Puppet function and data backend for Hiera
Switch branches/tags
Nothing to show
Clone or download
Pull request Compare This branch is 84 commits behind puppetlabs:master.
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.



A data backend for Hiera that can query the internal Puppet scope for data. The data structure and approach is heavily based on work by Nigel Kersten but made more configurable and with full hierarchy.

It also includes a Puppet function that works like extlookup() but uses the Hiera backends.


Hiera supports the concept of chaining backends together in order, using this we can create a very solid module author/module user experience.

Module Author

A module author wants to create a configurable module that has sane defaults but want to retain the ability for users to configure it.

We'll use a simple NTP config class as an example.

class ntp::config($ntpservers = hiera("ntpservers")) {
       content => template("ntp.conf.erb")

We create a class that takes as parameters a list of NTP servers.

The module author wants to create a works-out-of-the-box experience so creates a data class for the NTP module:

class ntp::data {
   $ntpservers = ["1.pool.ntp.org", "2.pool.ntp.org"]

Together this creates a default sane setup.

Module User

The module user has a complex multi data center setup, he wants to use the NTP module from the forge and configure it for his needs.

The user creates a set of default data for his organization, he can do this in data files or in Puppet. We'll show a Puppet example.

class data::common {
   $ntpservers = ["ntp1.example.com", "ntp2.example.com"]

Being part of the actual code this data is subject to strict change control. This is needed as its data that can potentially affect all machines in all locations.

The user has a fact called location that contains, for example, a name of the data center.

He decides to create JSON based data for the data centers, being just data that applies to one data center this data is not subject to as strict change controls and so does not live with the code:

He creates /var/lib/hiera/dc1.json with the following:

{"ntpservers" : ["ntp1.dc1.example.com", "ntp2.dc1.example.com"]}

Machines in dc1 will now use specific NTP servers while all the rest will use the data in data::common

The module user can now just declare the class on his nodes:

node "web1" {
   include ntp::config

For true one-off changes, the user can use the full paramterized class approach that will completely disable the Hiera handling of this data item. He could also use an ENC to supply this data.

node "web2" {
   class{"ntp::config": ntpservers => ["another.example.com"]}

This behavior is thanks to Hiera's ability to search through multiple backends for data picking the first match. We can have the JSON searched before the internal Puppet data.

To achieve this setup the module user needs to configure Hiera in /etc/puppet/hiera.yaml:

:backends: - json
           - puppet
:hierarchy: - %{location}
            - common
        :datadir: /var/lib/hiera

	:datasource: data

Converting from extlookup?

A simple converter is included called extlookup2hiera and it can convert from CSV to JSON or YAML:

$ extlookup2hiera --in common.csv --out common.json --json


It's not 100% ready for prime time, shortly a simple gem install hiera-puppet on your master will do it.

For the moment the Gem install will place the Puppet Parser Function where Puppet cannot find it, you should copy it out and distribute it to your master using Pluginsync or something similar


R.I.Pienaar / rip@devco.net / @ripienaar / www.devco.net