What's your juju on LXC experience been like?

Jonathan Lange jml at canonical.com
Mon Jun 18 15:58:52 UTC 2012

On 18 June 2012 16:42, Jorge O. Castro <jorge at ubuntu.com> wrote:
> At UDS we discussed that we need to redo the bits of documentation
> that mention LXC. More specifically, we need to tell people what to
> expect, like "oh yeah, the very first time you run this it'll download
> 300 megs of stuff, so that will be slow." And then there's making sure
> you're in the right group if you hadn't used it before, etc. I plan on
> just putting this all in the documentation so people can manage their
> expectations before they begin.

That'd be great. Those two things bit me hard when I first started. Is
it possible to automate the group stuff?

> So what I would like to do is collect your experiences with LXC,
> specifically any "gotchas" that you think might catch new LXC users,
> so if you have any tips or workarounds for LXC specific workflows
> you're doing, let me know!

Other things:
 * 'juju debug-log' isn't what you want, you actually want 'tail -f
/tmp/juju-local/jml-local/machine-agent.log'. Well, that's what I
want. I couldn't possibly speak for other user names.
 * it's just not obvious when things happen or when things are done.
Keep bouncing on 'juju status'.
 * services that are running don't necessarily appear in netstat (I
just ran 'juju deploy python-moinmoin' and wondered why I couldn't
browse to the IP – I didn't notice the port number in 'juju status'
and looked at 'netstat' instead on the assumption that this was
another slow-to-start, not-announced-when-its-done thing.)

Will keep you up-to-date. I'm charming up a new, not-yet-deployed
internal microservice (lp:libdep-service) that functions as one of the
backends for developer.ubuntu.com's auto packaging.


More information about the Juju mailing list