LOWERLAYERDOWN for docker0 bridge since upgrading to CoreOS766.3.0

[hookenz]

I've got some odd issues with container networking. Some services were not responding all of a sudden and I've isolated it to this one machine that is showing the LOWERLAYERDOWN error.

docker0@NONE: <NO-CARRIER,BROADCAST,MULTICAST,UP> mtu 1500 qdisc noqueue state LOWERLAYERDOWN group default

Yet all networks are all up.

Some googling suggests that recent versions of systemd-network ignore the global ipforward flag and that you have to set it per interface.

moby/moby#14185

However, mine are all in the forward state and UP. I set the IPForward=kernel against the primary network interface and rebooted and it is still saying "LOWERLAYERDOWN".

Some further investigation and several reboots and despite it saying LOWERLAYERDOWN the containers are able to ping other networks.. confused! is this LOWERLAYERDOWN message incorrect and should it be ignored?

[hookenz]

Just for reference. Another node running 723.3.0 shows

docker0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP group default

[hookenz]

Oh... might be a hardware problem... leaving open but investigating further... another node running 766 is not showing LOWERLAYERDOWN and is working correctly... but this one node isn't... BTW networking is appearing to function on it, because all the containers that have been scheduled to run on it just happen to use --net=host.

[hookenz]

False alarm. I've just discovered that this behaviour is normal if there are no containers attached to the bridge. I just fired up another container that doesn't use --net=host and then the bridge state changed on the host. So nevermind. It's all good.