puppetmaster was used to determine if `agent` or `apply` should be used. But puppetmaster is not required by puppet per default. Puppet may have a config or could find out by itself (...) where the puppet master is.
It changed the code so we only use `apply` if a manifest was passed, otherwise we use `agent`.
This also fixes the example, which did not work the way without this change.
~~~
# Run puppet agent and fail if anything goes wrong
- puppet
~~~
puppet may be configured to operate in `--noop` mode per default.
That is why we must pass a `--no-noop` to make sure, changes are going to be applied.
There is a growing pattern for using ansible to orchestrate runs of
existing puppet code. For instance, the OpenStack Infrastructure team
started using ansible for this very reason. It also turns out that
successfully running puppet and interpreting success or failure is
harder than you'd expect, thus warranting a module and not just a shell
command.
This is ported in from
http://git.openstack.org/cgit/openstack-infra/ansible-puppet
