Chef InSpec Migration Guide
How is Chef InSpec different from Serverspec
We’ve written a complete blog post about that topic: The Road to InSpec
Is Chef InSpec suitable for infrastructure testing?
Chef InSpec is a framework that allows you to run infrastructure testing as well as compliance testing. The compliance features are always optional and provide customers a way to use Chef InSpec for both use-cases. To ensure we build the best infrastructure testing, we migrate our cookbooks chef-cookbooks to InSpec.
Which Serverspec resources are available in InSpec?
The following resources are available in InSpec:
Some Serverspec resources are not available yet. We will implement those resources based on user feedback. If you need a resource that is not available in InSpec, please open an Github issue. The list of resources that are not available in InSpec:
- cgroup
- default_gateway
- ip6tables
- ipfilter
- ipnat
- linux_audit_system
- lxc
- mail_alias
- php_config
- routing_table
- selinux
- selinux_module
- x509_private_key
In addition Chef InSpec provides additional resources that are not available in Serverspec:
- apache_conf
- apt
- audit_policy
- auditd_conf
- bash
- csv
- shadow
- gem
- grub_conf
- inetd_conf
- ini
- json
- npm
- ntp_conf
- oneget
- pip
- powershell
- security_policy
- ssh_config
- sshd_config
- sys_info
How do I migrate my Serverspec tests to InSpec
For most cases, the migration to Chef InSpec is pretty straight forward. First, replace the current verifier in kitchen.yml
configuration with:
verifier:
name: inspec
Second, rename the directory test/integration/default/serverspec
to
test/integration/default/inspec
Third, remove the Serverspec-specific code from the test files.
require 'serverspec'
# Required by serverspec
set :backend, :exec
Chef InSpec is now configured with Test-Kitchen:
kitchen verify package-install-centos-72
-----> Starting Kitchen (v1.14.2)
-----> Verifying <package-install-centos-72>...
Detected alternative framework tests for `inspec`
Loaded
Target: ssh://vagrant@127.0.0.1:2200
PHP has
✔ php
✔ the pear.php.net channel
✔ the pecl.php.net channel
Test Summary: 3 successful, 0 failures, 0 skipped
Finished verifying <package-install-centos-72> (0m0.40s).
-----> Kitchen is finished. (0m3.31s)
Some real-world migrations are available:
Some general recommendations:
- use test-kitchen 1.14+
- in case of errors, increase the log level
kitchen verify package-install-centos-72 -l debug
Do I still need the backend configuration?
Chef InSpec does not attach backend information to test files. All tests are defined independently of any backend. Therefore a Serverspec test file:
require 'serverspec'
# Required by serverspec
set :backend, :exec
describe 'PHP' do
it 'has php' do
expect(command('php -v').exit_status).to eq(0)
end
it 'has the pear.php.net channel' do
expect(command('pear list-channels').stdout).to include('pear.php.net')
end
it 'has the pecl.php.net channel' do
expect(command('pear list-channels').stdout).to include('pecl.php.net')
end
end
will become the following Chef InSpec test file:
describe 'PHP' do
it 'has php' do
expect(command('php -v').exit_status).to eq(0)
end
it 'has the pear.php.net channel' do
expect(command('pear list-channels').stdout).to include('pear.php.net')
end
it 'has the pecl.php.net channel' do
expect(command('pear list-channels').stdout).to include('pecl.php.net')
end
end
As you can see, the Chef InSpec test files just focuses on tests and tries to avoid all clutter.
Nested describe blocks
Serverspec and RSpec allow you to define nested describe blocks. We did a survey and found out that most users use nested describe blocks only to improve their output report. We believe the code structure should not change to improve the output of a report. Nevertheless we understand that nested describe blocks help you to structure test code. A sample code block looks like:
describe 'chef-server-directories' do
describe file('/etc/opscode') do
it { should be_directory }
it { should be_owned_by 'root' }
end
describe file('/etc/opscode-analytics') do
it { should be_directory }
it { should be_owned_by 'opscode' }
it { should be_grouped_into 'opscode' }
end
describe file('/var/log/opscode') do
it { should be_directory }
it { should be_owned_by 'opscode' }
it { should be_grouped_into 'opscode' }
end
describe file('/var/opt/opscode') do
it { should be_directory }
it { should be_owned_by 'root' }
end
end
In Chef InSpec you would split up groups into files.
tests
├── server-directories.rb
├── other-tests.rb
└── further-tests.rb
Each file can have a top-level description of its content:
title "Chef Server Directories"
describe file('/etc/opscode') do
it { should be_directory }
it { should be_owned_by 'root' }
end
describe file('/etc/opscode-analytics') do
it { should be_directory }
it { should be_owned_by 'opscode' }
it { should be_grouped_into 'opscode' }
end
describe file('/var/log/opscode') do
it { should be_directory }
it { should be_owned_by 'opscode' }
it { should be_grouped_into 'opscode' }
end
describe file('/var/opt/opscode') do
it { should be_directory }
it { should be_owned_by 'root' }
end
Are you supporting the expect
syntax?
Of course. We still prefer the should
syntax for UX reasons. We did surveys with various types of customers like devops engineers, auditors, managers. All participants who preferred the expect
syntax have been Ruby experts. All non-Ruby developers found it easier to understand the should
syntax.
should
syntax with InSpec
describe command('php -v') do
its('exit_status') { should eq 0 }
end
describe command('pear list-channels') do
its('stdout') { should include('pear.php.net')}
end
describe command('pear list-channels') do
its('stdout') { should include('pecl.php.net')}
end
expect
syntax with InSpec
describe 'PHP' do
it 'has php' do
expect(command('php -v').exit_status).to eq(0)
end
it 'has the pear.php.net channel' do
expect(command('pear list-channels').stdout).to include('pear.php.net')
end
it 'has the pecl.php.net channel' do
expect(command('pear list-channels').stdout).to include('pecl.php.net')
end
end
Was this page helpful?