networkd tears down bridge ip address when the last device is pulled out from the bridge

[Vladimir Pouzanov]

In Ubuntu 18.04.1 LTS (systemd 237) pulling out the last device out of the bridge removes the bridge ip address.

given the following config:

# cat /etc/systemd/network/vmbr0.netdev
[NetDev]
Name=vmbr0
Kind=bridge

[Bridge]
HelloTimeSec=0
ForwardDelaySec=0
STP=no

# cat /etc/systemd/network/vmbr0.network
[Match]
Name=vmbr0

[Network]
Address=10.10.0.1/16
ConfigureWithoutCarrier=yes
DHCP=no
IPForward=yes
IPv6AcceptRA=no
LinkLocalAddressing=no

networkd would bring up vmbr0 with 10.10.0.1/16 on system boot despite it not having any devices, allowing services to bind to 10.10.0.1.

However, if you add a device and then remove it (e.g. by starting and then stopping a virtual machine connected to the bridge), networkd would tear down the interface:

3: vmbr0: <NO-CARRIER,BROADCAST,MULTICAST,UP> mtu 1500 qdisc noqueue state DOWN group default qlen 1000
    link/ether 16:19:4c:7f:e8:c4 brd ff:ff:ff:ff:ff:ff

thus making any service listening on 10.10.0.1 inaccessible.

previously on Xenial the bridge would stay intact.

[Vladimir Pouzanov (farcaller)]

Seems to be related to https://github.com/systemd/systemd/pull/7403 which isn't merged still.

[Launchpad Janitor (janitor)]

Status changed to 'Confirmed' because the bug affects multiple users.

[Scati Labs I+D (scatilabs)]

This issue has been fixed upstream: https://github.com/systemd/systemd/pull/10597

[Dan Streetman (ddstreet)]

This appears to be fixed in the latest version of systemd for Bionic and later.