Repair bridge and/or iptables rules without shutdown/restart

[bboreham]

From time to time we observe that bridge set up seems to have half-completed, so the initialization won't get run again on restart but it isn't properly set up.

Especially in a Kubernetes context this is hard to repair, and a reboot is generally easiest.

So this issue is to have some way to re-run the initialisation when needed.

[RRAlex]

I'm deploying on a small cluster (6 nodes: 3 small controllers, 3 workers). If I reboot the machines, it seems there is no more network connection to anywhere, but if I delete everything, reboot again and redeploy, it works again.

Is this related?
(k8s: 1.7.4, weave 2.0.4, etcd. 3.2.6, docker 17.06.1, ...)

[bboreham]

@RRAlex that's pretty much the opposite of this issue. A reboot forces the initialization to be re-run.

Could you please open a separate issue and include the logs from the weave container in the working and non-working cases.

[brb]

The "machine reboot required" part got fixed by #3204.