RoleSpec

A shell based test library for Ansible that works both locally and over Travis-CI.

• Typical Travis setup vs using RoleSpec
  • Travis on its own
  • RoleSpec
• Comparing real examples
  • The old typical Travis way
  • The same test case using RoleSpec
• Installation
  • Getting setup locally
• Ways to organize your tests
• Write your first test case
  • Basic test scaffold
  • Investigate the hosts file
  • Create a playbook
  • Investigate the test
  • Lint it
  • Run it
  • Test modes
  • Wrapping things up
• Example test cases
• Test API
  • System actions
  • Ansible actions and assertions
  • Basic assertions
  • Available variables
• Test code style
• Do you want to contribute?
• Author

Typical Travis setup vs using RoleSpec

Travis on its own

• Tons of duplication
• Locked into using Travis
• Running bash in YAML syntax
• Manual dependency management
• Tons of "Damn you Travis, please work" commits

RoleSpec

• Still uses Travis, so you can keep the badges and CI environment
• Runs on any Debian-based OS in case you want to use it locally
• Customized assertions optimized for Ansible
• ~75% less code written
• Automatic dependency resolution
• Automatic debug outputs
• Custom linter
• Your test cases end up being fully working examples/documentation
• Multiple test modes to optimize iteration speed

Comparing real examples

The snippets below come from testing a Rails deployment role.

The old typical Travis way

---

language: "python"
python: "2.7"

env:
  - SITE="tests/main.yml -i 'localhost,'"

install:
  - "pip install ansible"
  - "printf '[defaults]\\nroles_path = ../' > ansible.cfg"

before_script:
  - >
    sudo ansible-galaxy install debops.secret debops.etc_services \
    debops.postgresql debops.nginx debops.monit

    git config --global user.email '[email protected]' \
    && git config --global user.name 'Foo Bar'

    cd tests/testapp && git init && cd -

script:
  - "ansible-playbook $SITE --syntax-check"
  - "ansible-playbook $SITE --connection=local -vvvv"
  - >
    ansible-playbook $SITE --connection=local
    | grep -q "changed=0.*failed=0"
    && (echo "Idempotence test: PASS" && exit 0)
    || (echo "Idempotence test: FAIL" && exit 1)
  - sleep 5
  - >
    sudo cat /srv/users/testapp/.ssh/id_rsa
    | grep -q "ssh"
    && (echo "Private key: PASS" && exit 0)
    || (echo "Private key: FAIL" && exit 1)
  - >
    sudo groups testuser
    | grep -q "audio"
    && (echo "Group: PASS" && exit 0)
    || (echo "Group: FAIL" && exit 1)
  - >
    sudo stat -c "%a %n" /srv/users/testapp
    | grep -q "751"
    && (echo "Secure home: PASS" && exit 0)
    || (echo "Secure home: FAIL" && exit 1)
  - >
    sudo cat /etc/logrotate.d/testapp
    | grep -q "{.*}"
    && (echo "Rotated logs: PASS" && exit 0)
    || (echo "Rotated logs: FAIL" && exit 1)
  - >
    curl -k -s -o /dev/null -w "%{http_code}" https://localhost
    | grep -q "200"
    && (echo "SSL 200 - Testapp: PASS" && exit 0)
    || (echo "SSL 200 - Testapp: FAIL" && exit 1)
  - >
    curl -k -s -o /dev/null -w "%{http_code}" https://localhost/sidekiq
    | grep -q "200"
    && (echo "SSL 200 - Sidekiq: PASS" && exit 0)
    || (echo "SSL 200 - Sidekiq: FAIL" && exit 1)
  - >
    sudo monit status
    | grep -q "testapp"
    && (echo "Monitoring Testapp: PASS" && exit 0)
    || (echo "Monitoring Testapp: FAIL" && exit 1)
  - >
    sudo monit status
    | grep -q "sidekiq"
    && (echo "Monitoring Sidekiq: PASS" && exit 0)
  || (echo "Monitoring Sidekiq: FAIL" && exit 1)

The same test case using RoleSpec

#!/bin/bash

. "${ROLESPEC_LIB}/main"

install_ansible "v1.7.1"

cd "${ROLESPEC_TEST}/test_files/testapp" && git init && cd -

assert_playbook_runs
assert_playbook_idempotent
assert_playbook_idempotent_long

assert_permission "/srv/users/testapp" "751"
assert_user_in_group "testuser" "audio"

assert_in_file "/srv/users/testapp/.ssh/id_rsa" "ssh"
assert_in_file "/etc/logrotate.d/testapp" "{.*}"

assert_url "https://${ROLESPEC_FQDN}"
assert_url "https://${ROLESPEC_FQDN}/sidekiq"

assert_monitoring "testapp"
assert_monitoring "sidekiq"

Installation

If you're using it on Travis then you don't need to download anything.

