Skip to content

Issues with Bond Interface Failover on Bare Metal Ubuntu EKS-A 22.04 Using systemd-networkd

0

Hi All,

I am working with Bare Metal Ubuntu EKS-A 22.04 using systemd-networkd with Netplan for network configuration. I’ve set up a bond interface through the /etc/systemd/network files, and the settings in /proc/net/bonding/bond0 looks correct.

However, after configuring the bond interface, I lose connectivity. A restart of systemd-networkd doesn't resolve the issue; only a reboot restores the connection. Additionally, when I bring the physical or logical link down (either by disabling the interface or using ip link), I lose connectivity again and need to reboot the server.

Interestingly, when I use NetworkManager for configuring the bond interface, everything works as expected — failover and link status changes happen without issues.

Has anyone else experienced similar problems with bonding when using systemd-networkd in EKS-A? Is there any special configuration required for bonding in EKS-A, or is this a known issue with systemd-networkd?

In regular Ubuntu, NetworkManager is the default, but in EKS-A, systemd-networkd is used by default. Any insights or suggestions would be greatly appreciated.

1 Answer
0
Accepted Answer

The issue you're experiencing with bond interface failover on Ubuntu EKS-A 22.04 using systemd-networkd might be related to how network interfaces are named and managed in systemd-based systems.

In systems using systemd or udev versions 197 or greater (which includes Ubuntu 22.04), predictable network interface names are used by default. This can cause problems with network connectivity, especially when configuring bonded interfaces, as the system may not consistently identify network interfaces as expected.

You might want to try disabling predictable network interface names by adding the net.ifnames=0 option to your GRUB configuration. This can be done by modifying the GRUB_CMDLINE_LINUX line in /etc/default/grub and then rebuilding the grub configuration file.

Another potential issue could be related to security groups if you're using AWS infrastructure. When multiple network interfaces are attached to an instance, there have been cases where additional ENIs (Elastic Network Interfaces) are assigned incorrect security groups, which can cause connectivity problems.

Since NetworkManager works correctly for your bond configuration while systemd-networkd doesn't, there might be specific configuration parameters or settings in systemd-networkd that need adjustment for proper bond interface failover. Check your bond configuration parameters like bond mode, primary interface settings, and failover policies in your systemd-networkd configuration files.

Additionally, ensure that your routing configuration is properly handled for multiple network interfaces, as this is a common issue when working with bonded interfaces in Ubuntu.
Sources
AWS instance unusable after attaching Network Interface | AWS re:Post
Enable enhanced networking with ENA on your EC2 instances - Amazon Elastic Compute Cloud

answered a year ago

AWS
EXPERT

reviewed a year ago

  • All routes correct. I am not able to ping to bond interface IP. Not related to routes.

You are not logged in. Log in to post an answer.

A good answer clearly answers the question and provides constructive feedback and encourages professional growth in the question asker.