All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Debian 12 upgrade to kernel 6.1.0-34 & network interface issue
I did an apt upgrade on Debian 12 to the latest 6.1.0-34 kernel on several of my vps today, and ran into a network issue. I'll mention it here in case someone else has the same issue.
The network didn't start after rebooting. The network scripts use "eth0" but after rebooting, the interface name had changed to something else, like ens3.
virtio_net virtio0 ens3: renamed from eth0
Normally, this is fixed with a grub configuration in /etc/default/grub, and then update-grub.
GRUB_CMDLINE_LINUX="net.ifnames=0 biosdevname=0"
I connected with vnc and found, for some reason, there was no /etc/default/grub file. On several of my vps, the same problem. Either there was never a file, and it just became an issue with this update, or something happened to it, I'm not sure.
But there was a grub.ucf-dist (that had the right grub cmdline). I copied it to /etc/default/grub, did an update-grub, and rebooted. And that fixed it.
Comments
Do you do anything on your VMs that most likely don't do?
Which scripts?
might be the friendly cloud-init botching up?
Nothing unusual that I can think of.
I guess not scripts, but the configuration file: /etc/network/interfaces
So the issue was caused by the installed Debian template
I've had the problem happen before, maybe a year or two ago, which is why I was familiar with the solution. But these are newer vps and I probably hadn't updated /etc/default/grub with the fix, because the network interface names were already eth0.
I had this issue today with my vps at: Virtono, HostSailor, and HostDare.
I didn't have a problem at RackNerd, probably because I had already updated /etc/default/grub with the cmdline earlier because it came using a different interface name instead of eth0.
I also didn't have a problem at Host-C. There is an /etc/default/grub file there, and it doesn't have "net.ifnames=0 biosdevname=0" in GRUB_CMDLINE_LINUX. But, despite that, it uses "eth0" for the network interface still.
But the Host-C kernel is a little different from the others:
another reason to not use provider templates. Stopped doing that a while ago and thank god me cuz i can't go 1 by 1 fixing it
You want to say that you've manually installed system on all your hundreds servers? Insane. Or did you use any kind of automation?
Yes, insane, but also sound, providers make small modifications from default ISO release all the time, disk partitioning and even filesystems will be different across VMs. Imagine dealing with an issue like OP where hundreds of servers all act differently.
But yes, I'd also be very interested in a way to efficiently automate this across platforms, if there is any.
Same CP and likely the same provided template? I haven't used any of the above providers.
Fair point and I completely agree with it, but what I mean is that zGato must be insanely patient to install same system manually from an ISO each time. Hopefully I am wrong and there is some trick to make this kind of task more faster (having traditional automation tools like Terraform unavailable)
Maybe something similar to https://github.com/bin456789/reinstall ?
If you have 100 servers, then at that scale you will have moved away from VPS to dedicated servers so not relevant.
But for VPS you would use a unattended install: https://wiki.debian.org/AutomatedInstallation
Which would be no different to clicking in the Provider's UI.
Yes,
All providers should allow custom ISOs, and its a shame a few don't.
Unfortunately kind of manually in (almost) every VM I have. I do kinda have a working ISO preseed file and that skips me a lot of headache but doesn't save me from having to go provider 1 by 1 VNC and doing the whole process.
Plus some providers make it real hard to install your own O.S
For everything else, ansible playbook and that sets everything up.
I do still use a different SSH port and keypair for each server though, kinda mental
You must not know zGato and why he has hundreds (literally) of VPS in different parts of the world with widest ever seen variety of networks used.
and these images will be supported by all VPS providers that he uses?
that's kinda the main issue, most providers don't even let me mount an ISO so I have to do workarounds
True legend!
Yeah this is expected, I just wondered about provision part.
How many do you have? As once you have a lot of VPS, may be cheaper to get a small number of cheap dedicated servers instead and create VMs on the them?
see:
Monitoring publicly: https://dash.zgato.dev/ (as you may see, a dedicated server is not a solution for me)
And the rest (out of roughly 400) are yet to be updated with my new image and stuff so not rushing it
Nope.
Why/Any links to previous explanations
Edit: Just saw the above
Virtono and HostDare are using Virtualizer and HostSailor is using SolusVM. I'm not sure about the template. I just used their Debian 12 templates. I won the HostDare vps, so that one was free.
that is absolutely insane
i had some success with using reinstall scripts.
the one mentioned before works okay and there is also https://installnet.sh/ (and some updated versions can be found on github)
works fine for basic installations of debian. if you need special partitioning or users and stuff you will have to fork and modify them which can be a bit of a pain.
but if you do it right you never have to use VNC
Debian 12 v12.10 was released a few weeks ago, so providers won't have updated yet> @emgh said:
Also a risk that Providers won't also be using the same minor version
Debian 12 history:
12.1 - 22 July 2023
12.2 - 7 October 2023
12.3 - Cancelled
12.4 - 10 December 2023
12.5 - 10 February 2024
12.6 - 29 June 2024
12.7 - 31 August 2024
12.8 - 9 November 2024
12.9 - 11 January 2025
12.10 - March 15, 2025
Custom ISO means you know all your servers are the same.
It's not unusual for the provider template to be an old minor version (like 12.1), but I upgrade it after it's built.
I seem to have the same problem. I'm using Debian 12, and I used some scripts, and I also ran some programs, and then I rebooted directly, and then I couldn't connect to the server. I sent a ticket, and the technician checked and said that my network configuration disappeared!
Any idea to prevent this issue?
Just updated mine and not dare to restart now.