RSpec tests for your Puppet manifests & modules

Build Status Coverage Status

Installation

gem install rspec-puppet

Note for ruby 1.8 users: while rspec-puppet itself supports ruby 1.8, you'll need to pin rspec itself to ~> 3.1.0, as later rspec versions do not work on old rubies anymore.

Starting out with a new module

When you start out on a new module, create a metadata.json file for your module and then run rspec-puppet-init to create the necessary files to configure rspec-puppet for your module's tests.

Configure manifests for Puppet 4

With Puppet 3, the manifest is set to $manifestdir/site.pp. However Puppet 4 defaults to an empty value. In order to test manifests you will need to set appropriate settings.

Puppet configuration reference for manifest can be found online:

Configuration is typically done in a spec/spec_helper.rb file which each of your spec will require. Example code:

# /spec
base_dir = File.dirname(File.expand_path(__FILE__))

RSpec.configure do |c|
  c.module_path     = File.join(base_dir, 'fixtures', 'modules')
  c.manifest_dir    = File.join(base_dir, 'fixtures', 'manifests')
  c.manifest        = File.join(base_dir, 'fixtures', 'manifests', 'site.pp')
  c.environmentpath = File.join(Dir.pwd, 'spec')

  # Coverage generation
  c.after(:suite) do
    RSpec::Puppet::Coverage.report!
  end
end

Configuration

rspec-puppet can be configured by modifying the RSpec.configure block in your spec/spec_helper.rb file.

RSpec.configure do |c|
  c.<config option> = <value>
end

manifest_dir

Type Default Puppet Version(s)
String Required 2.x, 3.x

The path to the directory containing your basic manifests like site.pp.

module_path

Type Default Puppet Version(s)
String Required 2.x, 3.x, 4.x, 5.x

The path to the directory containing your Puppet modules.

default_facts

Type Default Puppet Version(s)
Hash {} 2.x, 3.x, 4.x, 5.x

A hash of default facts that should be used for all the tests.

hiera_config

Type Default Puppet Version(s)
String "/dev/null" 3.x, 4.x, 5.x

The path to your hiera.yaml file (if used).

default_node_params

Type Default Puppet Version(s)
Hash {} 4.x, 5.x

A hash of default node parameters that should be used for all the tests.

default_trusted_facts

Type Default Puppet Version(s)
Hash {} 4.x, 5.x

A hash of default trusted facts that should be used for all the tests (available in the manifests as the $trusted hash). In order to use this, the trusted_node_data setting must be set to true.

trusted_node_data

Type Default Puppet Version(s)
Boolean false >=3.4, 4.x, 5.x

Configures rspec-puppet to use the $trusted hash when compiling the catalogues.

trusted_server_facts

Type Default Puppet Version(s)
Boolean false >=4.3, 5.x

Configures rspec-puppet to use the $server_facts hash when compiling the catalogues.

confdir

Type Default Puppet Version(s)
String "/etc/puppet" 2.x, 3.x, 4.x, 5.x

The path to the main Puppet configuration directory.

config

Type Default Puppet Version(s)
String Puppet's default value 2.x, 3.x, 4.x, 5.x

The path to puppet.conf.

manifest

Type Default Puppet Version(s)
String Puppet's default value 2.x, 3.x

The entry-point manifest for Puppet, usually $manifest_dir/site.pp.

template_dir

Type Default Puppet Version(s)
String nil 2.x, 3.x

The path to the directory that Puppet should search for templates that are stored outside of modules.

environmentpath

Type Default Puppet Version(s)
String "/etc/puppetlabs/code/environments" 4.x, 5.x

The search path for environment directories.

parser

Type Default Puppet Version(s)
String "current" >= 3.2

This switches between the 3.x (current) and 4.x (future) parsers.

ordering

Type Default Puppet Version(s)
String "title-hash" >= 3.3, 4.x, 5.x

How unrelated resources should be ordered when applying a catalogue.

