State of our puppet use

[Ewoud Kohl van Wijngaarden]

Hello all,

In preparation for the meeting I thought I'd write something about the
state of our puppet use.

First of all, we manage access. I think this works reasonably well, but
the workflow for completeness.

* (New) infra member creates a class ovirt_infra::user::USERNAME
* It is submitted as a review to gerrit.ovirt.org
* Once merged, it's added to the production hostgroup
* On the next puppet runs the user is created

The second big thing we do, is manage our jenkins slaves. I'm not sure I
like the workflow because it's so static. I wonder why so many system
dependencies are needed and if we can't build using mock. This ensures a
clean build.

Another thing we must fix is renaming ovirt_infra::jenkins-slave to
ovirt_infra::jenkins_slave or ovirt_infra::jenkins::slave because a dash
is discouraged and broken in modern puppet versions.

The next IMHO big improvement is merging r10k deployments,
http://gerrit.ovirt.org/19141. Once that's done, I'd also like to add
tests through http://gerrit.ovirt.org/20367.

For the future I have some other ideas:

* Use puppet exported resources to automatically build the icinga
  configuration.
* Split linode01 into separate servers and manage those using puppet.
* Documentation

Comments, feedback or anything I missed?