This is an interesting spec test for module developers.
It illustrates how to cause Puppet to test the function
from the Puppet DSL rather than the Ruby DSL, fully
exercising the system from the perspective of the end
user.
(Note how Puppet[:code] is set, then the scope reset, then
the compile method called.)
Paired-with: Dan Bode <dan@puppetlabs.com>
This function aborts catalog compilation if any of the passed
values are not true or false. Note, this catches the string
values of true and false correct and will abort catalog
compilation if they are not boolean values.
Paired-with: Dan Bode <dan@puppetlabs.com>
Working with the stages in stdlib, I quickly ran into an issue where
most of the stages were before the main stage. This made it difficult
to declare any resources in a traditional "include" style class while
hiding the end user from the stages being associated with other module
classes.
For example, in class mcollective, a package would be declared in main.
However, if mcollective declared class mcollective::service in stage
infra_deploy and this was before main, there would be a dependency loop
between the package and the service.
There appears to be a convention around "chain your stages after main"
to avoid the need to create relatively empty shell classes.
While developing Puppet Modules with class parameters, data from the
user should be validated as per the Style Guide. Puppet should fail
early and hard in the situation of invalid data being passed into the
module.
This function provides a more concise method to the alternative of using
if statements in the Puppet manifests.
Many modules I'm working on need a standard but
relatively granular location in the catalog. For example,
any module that configures the packaging system should
run "early"
Add the following stages which have inter-dependencies
in the top to bottom order listed:
* setup
* deploy
* runtime
* setup_infra
* deploy_infra
* main
* setup_app
* deploy_app