  • manifest - Use the order in which the resources are declared in the manifest.
  • title-hash - Order the resources randomly, but in a consistent manner across runs (the order will only change if the manifest changes).
  • random - Order the resources randomly.

strict_variables

Type Default Puppet Version(s)
Boolean false >= 3.5, 4.x, 5.x

Makes Puppet raise an error when it tries to reference a variable that hasn't been defined (not including variables that have been explicitly set to undef).

stringify_facts

Type Default Puppet Version(s)
Boolean true >= 3.3, 4.x, 5.x

Makes rspec-puppet coerce all the fact values into strings (matching the behaviour of older versions of Puppet).

enable_pathname_stubbing

Type Default Puppet Version(s)
Boolean false 2.x, 3.x, 4.x, 5.x

Configures rspec-puppet to stub out Pathname#absolute? with it's own implementation. This should only be enabled if you're running into an issue running cross-platform tests where you have Ruby code (types, providers, functions, etc) that use Pathname#absolute?.

setup_fixtures

Type Default Puppet Version(s)
Boolean true 2.x, 3.x, 4.x, 5.x

Configures rspec-puppet to automatically create a link from the root of your module to spec/fixtures/<module name> at the beginning of the test run.

derive_node_facts_from_nodename

Type Default Puppet Version(s)
Boolean true 2.x, 3.x, 4.x, 5.x

If true, rspec-puppet will override the fdqn, hostname, and domain facts with values that it derives from the node name (specified with let(:node).

In some circumstances (e.g. where your nodename/certname is not the same as your FQDN), this behaviour is undesirable and can be disabled by changing this setting to false.

Naming conventions

For clarity and consistency, I recommend that you use the following directory structure and naming convention.

module/
  ├── manifests/
  ├── lib/
  └── spec/
       ├── spec_helper.rb
       │
       ├── classes/
       │     └── <class_name>_spec.rb
       │
       ├── defines/
       │     └── <define_name>_spec.rb
       │
       ├── applications/
       │     └── <application_name>_spec.rb
       │
       ├── functions/
       │     └── <function_name>_spec.rb
       │
       ├── types/
       │     └── <type_name>_spec.rb
       │
       ├── type_aliases/
       │     └── <type_alias_name>_spec.rb
       │
       └── hosts/
             └── <host_name>_spec.rb

Example groups

If you use the above directory structure, your examples will automatically be placed in the correct groups and have access to the custom matchers. If you choose not to, you can force the examples into the required groups as follows.

describe 'myclass', :type => :class do
  ...
end

describe 'mydefine', :type => :define do
  ...
end

describe 'myapplication', :type => :application do
  ...
end

describe 'myfunction', :type => :puppet_function do
  ...
end

describe 'mytype', :type => :type do
  ...
end

describe 'My::TypeAlias', :type => :type_alias do
  ...
end

describe 'myhost.example.com', :type => :host do
  ...
end

Defined Types, Classes & Applications

Matchers

Checking if the catalog compiles

You can test whether the subject catalog compiles cleanly with compile.

it { is_expected.to compile }

To check the error messages of your class, you can check for raised error messages.

it { is_expected.to compile.and_raise_error(/error message match/) }

Checking if a resource exists

You can test if a resource exists in the catalogue with the generic contain_<resource type> matcher.

it { is_expected.to contain_augeas('bleh') }

You can also test if a class has been included in the catalogue with the same matcher.

it { is_expected.to contain_class('foo') }

Note that rspec-puppet does none of the class name parsing and lookup that the puppet parser would do for you. The matcher only accepts fully qualified classnames without any leading colons. That is a class foo::bar will only be matched by foo::bar, but not by ::foo::bar, or bar alone.

If your resource type includes :: (e.g. foo::bar simply replace the :: with __ (two underscores).

it { is_expected.to contain_foo__bar('baz') }

You can further test the parameters that have been passed to the resources with the generic with_<parameter> chains.

it { is_expected.to contain_package('mysql-server').with_ensure('present') }

If you want to specify that the given parameters should be the only ones passed to the resource, use the only_with_<parameter> chains.

it { is_expected.to contain_package('httpd').only_with_ensure('latest') }

You can use the with method to verify the value of multiple parameters.

it do
  is_expected.to contain_service('keystone').with(
    'ensure'     => 'running',
    'enable'     => 'true',
    'hasstatus'  => 'true',
    'hasrestart' => 'true'
  )
end

