canonical-ci-engineering team mailing list archive

[Andy Doan <andy.doan@xxxxxxxxxxxxx>]

On 10/03/2013 06:57 AM, Evan Dandrea wrote:
> On 18 September 2013 23:22, Andy Doan <andy.doan@xxxxxxxxxxxxx> wrote:
> > this works (tangent question: how do i get this to use an m1.medium?).
>
> --constraints "mem=3840" should do it.
>
> > This *seems* to work - I see a new process running on my instance like:
> >
> >   python2 -c import sys; skip_imports=1; verbosity=0; exec
> > compile(sys.stdin.read(764), "assembler.py", "exec")
>
> What is this process? From the "error details" below, you appear to be
> correctly running gojuju and not pyjuju, so I'm assuming it's not the
> latter.
>
> > Then I run "juju status" and get:
> >
> > $ juju status -e openstack
> > error: Unable to connect to environment "openstack".
> > Please check your credentials or use 'juju bootstrap' to create a new
> > environment.
> >
> > Error details:
> > cannot log in to admin database
>
> Is there anything of note in /var/log/juju on the bootstrap node?
>
> Happy to walk through this with you on IRC if you're still having issues.

Things are working, but I'll note the 2 issues (as I remember them to be):

1) deployments of precise weren't working.
I noticed the primary issue was that cloud-init was failing. This meant 
the db was installed, but not populated. On a whim, I tried on raring 
and things "just worked". A few days after that precise started working 
again.

2) public IP and juju
juju deploy won't work correctly if you have assigned your node a public 
IP (and thus don't need sshuttle). For some reason the agent.conf files 
for each juju agent were created with incorrect listening addresses. I 
could manually tweak the files and get things to work. However, I just 
gave up and configure the public IP address as the last step in my process.