Use this .travis.yml as a guide, it would go in each of your role's repositories:

---

# Ensure Python 2.7.x is being used
language: 'python'
python: '2.7'

# Use system installed packages inside of the Virtual environment
virtualenv:
  system_site_packages: True

# Skip running these which boosts the Travis boot time
before_install: True
install: True

script:
  # Clone the RoleSpec repo, feel free to use --branch xxx to use something
  # other than the master branch (latest stable)
  - 'git clone --depth 1 https://github.com/nickjj/rolespec'

  # The location of YOUR test suite
  - 'cd rolespec ; bin/rolespec -r https://github.com/you/some-test-suite'

You can also use RoleSpec locally, perhaps in a container or virtual machine.

git clone https://github.com/nickjj/rolespec
cd rolespec ; sudo make install

Getting setup locally

You'll probably want to run tests locally in a container or VM so you can iterate on them quicker. Then once you're ready you could push it out to Travis. We will go over on how to do this shortly.

Ways to organize your tests

Dedicated test suite

It would consist of 1 repository that contains isolated test cases for each role you have. This is how we do it for DebOps. Check out the DebOps test suite for a working example.

This allows you to not pollute your role's commit history with things like "Travis is a jerk face, attempt 42 finally worked!". It also makes it convenient for adding new tests.

A tests/ directory in each role

Not supported right now but it could be in the future. I'm looking for feedback to see if the demand is there. Let me know by opening an issue or by contacting me, #debops on Freenode or @nickjanetakis.

Write your first test case

Let's create a new test in a container/VM. I'm going to assume by now you have installed RoleSpec.

First off we'll want to init a new working directory. This is where all of your roles and tests will be stored. It can be located anywhere you want. Run this:

rolespec -i ~/foo

From this point on I'm going to assume you're in your working directory. All paths will be relative to that.

For this example let's make pretend we have the following setup:

• Your role name is foo
• Your Ansible Galaxy name is someperson
• Your role is on GitHub at github.com/someperson/ansible-foo
• Your tests are on GitHub at github.com/someperson/test-suite
• Your test is located in the ansible-foo directory in the test-suite repo

NOTE: Galaxy and GitHub are not necessary for any of this, it is just an example.

Let's create a role locally and make it do the least amount possible just so we can test it.

mkdir -p roles/someperson.foo/tasks && touch roles/someperson.foo/tasks/main.yml

Basic test scaffold

rolespec -n tests/ansible-foo to create a new test case for this role.

Investigate the hosts file

RoleSpec provides you with many variables and will also do find/replaces on your test to replace placeholders at runtime. The hosts file is one spot where you will use a placeholder.

You will notice it contains nothing except placeholder_fqdn. You can put it in 1 or more groups if you want. All instances of that string will get swapped to the real fully qualified domain name of the host.

Create a playbook

You don't have to make one because RoleSpec will generate one at runtime for you. It will consist of running the play against the FQDN of the host (all groups essentially) and set the role you're testing.

If you want more control over the generated playbook then you can supply a custom playbook of your own, it must be located at tests/ansible-foo/playbooks/test.yml.

Investigate the test

Open up tests/ansible-foo/test and read through it. It's commented and explains everything.

Lint it

You can optionally run rolespec -l to run a linter against all of your tests. It will report back missing files, warn you if you're missing key things in your test script/yaml files and perform a syntax check.

• RED results will cause your test to not run
• YELLOW results are warnings that you should fix but are pretty ok to ignore
• No results is great, that means everything is syntactically valid and well formed

Try running it now, you may see some feedback.

Run it

rolespec -r foo

It should run successfully and you'll be greeted with passing tests at the end. Here's a cool tip too, if you run bash -x rolespec -r foo instead you will be provided with an in depth debug output as it runs.

Test modes

By default RoleSpec will run the full setup/teardown stack. That includes tasks like installing system packages, installing Ansible, running the playbook and the assertions. This is good to run when you want to do a full test.