The same holds for the only_with method, which in addition verifies the exact set of parameters and values for the resource in the catalogue.

it do
  is_expected.to contain_user('luke').only_with(
    'ensure' => 'present',
    'uid'    => '501'
  )
end

You can also test that specific parameters have been left undefined with the generic without_<parameter> chains.

it { is_expected.to contain_file('/foo/bar').without_mode }

You can use the without method to verify that a list of parameters have not been defined

it { is_expected.to contain_service('keystone').without(
  ['restart', 'status']
)}

Checking the number of resources

You can test the number of resources in the catalogue with the have_resource_count matcher.

it { is_expected.to have_resource_count(2) }

The number of classes in the catalogue can be checked with the have_class_count matcher.

it { is_expected.to have_class_count(2) }

You can also test the number of a specific resource type, by using the generic have_<resource type>_resource_count matcher.

it { is_expected.to have_exec_resource_count(1) }

This last matcher also works for defined types. If the resource type contains ::, you can replace it with __ (two underscores).

it { is_expected.to have_logrotate__rule_resource_count(3) }

NOTE: when testing a class, the catalogue generated will always contain at least one class, the class under test. The same holds for defined types, the catalogue generated when testing a defined type will have at least one resource (the defined type itself).

Relationship matchers

The following methods will allow you to test the relationships between the resources in your catalogue, regardless of how the relationship is defined. This means that it doesn’t matter if you prefer to define your relationships with the metaparameters (require, before, notify and subscribe) or the chaining arrows (->, ~>, <- and <~), they’re all tested the same.

it { is_expected.to contain_file('foo').that_requires('File[bar]') }
it { is_expected.to contain_file('foo').that_comes_before('File[bar]') }
it { is_expected.to contain_file('foo').that_notifies('File[bar]') }
it { is_expected.to contain_file('foo').that_subscribes_to('File[bar]') }

An array can be used to test a resource for multiple relationships

it { is_expected.to contain_file('foo').that_requires(['File[bar]', 'File[baz]']) }
it { is_expected.to contain_file('foo').that_comes_before(['File[bar]','File[baz]']) }
it { is_expected.to contain_file('foo').that_notifies(['File[bar]', 'File[baz]']) }
it { is_expected.to contain_file('foo').that_subscribes_to(['File[bar]', 'File[baz]']) }

You can also test the reverse direction of the relationship, so if you have the following bit of Puppet code

notify { 'foo': }
notify { 'bar':
  before => Notify['foo'],
}

You can test that Notify[bar] comes before Notify[foo]

it { is_expected.to contain_notify('bar').that_comes_before('Notify[foo]') }

Or, you can test that Notify[foo] requires Notify[bar]

it { is_expected.to contain_notify('foo').that_requires('Notify[bar]') }
Match target syntax

Note that this notation does not support any of the features you're used from the puppet language. Only a single resource with a single, unquoted title can be referenced here. Class names need to be always fully qualified and not have the leading ::. It currently does not support inline arrays or quoting.

These work

  • Notify[foo]
  • Class[profile::apache]

These will not work

  • Notify['foo']
  • Notify[foo, bar]
  • Class[::profile::apache]
Recursive dependencies

The relationship matchers are recursive in two directions:

  • vertical recursion, which checks for dependencies with parents of the resource (i.e. the resource is contained, directly or not, in the class involved in the relationship). E.g. where Package['foo'] comes before File['/foo']:
class { 'foo::install': } ->
class { 'foo::config': }

class foo::install {
  package { 'foo': }
}

class foo::config {
  file { '/foo': }
}
  • horizontal recursion, which follows indirect dependencies (dependencies of dependencies). E.g. where Yumrepo['foo'] comes before File['/foo']:
class { 'foo::repo': } ->
class { 'foo::install': } ->
class { 'foo::config': }

class foo::repo {
  yumrepo { 'foo': }
}

class foo::install {
  package { 'foo': }
}

class foo::config {
  file { '/foo': }
}
Autorequires

Autorequires are considered in dependency checks.

