NetworkManager cannot manage macvlan devices inside container

Under the following conditions:

• Using container instances (e.g., LXC)
• macvlan networking is defined for a particular network interface (e.g., eth0 defined as type macvlan with parent interface eno0)
• if NetworkManager is running

Then:

Attempting to bring up the interface will fail with the following error:

[root@centos-8-stream-cloud ~]# ifup eth0
Error: Connection activation failed: No suitable device found for this connection (device lo not available because device is strictly unmanaged).

Note: The requested interface is eth0 not lo. The error is for lo. The interface lo is up.

[root@centos-8-stream-cloud ~]# ip addr show lo
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN group default qlen 1000
    link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
    inet 127.0.0.1/8 scope host lo
       valid_lft forever preferred_lft forever
    inet6 ::1/128 scope host 
       valid_lft forever preferred_lft forever

Workaround:

The following workarounds can be used if you don't require network access to manage the instance and only if network-scripts is installed.

• Disable NetworkManager. Then ifup eth0 brings up the interface with no error.
• Add NM_CONTROLLED=no to /etc/sysconfig/network-scripts/ifcfg-eth0. Note that this is the default for cloud-init, but RedHat has patched this out of their version. Also adding NM_CONTROLLED=no to ifcfg-lo does not alter the behavior or error message.

However, cloud instances are often only available over the network. Where networking is expected to work and the primary access method is via ssh, these instances are unusable and unfixable. This presents a major problem.

Desired outcome:

Given the fact that the interface comes up just fine without NetworkManager, NetworkManager needs to get over its hangup regarding macvlan interfaces. The interface works fine. It's just NetworkManager's inability to recognize this and do the right thing.