[SOLVED] Proxmox 6 to 7 upgrade : network interfaces names changed

[nathanaelh]

Hello, I upgraded successfully an OVH-soyoustart server. I just wanted to tell you that the interfaces names changed from eno3 & eno4 to eth0 & eth1. So after the reboot I had no network ; I gained access to the IPMI console (network KVM) and noticed that change. So I upgraded my /etc/network/interfaces file and it worked.

By the way I confirm that the Mac Address of the Bridge changed, as explained in the migration guide.

I do not know it this is specific to my server, or to Proxmox or to Debian 10 to 11, but as I was not aware of that, I thought this could be usefull to other.

Thanks for the upgrade.

Nathanaël

 

[Moayad]

Thank you for sharing your experience

I do not know it this is specific to my server, or to Proxmox or to Debian 10 to 11

This is from Debian 11, the below link has more details.

https://sources.debian.org/src/bridge-utils/1.7-1/debian/NEWS/#L3-L23

 

[nathanaelh]

Thanks for the link.

Do you know why the interfaces (physical, not bridge) names changed ?

 

[Moayad]

Do you know why the interfaces (physical, not bridge) names changed ?

read the below link
https://www.freedesktop.org/wiki/Software/systemd/PredictableNetworkInterfaceNames/

 

[hellfire]

Just want to mention that I'm just upgrading from pve6 to pve7 and the predictive network interface names changed as well. Probably some Naming convention change...

ens3f0 -> ens24s0f0
ens3f1 -> ens24s0f1
ens3f2 -> ens24s0f2
ens3f3 -> ens24s0f3
ens4f0 -> enp216s0f0
ens4f1 -> enp216s0f1

luckily eno0 the management network device was unchanged, so no need for using ipmi.

(I realized that I had active the non-enterprise repo active, even if the cluster has a subscription. But since it was active I did not want to change that now. If I think about it, it would have been better).

Maybe it's good to remember the mac addresses of the network devices. Just in case one is not able to determine what the new names are linked to the old ones, with a line like this:

for dev in /sys/class/net/*;do echo $(basename $dev) $(cat $dev/address);done >/root/macs.txt

 

[Stronginthearm]

I just encountered this problem on one of my cluster nodes and wanted to share the solution:

With some hardware apparently you have to set the naming policy for systemd in the /etc/systemd/network/99-default.link file:
[Link] NamePolicy=path

I already had that, but with the new version systemd seems to also explicitly wants to have a [Match]-section, so this is the 99-default.link that fixed the problem for me:

[Match]
OriginalName=*

[Link]
NamePolicy=path

 

[treki]

Hello, I upgraded successfully an OVH-soyoustart server. I just wanted to tell you that the interfaces names changed from eno3 & eno4 to eth0 & eth1. So after the reboot I had no network ; I gained access to the IPMI console (network KVM) and noticed that change. So I upgraded my /etc/network/interfaces file and it worked.

By the way I confirm that the Mac Address of the Bridge changed, as explained in the migration guide.

I do not know it this is specific to my server, or to Proxmox or to Debian 10 to 11, but as I was not aware of that, I thought this could be usefull to other.

Thanks for the upgrade.

Nathanaël

It was exactly the same for me on the OVH (stor) server. Many Thanks