Type matcher

When testing custom types, the be_valid_type matcher provides a range of expectations:

  • with_provider(<provider_name>): check that the right provider was selected
  • with_properties(<property_list>): check that the specified properties are available
  • with_parameters(<parameter_list>): check that the specified parameters are available
  • with_features(<feature_list>): check that the specified features are available
  • with_set_attributes(<param_value_hash>): check that the specified attributes are set

Type alias matchers

When testing type aliases, the allow_value and allow_values matchers are used to check if the alias accepts particular values or not:

describe 'MyModule::Shape' do
  it { is_expected.to allow_value('square') }
  it { is_expected.to allow_values('circle', 'triangle') }
  it { is_expected.not_to allow_value('blue') }
end

Writing tests

Basic test structure

To test that

sysctl { 'baz'
  value => 'foo',
}

Will cause the following resource to be in included in catalogue for a host

exec { 'sysctl/reload':
  command => '/sbin/sysctl -p /etc/sysctl.conf',
}

We can write the following testcase (in spec/defines/sysctl_spec.rb)

describe 'sysctl' do
  let(:title) { 'baz' }
  let(:params) { { 'value' => 'foo' } }

  it { is_expected.to contain_exec('sysctl/reload').with_command("/sbin/sysctl -p /etc/sysctl.conf") }
end

Specifying the title of a resource

let(:title) { 'foo' }

Specifying the parameters to pass to a resources or parameterised class

Parameters of a defined type, class or application can be passed defining :params in a let, and passing it a hash as seen below.

let(:params) { {'ensure' => 'present', ...} }

For passing Puppet's undef as a paremeter value, you can simply use :undef and it will be translated to undef when compiling. For example:

let(:params) { {'user' => :undef, ...} }

For references to nodes or resources as seen when using require or before properties, or an application resource you can pass the string as an argument to the ref helper:

let(:params) { 'require' => ref('Package', 'sudoku') }

Which translates to:

mydefine { 'mytitle': require => Package['sudoku'] }

Another example, for an application setup (when using app_management):

let(:params) { { 'nodes' => { ref('Node', 'dbnode') => ref('Myapp::Mycomponent', 'myapp') } } }

Will translate to:

site { myapp { 'myimpl': nodes => { Node['dbnode'] => Myapp::Mycomponent['myimpl'] } } }

Specifying the FQDN of the test node

If the manifest you're testing expects to run on host with a particular name, you can specify this as follows

let(:node) { 'testhost.example.com' }

Specifying the environment name

If the manifest you're testing expects to evaluate the environment name, you can specify this as follows

let(:environment) { 'production' }

Specifying the facts that should be available to your manifest

By default, the test environment contains no facts for your manifest to use. You can set them with a hash

let(:facts) { {'operatingsystem' => 'Debian', 'kernel' => 'Linux', ...} }

Facts may be expressed as a value (shown in the previous example) or a structure. Fact keys may be expressed as either symbols or strings. A key will be converted to a lower case string to align with the Facter standard

let(:facts) { {'os' => { 'family' => 'RedHat', 'release' => { 'major' => '7', 'minor' => '1', 'full' => '7.1.1503' } } } }

You can also create a set of default facts provided to all specs in your spec_helper:

RSpec.configure do |c|
  c.default_facts = {
    'operatingsystem' => 'Ubuntu'
  }
end

Any facts you provide with let(:facts) in a spec will automatically be merged on top of the default facts.

Specifying top-scope variables that should be available to your manifest

You can create top-scope variables much in the same way as an ENC.

let(:node_params) { { 'hostgroup' => 'webservers', 'rack' => 'KK04', 'status' => 'maintenance' } }

You can also create a set of default top-scope variables provided to all specs in your spec_helper:

RSpec.configure do |c|
  c.default_node_params = {
    'owner'  => 'itprod',
    'site'   => 'ams4',
    'status' => 'live'
  }
end

NOTE Setting top-scope variables is not supported in Puppet < 3.0.

Specifying extra code to load (pre-conditions)

If the manifest being tested relies on another class or variables to be set, these can be added via a pre-condition. This code will be evaluated before the tested class.

let(:pre_condition) { 'include other_class' }

This may be useful when testing classes that are modular, e.g. testing apache::mod::foo which relies on a top-level apache class being included first.

