juju bootstrap silently fails with invalid origin, creates impression of working setup that fails in unpredictable ways

[Peter Petrakis]

In following the getting started guide:
https://juju.ubuntu.com/docs/getting-started.html

I decided to take a slightly different path seeing that I was using an
existing precise-server, and juju was already in the repository, I didn't
add the ppa and continued with the instructions.

When I got to ~/.juju/environments.yaml I saw the "ppa" line and
changed it to universe, as that's where juju lives according to apt-cache.

default: sample
environments:
  sample:
    type: local
    control-bucket: juju-a14dfae3830142d9ac23c499395c2785999
    admin-secret: 6608267bbd6b447b8c90934167b2a294999
    default-series: precise
    juju-origin: universe
    data-dir: /home/ppetraki/Sandbox/juju-local

# vim:ts=2:sw=2:et:

juju bootstrap completed with out error. In fact all the commands prescribed
to create the wordpress instance completed without error.

Juju status reported a "pending" state for my instances with no public ip
addresses. I was quite puzzled.

I sought help on #juju and found it. After destroying the environment
and recreating the steps with the debug log running we discovered a
fatal error.

Reading package lists...
+ setup_networking
+ apt-get install -o Dpkg::Options::= --force-confnew --force-yes -fuy resolvconf
Reading package lists...
Building dependency tree...
Reading state information...
resolvconf is already the newest version.
0 upgraded, 0 newly installed, 0 to remove and 2 not upgraded.
+ cat
/tmp/juju-create8LcGCo: line 28: /etc/resolvconf/run/resolv.conf: No such file or directory
+ cat
+ cat
+ setup_juju
+ '[' '!' -e /var/lib/juju ']'
+ mkdir /var/lib/juju
+ '[' '!' -e /var/log/juju ']'
+ mkdir /var/log/juju
+ '[' universe = ppa ']'
+ '[' universe = proposed ']'
+ [[ universe = lp:* ]]
+ '[' universe = distro ']'
+ echo 'Unknown juju origin policy universe: expected [ppa|branch|disto]'
Unknown juju origin policy universe: expected [ppa|branch|disto]
+ exit 1

2012-09-17 11:05:29,684 Machine:0: juju.agents.machine DEBUG: Units changed old:set(['wordpress-db/0']) new:set(['wordpress/0', 'wordpress-db/0'])
2012-09-17 11:05:29,684 Machine:0: juju.agents.machine DEBUG: Starting service unit: wordpress/0 ...
2012-09-17 11:05:29,760 Machine:0: unit.deploy DEBUG: Downloading charm cs:precise/wordpress-7 to /home/ppetraki/Sandbox/juju-local/ppetraki-sample/charms
...

If it wasn't for the fact that the error logs showed the enum set for valid
origin values I wouldn't have known what to set it to, without resorting to google etc.

<jimbaker> ppetraki, juju bootstrap should have immediately rejected a bad juju-origin. it didn't. instead you saw the error in the log
<ppetraki> jimbaker, ok, so I'll file a bug on that
<jimbaker> i just tried with ec2
 and same thing

Also, in retrospect, I noticed that there is no annotated default environments.yaml file provided
with juju. It's customary with UNIX to provide an annotated config file or at least a manpage describing
the config file to avoid things like this from even happening to begin with and provide ready reference to
solve said issue without leaving the console.