On Mon, 2018-03-26 at 14:08 +0200, Thomas Haller wrote:
In the log, the first activation attempt fails,
Why does it fail? It should be statically assigned IPv4.
that cause NetworkManager to tear down the interface, and remove all IP addresses (and routes). It fails, because no IP address is received within timeout.
Do you mean an IPv6 activation attempt fails so it tears down everything including IPv4 and IPv6? So this is just more fallout of the router incorrectly advertising DHCP6 configuration?
If you don't want NM to manage this device, mark it as unmanaged (or just don't create a connection that is configured to autoconnect).
It's not so much that I don't want NM to manage the device. I just want it to know that routing will be handled by a routing daemon and protocol. What is interesting is that when this all happens, it's only the default route that is not restored by zebra. All other routes installed by zebra are back up once NM succeeds in configuring the interface.
Most likely, you just want to delete the connection profile for that interface, `nmcli connection delete "$PROFILE"`.
Meaning not have NM manage it and let RH's native networking scripts handle configuring it?
You could also configure the route in NetworkManager (by setting ipv6.gateway).
It is the IPv4 default route that is getting removed. But even if I were to configure that route using ipv4.gateway, that defeats the purpose of using a dynamic routing protocol, which is what is used here. Cheers, b.
Attachment:
signature.asc
Description: This is a digitally signed message part