting more and more rare these days for large corps, where that's more normal.
The Harvester team and Suse in general, has been pretty good about keeping most, even if it's not all, of the work out and about for contributions and I 100% appreciate it.
29 Views
Open in Slack
PreviousNext
Powered by