Adding 2.5" SATA SSD via PCIe adapter kills ethernet connection

[filou]

I am trying to add a 2.5" SSD to my system using the Accelsior S PCIe adapter card.

When I add the card to the system and boot, I can see the disk drive in fdisk as well as format, add/remove partitions etc.

But then my ethernet connection is broken:

Jan 13 14:26:08 pve-1 networking[887]: error: vmbr0: bridge port enp111s0 does not exist
Jan 13 14:26:08 pve-1 /usr/sbin/ifup[887]: error: vmbr0: bridge port enp111s0 does not exist
Jan 13 14:26:08 pve-1 networking[887]: warning: vmbr0: apply bridge ports settings: bridge configuration failed (missing ports)
Jan 13 14:26:08 pve-1 /usr/sbin/ifup[887]: warning: vmbr0: apply bridge ports settings: bridge configuration failed (missing ports)
Jan 13 14:26:08 pve-1 networking[887]: error: >>> Full logs available in: /var/log/ifupdown2/network_config_ifupdown2_112_Jan-13-2024_14:26:08.612247 <<<
Jan 13 14:26:08 pve-1 /usr/sbin/ifup[887]: >>> Full logs available in: /var/log/ifupdown2/network_config_ifupdown2_112_Jan-13-2024_14:26:08.612247 <<<

Once the card is removed, the ethernet connection works again and vmbr0 is able to bridge to enp111s0 again.

Jan 13 14:43:40 pve-1 kernel: igb 0000:6f:00.0 enp111s0: igb: enp111s0 NIC Link is Up 1000 Mbps Full Duplex, Flow Control: RX/TX
Jan 13 14:43:40 pve-1 kernel: vmbr0: port 1(enp111s0) entered blocking state
Jan 13 14:43:40 pve-1 kernel: vmbr0: port 1(enp111s0) entered forwarding state

What could be causing this behavior? Any ideas what do I need to change or troubleshoot in order to get ethernet working again when the PCIe card is in use?

Thanks in advance

kernel: Linux version 6.5.11-7-pve

 

[ness1602]

Probably the MB is changing etherner name once you add pcie card. Check the dmesg for card name after adding it.

 

[leesteken]

Very common issue, once you recognize that the PCI ID changes because a device was added (or removed) and the network device name is based on the PCI ID.
Use the Proxmox host console and run ip a to find the new name and edit /etc/network/interfaces accordingly.

 

[filou]

Probably the MB is changing etherner name once you add pcie card. Check the dmesg for card name after adding it.

Thank you, this was exactly the issue.

Use the Proxmox host console and run ip a to find the new name and edit /etc/network/interfaces accordingly.

The problem was, I was not able to access the console since the host was no longer available on the network! I had to bring in a monitor and keyboard to connect directly, but was then able to change /etc/network/interfaces to get everything working again.

This seems fairly important that any added (or removed) PCI device will potentially take down your entire host!

Is there any way to define a specifc name for the ethernet connection, one that would not change – in order to avoid this problem in the future?

Thanks both for your help pointing me in the right direction!

 

[leesteken]

This seems fairly important that any added (or removed) PCI device will potentially take down your entire host!

Is there any way to define a specifc name for the ethernet connection, one that would not change – in order to avoid this problem in the future?

Yes, add some udev-rules to use the MAC instead. This is not an uncommon issue (on other Linux systems as well as Proxmox) and you might find work-arounds on this forum (now that the problem is clear).