Description of problem:

With a specific set of profiles, on boot, profiles with lower priority are used over profiles with higher priority. Bonds and OVS are involved and possibly a complicated auto-connection dependency graph, but the intended activation can be achieved manually without any issues so there is no reason NM could not achieve the same automatically on boot, so the perception form the user is that priority is being ignored.

On boot, no manual intervention, this is the status:

[root@fedora system-connections]# nmcli c show
NAME UUID TYPE DEVICE
Wired connection 1 40b0b9db-c803-3e7f-abb2-39d8593e3485 ethernet eth0
ovs-if-br-ex 4cff5623-604b-4cd0-9a5d-62915df9460c ovs-interface br-ex
bond-slave-eth1 a38b0380-f8a0-4098-af0e-ece8106d4cab ethernet eth1
bond-slave-eth2 7588cca3-1cb6-4213-9b90-956aa77821c2 ethernet eth2
br-ex cf097898-974d-4ef2-a804-917a04d16090 ovs-bridge br-ex
ovs-if-phys0 1141ecb1-06e5-4e55-a224-228307d3adc6 bond bond0
ovs-port-br-ex 43575ecf-f90a-464d-8843-9098a69dc599 ovs-port br-ex
ovs-port-phys0 a117fb62-a338-4c42-b4a5-6f855cd82dbd ovs-port bond0
bond-bond0 a9732df4-4570-4f64-853f-f8b21d6c8a09 bond –
bond-slave-eth1-slave-ovs-clone 90e70007-17ac-496a-aae0-6f134e011314 ethernet –
bond-slave-eth2-slave-ovs-clone 7ca5b4cd-6d15-4693-a7df-1534782d9b3a ethernet –

bond-slave-eth1-slave-ovs-clone and bond-slave-eth2-slave-ovs-clone have higher priority than bond-slave-eth1 and bond-slave-eth2, can be manually activated without any issues and are expected to be activated after a reboot over bond-slave-eth1 and bond-slave-eth2.

Attached connections and NM trace log

Version-Release number of selected component (if applicable):
1.32.12-2.fc35

How reproducible:
Always