Hi All,
We are going to start using puppet in anger as we start developing some production SL6 services (worker nodes for instance), and I thought it would be worthwhile listing out some of the design decisions we're making before we get too far in so more experienced admins can tell us what's stupid. The plan is to implement this from scratch for the SL6 deployment and not try to translate our existing cfengine set-up.
So here we go.
1) We will use Puppet 3.2 from puppet labs rather than the 2.6 version from EPEL
2) We will use the Roles and Profiles ideas suggested by Kashif with a naming scheme like Role::RALPP::Prod::<name> and Profile::RALPP::Prod::<name>
3) We're going to extend the naming scheme and call the modules that actually configure things "Atoms" (the best name we could come up with over Coffee) with a similar naming convention i.e. Atom:RALPP:Prod:<name>
4) We will use hiera and templates to abstract out as much site local config as possible from the Atoms
We will also probably use augaus to edit files in place where using templates is not practical but don't know enough about this yet to be certain on that.
What other extra modules are people using that you've found useful? Looking around there often seem to be several tools/modules/ways of doing the same things so choosing a common approach would help in sharing.
Thanks,
Chris.
--
Dr Chris Brew
Scientific Computing Manager
Particle Physics Department
STFC - Rutherford Appleton Laboratory
Harwell Oxford,
Didcot
OX11 8TZ
+44 1235 446326
--
Dr Chris Brew
Scientific Computing Manager
Particle Physics Department
STFC - Rutherford Appleton Laboratory
Harwell Oxford,
Didcot
OX11 8TZ
+44 1235 446326
--
Scanned by iCritical.
|