Network interfaces renamed after upgrade to 2079.3.0 #2578
Container Linux Version
What hardware/cloud provider/hypervisor is being used to run Container Linux?
Bare Metal K8s with Cillium for networking and Rook for storage
No changes per the release notes found here https://coreos.com/releases/#2079.3.0
network interface ids changed and broke things
The text was updated successfully, but these errors were encountered:
Those release notes are just saying that there were no changes between 2079.2.0 and 2079.3.0. We'll use clearer wording in the future. For the full set of changes between 2023.5.0 and 2079.3.0, see the releases page. In this case, the systemd 241 update was probably the culprit.
In order to fix the regression, we'll need more information from you. What were the interface names before and after the upgrade? Could you post the output of
I'll retitle this bug for clarity.
This is the requested output for a compute node, where the renaming of the interfaces atleast results in a working node. For more complex nodes like network nodes, the BGP configuration, sysctl and iptables settings completely fall apart. All nodes have Mellanox NICs and in the case of the compute nodes the name was always
@gigatexal @trevex thanks for the follow up.
The network card(s) at
While we track down what's going on, you may try experimenting and getting back the old behavior via https://www.freedesktop.org/software/systemd/man/systemd-udevd.service.html#net.naming-scheme=.
I haven't gotten around setting
This would indeed lead me to believe that systemd is the "culprit".