Juju-Vagrant boxes: Flaky API connections shortly after booting

[Abel Deuring]

Recent versions of the Juju-Vagrant images (see
https://juju.ubuntu.com/docs/config-vagrant.html ) do not work after a
reboot because a script that is invoked to configure NAT has
difficulties to access the state server; the web server for
http://127.0.0.1:6080 (mention in the web page above), who also wants to
connect to the state server, fails quite often too.

Comparing MongoDB logs, Juju logs and logs from the script, it turns out
that the errors occur shortly after the mongo server is restarted.

More precisely, the mongo server is always started _three_ times when
the machine boots. Example log output:

machine-0: 2014-07-10 13:00:00 INFO juju.mongo mongo.go:147 Ensuring
mongo server is running; data directory /home/vagrant/.juju/local; port
37017
machine-0: 2014-07-10 13:00:06 INFO juju.mongo mongo.go:147 Ensuring
mongo server is running; data directory /home/vagrant/.juju/local; port
37017
machine-0: 2014-07-10 13:00:13 INFO juju.mongo mongo.go:147 Ensuring
mongo server is running; data directory /home/vagrant/.juju/local; port
37017

I have a workaround for the Vargant boxes nearly ready for this problem
(patiently try to query the server), but I'm wondering if it is
necessary to restart the server two times -- and I'm wondering if only
the Vagrant images are affected by this problem.

cheers
Abel