Mixing AMIs?

[Clint Byrum]

Excerpts from Torsten Spindler's message of Fri Apr 15 00:57:26 -0700 2011:
> Hello,
> 
> I just started playing around with ensemble, setup my environment and
> run the mysql/wordpress example. Thanks to Kapil's help on IRC it all
> worked in the end :)
> 
> For testing ensemble I'd like to write a formula for deploying a Jenkins
> instance on EC2, as I just looked into this for a sprint. However, the
> PPA for Jenkins is for Natty, not Lucid, which is used by default for
> ensemble. 
> 
> On
> http://readthedocs.org/docs/ensemble/en/latest/provider-configuration-ec2.html
> I've seen that I can change the AMI via 'param default-image-id'. Is
> there also a way to use different AMIs for different formulas? E.g. I
> want to use a solid LTS AMI for my webservers and databases, but for
> testing my development code I want to use the latest development machine
> image that I can get hold on.
> 
> A bit unrelated, but how realistic is this scenario for the future: An
> Ubuntu using company deploys it on several thousand desktops and
> laptops. They are in dear need having an infrastructure that supports
> this environment: Mirrors for the archive, a repository for their own
> packages, a feature and bug tracking system, a testing system for
> updates, a repository server for their code, a wiki for collaboration
> and a proprietary mail / document storage servers. With help of
> Landscape Dedicated Server (systems management and web frontend),
> ensemble (service management), puppet (config management), and UEC
> (infrastructure) the company can set this environment up in a matter of
> days, rather than weeks or months. Thoughts on this 'vision'?

Torsten this hits the sweet spot for how I think of ensemble, and I think
the whole dev team would agree with me there. Building up and scaling
out an infrastructure *inside* an organization should be very straight
forward, allowing companies to focus on their products, not their tools.

As was said before, this is the kind of scenario Ensemble will shine in.
Ops guys will just build up VM/IO capacity as needed, or pay a public
cloud provider until they have it in house, and deploy the existing open
source apps, tweaking the settings knobs to work for their company.