Zygmunt Krynicki zygmunt.krynicki at linaro.org
Wed Aug 31 15:24:59 UTC 2011

W dniu 31.08.2011 16:39, Paul Larson pisze:
> On Wed, Aug 31, 2011 at 6:57 AM, Zygmunt Krynicki<
> zygmunt.krynicki at linaro.org>  wrote:
>> Edge is just a point on the roadmap. If staging is fast enough to
>> scrap/deploy then we may just use staging all the time.
> I was actually thinking it might be nice to have staging as more of a fluid
> concept using VMs.  We could set up a semi-permanent staging VM that gets
> updated with trunk of everything every {day, week, whatever}, but also have
> other VMs available for upgrade testing, and testing individual changes.
>   That way, for instance, if you need to test a dispatcher change that
> affects a certain type of board you don't have at home, you can simply wait
> for one of those boards to become available in the lab, mark it offline,
> bring up a vm and do some testing with it.  Likewise, it could be used to
> setup and test an invasive change to the dashboard.  All that would be
> required is a few static port mappings on the firewall, and a server capable
> of handling a few KVM sessions (which I'd like to look at getting soon for
> this purpose and others).

I agree with usefulness of staging as a fast-to-production model While 
using VMs is a possibility it does not really preclude me from 
experimenting with this design.

For safety reasons a VM might be needed once we start to run stuff as 
root. Currently all edge stuff is owned by a normal user and runs as the 
web server.

One thing that worries me is KVMs lack of (can anyone correct me if I'm 
wrong) USB pass-through support. This means we may have a hard time 
reaching USB serial adapters or USB-to-device links in general. For that 
reason my testing will use headless virtualbox instances.

I have created lp:~zkrynicki/+junk/edge where this stuff is being created.


More information about the linaro-dev mailing list