Sometimes you just want to quickly iterate on a playbook and you don't care about resetting all of the system packages, etc.. You can run RoleSpec in playbook mode like so:

rolespec -r foo -p

Last up is turbo mode which skips everything except running your assertions. This allows you to work against a static state of the system. Perfect for when you want to write a bunch of assertions against a known setup. You can run that like so:

rolespec -r foo -t

Wrapping things up

If you ever get lost then run rolespec -h to bring up the help menu. Also don't forget that each test is basically a standalone guide on how to use your role. Feel free to use inventory/group_vars or meta/main.yml in your test if you need to.

Example test cases

You can view over 50 working examples in the DebOps test suite.

Test API

System actions

Do not use quotes when calling any system functions, they must be passed as arguments.

• install <space separated list of apt packages>
• purge <space separated list of apt packages>
• start <service name>
• stop <service name>

Ansible actions and assertions

• install_ansible [branch=devel]

  • Installs a specific version of Ansible

You may optionally pass ansible-playbook arguments to any of the functions below.

• assert_playbook_syntax

  • Performs just a syntax check

• assert_playbook_runs

  • Performs a syntax check and runs the playbook once

• assert_playbook_check_runs

  • Performs a syntax check and runs ansible in check mode and runs the playbook once

• assert_playbook_idempotent

  • Re-runs the playbook checking for 0 changes

• assert_playbook_idempotent_long

  • Re-runs the playbook checking for 0 changes with periodic output

Basic assertions

Add an ! as an optional last argument to any of the functions below to negate them.

• assert_in <command output or string> <search pattern>
• assert_in_file <path> <search pattern>
• assert_path <path>
• assert_permission <path> <octal permission>
• assert_group <space separated list of groups>
• assert_user_in_group <user> <group>
• assert_running <process name>
• assert_monitoring <process name>
• assert_iptables_allow <port or service name>
• assert_url <full url> [status code=200]
• assert_tcp <hostname> <port> [return code=0]
• assert_exit_code <command> <expected exit code>

Available variables

• ROLESPEC_ANSIBLE_INSTALL

  • The path where Ansible has been installed to

• ROLESPEC_ANSIBLE_SOURCE

  • The address where Ansible has been cloned from

• ROLESPEC_ANSIBLE_ROLES

  • The roles_path which gets set in ansible.cfg

• ROLESPEC_ANSIBLE_CONFIG

  • The path where ansible.cfg exists

• ROLESPEC_LIB

  • The path where RoleSpec's libs exist

• ROLESPEC_VERSION

  • The version of RoleSpec

• ROLESPEC_RELEASE_NAME

  • The release name of the host's OS

• ROLESPEC_FQDN

  • The fully qualified domain name of the host

• ROLESPEC_TRAVIS

  • Is the host running on Travis-CI?

• ROLESPEC_TRAVIS_ROLES_PATH

  • The path where roles are downloaded from Travis-CI

• ROLESPEC_TURBO_MODE

  • Has turbo mode been enabled?

• ROLESPEC_DEVELOPMENT_MODE

  • Has development mode been enabled?

• ROLESPEC_ROLES

  • The path where roles are downloaded from Ansible Galaxy

• ROLESPEC_ROLE

  • The name of the role as it exists on the file system

• ROLESPEC_ROLE_NAME

  • The name of the role without any galaxy or repository prefix

• ROLESPEC_TEST

  • The path of the test directory for the current role being tested

• ROLESPEC_HOSTS

  • The path of its hosts file

• ROLESPEC_META

  • The path of its meta file

• ROLESPEC_PLAYBOOK

  • The path of its playbook file

• ROLESPEC_SCRIPT

  • The path of its test file

• ROLESPEC_POSTGRESQL_LIBS

  • A list of packages to purge before installing PostgreSQL

• ROLESPEC_MYSQL_LIBS

  • A list of packages to purge before installing MySQL

Test code style

Up to you but so far I'm digging this, each section gets separated by 2 lines:

1. Header comments
2. Source the RoleSpec lib
3. Stop service / purge apt packages
4. Install apt packages
5. Any type of setup code that needs to happen before the playbook is ran
6. install_ansible [version]
7. assert_playbook_runs and optionally assert_playbook_idempotent
8. All of your tests, separated by 0 or 1 lines
9. Any cleanup code that needs to happen, such as stopping a server

Do you want to contribute?

Sounds great, check out the contributing guide for the details.

Author

Nick Janetakis

• Email: [email protected]
• Twitter: @nickjanetakis
• GitHub: nickjj