The value may be a raw string to be inserted into the Puppet manifest, or an array of strings (manifest fragments) that will be concatenated.

Specifying extra code to load (post-conditions)

In some cases, you may need to ensure that the code that you are testing comes before another set of code. Similar to the :pre_condition hook, you can add a :post_condition hook that will ensure that the added code is evaluated after the tested class.

let(:post_condition) { 'include other_class' }

This may be useful when testing classes that are modular, e.g. testing class do_strange_things::to_the_catalog which must come before class foo.

The value may be a raw string to be inserted into the Puppet manifest, or an array of strings (manifest fragments) that will be concatenated.

Specifying the path to find your modules

I recommend setting a default module path by adding the following code to your spec_helper.rb

RSpec.configure do |c|
  c.module_path = '/path/to/your/module/dir'
end

However, if you want to specify it in each example, you can do so

let(:module_path) { '/path/to/your/module/dir' }

Specifying trusted facts

When testing with Puppet >= 4.3 the trusted facts hash will have the standard trusted fact keys (certname, domain, and hostname) populated based on the node name (as set with :node).

By default, the test environment contains no custom trusted facts (as usually obtained from certificate extensions) and found in the extensions key. If you need to test against specific custom certificate extensions you can set those with a hash. The hash will then be available in $trusted['extensions']

let(:trusted_facts) { {'pp_uuid' => 'ED803750-E3C7-44F5-BB08-41A04433FE2E', '1.3.6.1.4.1.34380.1.2.1' => 'ssl-termination'} }

You can also create a set of default certificate extensions provided to all specs in your spec_helper:

RSpec.configure do |c|
  c.default_trusted_facts = {
    'pp_uuid'                 => 'ED803750-E3C7-44F5-BB08-41A04433FE2E',
    '1.3.6.1.4.1.34380.1.2.1' => 'ssl-termination'
  }
end

Testing Exported Resources

You can test if a resource was exported from the catalogue by using the exported_resources accessor in combination with any of the standard matchers.

You can use exported_resources as the subject of a child context:

context 'exported resources' do
  subject { exported_resources }

  it { is_expected.to contain_file('foo') }
end

You can also use exported_resources directly in a test:

it { expect(exported_resources).to contain_file('foo') }

Testing applications

Applications in some ways behave as defined resources, but are more complex so require a number of elements already documented above to be combined for testing.

A full example of the simplest rspec test for a single component application:

require 'spec_helper'

describe 'orch_app' do
  let(:node) { 'my_node' }
  let(:title) { 'my_awesome_app' }
  let(:params) do
    {
      'nodes' => {
        ref('Node', node) => ref('Orch_app::Db', title),
      }
    }
  end

  it { should compile }
  it { should contain_orch_app(title) }
end

Each piece is required:

  • You must turn on app_management during testing for the handling to work
  • The :node definition is required to be set so later on you can reference it in the :nodes argument within :params
  • Applications act like defined resources, and each require a :title to be defined
  • The :nodes key in :params requires the use of node reference mappings to resource mappings. The ref keyword allows you to provide these (a normal string will not work).

Beyond these requirements, the very basic should compile test and other matchers as you would expect will work the same as classes and defined resources.

Note: for the moment, cross-node support is not available and will return an error. Ensure you model your tests to be single-node for the time being.

Functions

Matchers

All of the standard RSpec matchers are available for you to use when testing Puppet functions.

it 'should be able to do something' do
  subject.execute('foo') == 'bar'
end

For your convenience though, a run matcher exists to provide easier to understand test cases.

it { is_expected.to run.with_params('foo').and_return('bar') }

Writing tests

Basic test structure

require 'spec_helper'

describe '<function name>' do
  ...
end

Specifying the name of the function to test

The name of the function must be provided in the top level description, e.g.

describe 'split' do

Specifying the arguments to pass to the function

You can specify the arguments to pass to your function during the test(s) using either the with_params chain method in the run matcher

it { is_expected.to run.with_params('foo', 'bar', ['baz']) }

Or by using the execute method on the subject directly

it 'something' do
  subject.execute('foo', 'bar', ['baz'])
end

Passing lambdas to the function

A lambda (block) can be passed to functions that support either a required or optional lambda by passing a block to the with_lambda chain method in the run matcher.

