Juju LXC instances are routable after reboot

[dooferlad]

This is a fix for https://bugs.launchpad.net/juju-core/+bug/1442012

This branch adds a maintenance step that is run on daemon start up for each instance that has already been provisioned. For all instance types other than LXC this does nothing.

If jujud finds an LXC instance it will run through the same network provisioning steps as it did when first creating the machine. This already checked to see if a route or IP Tables rule existed before adding them, so there is no harm in running it again, even if the network is set up correctly already. What we have found so far is that routes are not saved across reboots of host machines. We could have solved this by writing a route save and restore init task, but it is difficult to find if such a task already exists. This way we keep all Juju related network configuration inside Juju.

This has been live tested on MAAS by:
juju bootstrap -e maas --upload-tools
juju add-machine -e maas lxc:0

watch juju status # until 0/lxc/0 is in state running
juju ssh 0/lxc/0 "echo hello world"

juju ssh 0 "sudo reboot"
watch juju status # until 0/lxc/0 is in state running
juju ssh 0/lxc/0 "echo hello world"

Repeat with more containers on the same host.

(Review request: http://reviews.vapour.ws/r/1642/)

[dooferlad]

$$merge$$

[jujubot]

Status: merge request accepted. Url: http://juju-ci.vapour.ws:8080/job/github-merge-juju

[jujubot]

Build failed: Does not match ['fixes-1453805']
build url: http://juju-ci.vapour.ws:8080/job/github-merge-juju/3199

[dooferlad]

$$merge$$

[jujubot]

Status: merge request accepted. Url: http://juju-ci.vapour.ws:8080/job/github-merge-juju