Working on a new OpenStack deployment we noticed some weird behavior with our nodes. It began to show up as soon as we had it all deployed. Some of the nodes where unable to communicate with MAAS and to some of the other nodes in the cluster.
All the configuration seemed OK, but then we noticed everything started working as soon as we disabled one of the ports in the bond. Looking at the switches we saw that the node port pair in a bond where both publishing the same MAC address at the same time. The expected behavior is for the active port to publish the MAC address.
We compared the configuration of this new environment and a older one. The configuration in Netplan is identical. We are using the same hardware (Dell PE R440).
After some testing we were able to find a workaround by manually adding the parameter "primary: " to the netplan config.
The Kernel documentation references this option. https://www.kernel.org/doc/Documentation/networking/bonding.txt
(Search for fail_over_mac)
As this is a manual process, it is not a good solution.
I was wondering if there is a way for MAAS to add either the “primary” or “fail_over_mac” parameter during the deployment?
MAAS Version: Snap 2.9.1 (9153-g.66318f531)