← Back to team overview

cf-charmers team mailing list archive

Re: Dealing with Warden-within-LXC issue.

 

On Mon, Apr 14, 2014 at 4:42 PM, Alexander Lomov <lomov.as@xxxxxxxxx> wrote:

> Hey, all.
>
> Since we need to implement deployment DEA to local environment and I'm not
> sure that I see problem clearly, I'd like to start discussion of this
> topic.
>
> 1. First question for me is:
> Is it worth to try to deploy DEA into juju kvm local environment? I mean
> something like approach described in this article:
> http://astokes.org/juju-deploy-to-lxc-and-kvm-in-the-local-provider/.
> Deploying to kvm can solve this issue.
>

yes, that article describes the technique to use for containers for most
things, and kvm just for the dea, that's a good workflow and can be
captured into a deployer config as well  (to: kvm:0)

make sure to install uvtool package for the kvm support.


> 2. Strategy to solve LXC/Warden problem:
> We can perform following steps in order to solve this issue:
>         a. create lxc container
>         b. deploy Warden to it
>         c .try to create Warden container using warden shell
>         d. after that we will change configuration step by step depending
> on errors we will get.
>
> It will require us to change default LXC settings, so we will need to
> provide user of juju charms with list of changes to LXC settings.
> Is it OK for now to make user change his defaults?
> What do you think about strategy for solving this task?
> What are your suggestions on this point?
>
>
for now go with #1 (it works and can be captured to bundle), when local
provider has app armor profiles configurable for lxc we can consider
debugging #2.

cheers,

Kapil

Follow ups

References