Test servers
Jeroen van Meeuwen
kanarip at kanarip.com
Fri Aug 22 22:55:17 UTC 2008
Mike McGrath wrote:
> So we're in kind of an awkward but probably beneficial spot right now for
> our test server infrastructure.
>
> Its _ALL_ down right now. And we need to build it back up. We've talked
> about another layer of test servers and I think its probably time for that
> so here's what I propose.
>
> Rebuild our test servers external to PHX. There is an open ticket for
> this already, we're just going to be doing it finally.
>
> Create a staging environment in the publictest and test areas in PHX.
>
> Staging will be identical to our production environment where possible
> with the exception of production data and where new features are being
> implemented and are not in production yet.
>
> Staging will be in puppet. All of it. Just like production is and
> they'll get rebuilt regularly. We can use puppet's environments to do
> this but I'm going to talk with kanarip about implementation and such. We
> actually have a very basic testing environment setup in puppet now but its
> not really tested or documented yet. Staging will be different.
>
So, a few things on staging environments:
Puppet can stage the following things:
- the manifest to use (e.g. site.pp) for an environment, including all
imported files (e.g. classes/*.pp etc)
- the modulepath to use
This means that you can split into environments what you use from the
modules in those environments, but not what is in configs/ right now.
For what is in configs/ to become staged, it needs to go into modules.
Preferably, we would stage using git branches (development -> testing ->
production, for example).
A simple plan without too many details:
- What we have right now becomes the production branch (configs/ can
stay where it is)
- We branch off to development (and testing if desirable), and start
moving configs/ content into the modules for as far as these are related
to the modules (e.g. templates, new files, but not "simple" files).
- Changes go into the development branch first, then are pulled into /
pushed to production.
Any other thoughts?
Kind regards,
Jeroen van Meeuwen
-kanarip
More information about the Fedora-infrastructure-list
mailing list