Proxmox switches network interfaces off?

[silke]

I have a very nasty problem and hope you have an idea how to get going again...

A few days ago I successfully installed proxmox on a new machine, everything worked and I created a few VMs (for testing but still I would not like to loose them).
Then I noticed that Intel vPro/AMT was not yet configured. So I configured everything in the BIOS and could connect.

But now the problems started: As soon as Proxmox started the LEDs at the NICs went dark and I not no network connectivity. As a test I started the installation USB stick up to the network configuration and it still worked (got a DHCP address as initial IP). So the NIC does not seem broken. "ip a" shows the interface with the correct IP but it cannot be reached,

Since it all started after the vPro configuration I disabled it but still the same: as soon as Proxmox takes over control the network is dead.

I am not an expert with these tools but what might make things even more difficult that common tools are not installed, like ifconfig or nmcli and I cannot install them since I have no net.

Any idea what is going on and what I can try? The only idea I have left is a complete reinstall but I would like to avoid this, since I would like to rescue some settings and the VMs.

 

[news]

please see: https://forum.proxmox.com/threads/intel-vpro-amt-ip-not-reachable.153694/

 

[maxim.webster]

Any idea what is going on and what I can try?

Best guess: Network Configuration in /etc/network/interfaces is broken, ethernet device names (en0, enpXsY) do not match reality.

 

[silke]

please see: https://forum.proxmox.com/threads/intel-vpro-amt-ip-not-reachable.153694/

vPro *is* reachable -- as long as Proxmox did not boot. The cable is in the correct plug.

Best guess: Network Configuration in /etc/network/interfaces is broken, ethernet device names (en0, enpXsY) do not match reality.

I just checked -- everything as it should be vmbr0 has the correct static IP.

Since the LEDs are completely off as soon as the OS has booted, my guess is the the physical NICS themselves are somehow forced down, though I have no idea how this can happen -- and how to fix it.

 

[silke]

Well, I accepted the inevitable and reinstalled Proxmox. Now the network is active again with both vPro and the normal networking. Expensive experience but luckily at an early state. As soon as it goes productive I will have backups for everything -- I hope ;-)

Still if anyone has an idea what might have happened I would be interested in hearing. It is always good to know how to handle such failures.

 

[silke]

Sorry to bring this up again but it happened again but this time I don't think it has to do with vPro.
All I did was adding two additional Nvme disks. Again the LEDs at the RJ45 ports are off and no network.

After some digging around I found the reason and yes it was /etc/network/interfaces. There the name of the active interface was changed to a wrong name. Instead of enp90s0 there suddenly was enp89s0 in the file. How can this happen? I am sure I never changed this file. Of course I will make a backup but can the system be protected against this sort of surprise?

After I changed it back und restarted the network everything started working again but still not a nice newbee experience :-(

 

[maxim.webster]

After I changed it back und restarted the network everything started working again but still not a nice newbee experience :-(

It is related to the way the kernel enumerates network devices and adding PCIe devices (incl. NVMe) affects it.

You can change the default behaviour, see https://www.freedesktop.org/software/systemd/man/latest/systemd.net-naming-scheme.html

 

[Neobin]

After some digging around I found the reason and yes it was /etc/network/interfaces. There the name of the active interface was changed to a wrong name. Instead of enp90s0 there suddenly was enp89s0 in the file. How can this happen? I am sure I never changed this file.

Not the config file content changed, but the actual network interface name(s).

Of course I will make a backup but can the system be protected against this sort of surprise?

https://pve.proxmox.com/pve-docs/chapter-sysadmin.html#network_override_device_names

 

[silke]

Not the config file content changed, but the actual network interface name(s).

Sure? I have three identical machines and the other two both have enp90s0 both as interface name and in /etc/network/interfaces.
The "problem machine" also has enp90s0 as interface name but had enp89s0 in /etc/network/interfaces.

For my uninformed mind it looks like only the interfaces file was wrong.
The whole thing looks very strange to me. Why not keep it like in the old days when there was just eth0, eth1 and so on? Just a name for every network interface and that's it (plus a few for loopback, wifi and such). I think Rapian OS still has it this way (also Debian based).