Network interface down after physical change

[fabiolanza]

Hi,

I turned off my server to install a new NVMe M.2 stick. I turned it on and after Proxmox finishes booting the network interface is down. I tried to bring it up with "ifup -a --force" but it did not work. Do you know why this might have happened?

My entire server is down and I am quite in a tough situation. Here are some photos:

Thanks,

Fabio

 

[fabiolanza]

I fixed it, I just had to run the command "ip link set dev <interface> up" and then rename the interface in the config file to the interface name. I am not sure but the name changed with the server change. Thanks.

 

[Moayad]

Hi,

I am not sure but the name changed with the server change. Thanks.

usually the systemd change name interface, you can see more details here[0].
But you can fix that if you tell the systemd to put a static name for your interface

[0] https://www.freedesktop.org/wiki/Software/systemd/PredictableNetworkInterfaceNames/

I hope this helps

 

[spirit]

the problem is that nvme drive are pci device, and nic new scheme also use pci slot in naming (enp37 -> this is slot37)

and some motherboard change the number when you add/remove nvme drive.

 

[bobiii84]

Just to help anyone else with this frustration, I recently removed my nvme, to replace the bands holding my heatsync on. I had no idea this issue could happen. Mine is off in a closet somewhere so I had to bring it over to connect. No internet. The NIC was in a DOWN state.

Eventually I discovered that the interface I was seeing was the "new" interface it assigned, and it wrote over my interfaces file in /etc/network/. The interface portion was in there, but nothing about the bridge or IP or gateway, I had to add it all in. Here is an example of what was in mine for your reference. Sorry for anyone dealing with this, took me two hours and help from a friend to solve :/

Just remember to put in YOUR internal IP, subnet mask and gateway. If you are unsure of your settings, you can always change it from "manual" to "static" give the interface the IP, subnet mask and gateway just to get into your web interface and look. Good luck!

 

[leesteken]

It is a small disaster that PCIe IDs shift when adding devices, which causes the modern way of named the network devices to shift also. Especially when (also) using PCI passthrough and VMs that start automatically on boot, which tend to freeze or restart the host when the wrong devices are chosen.
IMHO, to get out of such a crash-loop is to temporarily remove the amd_iommu or intel_iommu kernel parameters using the GRUB or systemd-boot edit option when starting the machine. Maybe this will help someone sometime when unsuspectingly removing or adding a device to their machine.
At least that way, you can boot Proxmox and change /etc/network/interfaces from the console and fix the /etc/pve/qemu-server/*.conf hostpci options.

 

[i_am_jam]

This helped me out after losing the webUI following adding a new PCI NIC to a server today.
Manually updated /etc/network/interfaces with the newly assigned PCIe IDs for the existing configured interfaces and restarted.

 

[k.jings]

Don't know when or why it happened - thanks to this thread, I was able to take care of several VMs which went offline that rely on the /etc/network/interfaces network definition method. Interestingly, a VM that uses systemd-networkd (with yaml config files under /etc/netplan) continued to work fine even though the interface names did not match. Another VM using NetworkManager (config files in /etc/NetworkManager/system-connections) - also had mismatched interface names yet continued to work fine. Just chiming in.

 

[leesteken]

Don't know when or why it happened - thanks to this thread, I was able to take care of several VMs which went offline that rely on the /etc/network/interfaces network definition method. Interestingly, a VM that uses systemd-networkd (with yaml config files under /etc/netplan) continued to work fine even though the interface names did not match. Another VM using NetworkManager (config files in /etc/NetworkManager/system-connections) - also had mismatched interface names yet continued to work fine. Just chiming in.

Systems that use a static configuration file (/etc/network/interface) cannot handle names that change. Systems that use a dynamic configuration and actively look for network devices and don't use fixed names (NetworkManager) can deal with changing names. I'm not surprised.

 

[Joe Piler]

It is a small disaster that PCIe IDs shift when adding devices, which causes the modern way of named the network devices to shift also. Especially when (also) using PCI passthrough and VMs that start automatically on boot, which tend to freeze or restart the host when the wrong devices are chosen.
IMHO, to get out of such a crash-loop is to temporarily remove the amd_iommu or intel_iommu kernel parameters using the GRUB or systemd-boot edit option when starting the machine. Maybe this will help someone sometime when unsuspectingly removing or adding a device to their machine.
At least that way, you can boot Proxmox and change /etc/network/interfaces from the console and fix the /etc/pve/qemu-server/*.conf hostpci options.

Facing a similar issue I think...

Before i had the onboard NIC and 1x PCI nic

Onboard : enp5s0 : vmbr0

pci nic : enp4s0 (not in use)

Yesterday morning I added one more pci nic, now none of the nics are up ip -a hows:

enp6s0: down
enp5s0: down
enp4s0: down
vmbr0: up (but not functioning due to the fact enp5s0 is down)

ifup and variations thereof do nothing.

(The idea is to use the 2 pci nics passed through to a vm running PFsense and use the onboard nic for vm traffic and proxmox)

 

[pinomat]

Hello,

Same issue here. I installed a new GPU in my Intel NUC and lost my Network Card... It took me 30 min to figure out that the name of my network card changed.
The interface was present but was DOWN.

I changed the name according to the name reported in ip link show in /etc/network/interfaces then reboot and all was back to the normal.

Have a good day

 

[r100gs]

Hello, I have the same problem.

It does not connect. It's a fresh install on an alder lake system.
Installing an other life Linux works fine.

Maybe sb can help me.

Best regards,
Stefan

 

[leesteken]

Hello, I have the same problem.

It does not connect. It's a fresh install on an alder lake system.
Installing an other life Linux works fine.

Maybe sb can help me.

Best regards,
Stefan

enp1s0 says NO-CARRIER, which usually means that there is no signal on the wire (which is different from the original issue of this thread). Check or replace the cable, make sure it is plugged in properly on both ends. Does the other side (router?) see a working cable? Check if the network controller reported some errors in journalctl -b 0 (use the arrow keys).

 

[tjsyl]

Just to help anyone else with this frustration, I recently removed my nvme, to replace the bands holding my heatsync on. I had no idea this issue could happen. Mine is off in a closet somewhere so I had to bring it over to connect. No internet. The NIC was in a DOWN state.

Eventually I discovered that the interface I was seeing was the "new" interface it assigned, and it wrote over my interfaces file in /etc/network/. The interface portion was in there, but nothing about the bridge or IP or gateway, I had to add it all in. Here is an example of what was in mine for your reference. Sorry for anyone dealing with this, took me two hours and help from a friend to solve :/

Just remember to put in YOUR internal IP, subnet mask and gateway. If you are unsure of your settings, you can always change it from "manual" to "static" give the interface the IP, subnet mask and gateway just to get into your web interface and look. Good luck!

You are a life saver. Thank you.

 

[ivanromeromendoza]

New Hardware, same system Disk?

command "ip address", take note of new interface like ens0xxx and change it en /etc/network/interface "bridge-ports" and Voila!