it { is_expected.to run.with_lambda { |x| x * 2 }

Testing the results of the function

You can test the result of a function (if it produces one) using either the and_returns chain method in the run matcher

it { is_expected.to run.with_params('foo').and_return('bar') }

Or by using any of the existing RSpec matchers on the subject directly

it 'something' do
  subject.execute('foo') == 'bar'
  subject.execute('baz').should be_an Array
end

Testing the errors thrown by the function

You can test whether the function throws an exception using either the and_raises_error chain method in the run matcher

it { is_expected.to run.with_params('a', 'b').and_raise_error(Puppet::ParseError) }
it { is_expected.not_to run.with_params('a').and_raise_error(Puppet::ParseError) }

Or by using the existing raises_error RSpec matcher

it 'something' do
  expect { subject.execute('a', 'b') }.should raise_error(Puppet::ParseError)
  expect { subject.execute('a') }.should_not raise_error(Puppet::ParseError)
end

Accessing the parser scope where the function is running

Some complex functions require access to the current parser's scope, e.g. for stubbing other parts of the system.

before(:each) { scope.expects(:lookupvar).with('some_variable').returns('some_value') }
it { is_expected.to run.with_params('...').and_return('...') }

Note that this does not work when testing manifests which use custom functions. Instead, you'll need to create a replacement function directly.

before(:each) do
    Puppet::Parser::Functions.newfunction(:custom_function, :type => :rvalue) { |args|
        raise ArgumentError, 'expected foobar' unless args[0] == 'foobar'
        'expected value'
    }
end

Hiera integration

Configuration

Set the hiera config symbol properly in your spec files:

let(:hiera_config) { 'spec/fixtures/hiera/hiera.yaml' }
hiera = Hiera.new(:config => 'spec/fixtures/hiera/hiera.yaml')

Create your spec hiera files

spec/fixtures/hiera/hiera.yaml

---
:backends:
  - yaml
:hierarchy:
  - test
:yaml:
  :datadir: 'spec/fixtures/hiera'

spec/fixtures/hiera/test.yaml

---
ntpserver: ['ntp1.domain.com','ntpXX.domain.com']
user:
  oneuser:
    shell: '/bin/bash'
  twouser:
    shell: '/sbin/nologin'

Use hiera in your tests

  ntpserver = hiera.lookup('ntpserver', nil, nil)
  let(:params) { 'ntpserver' => ntpserver }

Enabling hiera lookups

If you just want to fetch values from hiera (e.g. because you're testing code that uses explicit hiera lookups) just specify the path to the hiera config in your spec_helper.rb

RSpec.configure do |c|
  c.hiera_config = 'spec/fixtures/hiera/hiera.yaml'
end

spec/fixtures/hiera/hiera.yaml

---
:backends:
  - yaml
:yaml:
  :datadir: spec/fixtures/hieradata
:hierarchy:
  - common

Please note: In-module hiera data depends on having a correct metadata.json file. It is strongly recommended that you use metadata-json-lint to automatically check your metadata.json file before running rspec.

Producing coverage reports

You can output a basic resource coverage report with the following in your spec_helper.rb

RSpec.configure do |c|
  c.after(:suite) do
    RSpec::Puppet::Coverage.report!
  end
end

This checks which Puppet resources have been explicitly checked as part of the current test run and outputs both a coverage percentage and a list of untouched resources.

A desired code coverage level can be provided. If this level is not achieved, a test failure will be raised. This can be used with a CI service, such as Jenkins or Bamboo, to enforce code coverage. The following example requires the code coverage to be at least 95%.

RSpec.configure do |c|
  c.after(:suite) do
    RSpec::Puppet::Coverage.report!(95)
  end
end

Resources declared outside of the module being tested (i.e. forge dependencies) are automatically removed from the coverage report. There is one exception for this though: prior to Puppet 4.6.0, resources created by functions (create_resources(), ensure_package(), etc) did not have the required information in them to determine which manifest they came from and so can not be excluded from the coverage report.

For a list of other module development tools see https://puppet.community/plugins/