Skip to content
This repository has been archived by the owner on Mar 10, 2020. It is now read-only.
/ pennyworth Public archive

Test framework for running integration tests using virtual machines

License

Notifications You must be signed in to change notification settings

SUSE/pennyworth

Repository files navigation

Pennyworth

Code Climate Gem Version

Pennyworth is a tool for running integration tests inside a network of virtual machines. It allows the user to define virtual machines, build them as Vagrant boxes and run them using libvirt and kvm in a coordinated fashion in order to run the tests. These tests can be written in any language/framework, but the preferred combination is Ruby/RSpec, for which helpers are provided.

Pennyworth is a spin-off of the Machinery project. It is used there to manage the environment of the integration tests.

Contents

Installation

Pennyworth is tested on openSUSE 13.2 and SUSE Linux Enterprise Server 12. It may not work with other openSUSE versions, Linux distributions, or operating systems.

The following steps will make Pennyworth run on a vanilla openSUSE 13.2 system.

  1. Install required packages

    Install Git:

    $ sudo zypper in git
    

    Install basic Ruby environment:

    $ sudo zypper in ruby rubygem-bundler
    

    After the installation, make sure that your ruby20 version is at least 2.0.0.p247-3.11.1:

    $ rpm -q ruby20
    

    With lower versions, bundle install won't work because of a bug.

    Install packages needed to detect a base system:

    $ sudo zypper in lsb-release
    

    Install packages needed to compile Gems with native extensions:

    $ sudo zypper in gcc-c++ make ruby-devel libvirt-devel libxslt-devel libxml2-devel
    
  2. Clone Pennyworth repository and install Gem dependencies required to run the setup

    $ git clone https://github.com/SUSE/pennyworth.git
    $ cd pennyworth
    $ bundle config build.nokogiri --use-system-libraries
    $ bundle install --without test
    
  3. Run the setup

    $ bin/pennyworth setup
    
  4. Install remaining Gem dependencies

    $ bundle install --without ""
    

    Specifying an empty value with --without "" is necessary because this option is “remembered”. If absent, Bundler would use its value from the last invocation (test).

  5. Restart your system

    This refreshes information about current user's groups and network setup.

  6. Done!

    You can now start using Pennyworth.

Overview

Pennyworth is a command-line tool built around defining virtual machines, building them, and running integration tests on them.

The usual workflow is:

  1. Define machines used to run integration tests and build them. The result is a Vagrant box.

  2. Import built boxes into Vagrant.

  3. Run the boxes as needed and execute your test on them. Ruby users can use a RSpec helper to simplify this step.

All these tasks are driven by commands described in the Usage section.

Defining and Building Machines

To define and build machines, Pennyworth uses Kiwi.

When a machine is built, the resulting Vagrant box can be uploaded to a web server (manually). Pennyworth running on some other machine can then import it instead of using a locally-built box, which can save time and ensure everyone has exactly the same environment for running tests.

Running Boxes

To run the boxes, Pennyworth uses Vagrant. However, instead of the default VirtualBox backend it uses KVM driven by libvirt. The theory is that using open source and SUSE-supported technology will be more reliable and performant, which should outweigh the somewhat complicated setup.

Usage

Command Line

Pennyworth is a command-line tool. You can invoke it by using the bin/pennyworth command. It accepts subcommands (similarly to git or bundle).

For example:

  • Building an image $pennyworth build_base example_image --definitions-dir=example_dir

  • Listing available VMs $pennyworth list --definitions-dir=example_dir

  • Starting a VM $pennyworth up my_machine

For more information about the commands, see the Pennyworth man page.

RSpec Helper

Pennyworth contains an RSpec helper that helps with running integration tests using Pennyworth.

To use the helper, first require it:

require "pennyworth/spec"

Replace <pennyworth-dir> with a directory into which you installed Pennyworth.

In your specs, you can now use the start_system method to start a VM, e.g.:

describe "my pet feature" do
  it "works flawlessly" do
    vm = start_system(box: "box")

    # ...
  end
end

The start_system method can either start an existing Vagrant box, a generic VM image runnable by libvirt or connect to an already running system. The method returns an object, which can be used to access the system for testing.

Using Vagrant VMs

To start a Vagrant VM, pass its name using the box option. The name is looked up in the Vagrantfile in the directory provided with the config.vagrant_dir option in RSpec.

Using generic VM images

To start a generic VM image, pass its file path using the image option. The image is ran in KVM and made available for accessing it for tests.

Using existing hosts

For connecting to an existing running system, pass the name of the system with the host option. The name is looked up in a configuration file, which by default is ~/.pennyworth/hosts.yaml. The system is accessed with the address stored in this file, so that the same name can be used in the tests, even when the actual system used for tests is changing.

To prevent tests running simultaneously on the same machine to interfere with each other, there is a locking mechanism, which automatically makes sure that only one test is running on a system at the same time. For this to work a lock service is required. The address of the lock service is also defined in the hosts.yaml configuration file. The lock service has to conform to the API implemented by glockd and has to run at the address specified in the configuration file.

To simplify setup of tests with existing hosts, there are some helper commands in the pennyworth command line application. Get an overview by running pennyworth host.

To initially set up the test infrastructure run

pennyworth host setup http://ci.example.org/pennyworth/hosts.yaml

with a URL corresponding to where you store the configuration of your test hosts. This will create a configuration file which includes the configuration from http://ci.example.com/pennyworth/hosts.yaml as a reference. The configuration will dynamically be fetched from the URL when tests are run. You can add local configuration to the ~/.pennyworth/hosts.yaml file. If it defines values for keys from the remote file it overwrites the value by the local attribute. This setup makes it easy to distribute a common configuration between several systems and users, while still allowing users to have their local settings.

Pennyworth automatically cleans up the hosts after running the tests using snapper snapshots. The snapshot which the system is rolled back to is specified by the base_snapshot_id of the host entries. Before rolling back a new snapshot named pennyworth is created which can be used for debugging test failures.

In some cases (e.g. while working on a test) the automatic rollback might not be desired. It can easily be disabled temporarily by setting the SKIP_HOST_CLEANUP environment variable, e.g.

SKIP_HOST_CLEANUP=true rspec

An example for the configuration file is:

---
lock_server_address: lockserver.example.org:9999
hosts:
  test_host_1:
    address: host1.example.org
    base_snapshot_id: 2
  test_host_2:
    address: host2.example.org
    base_snapshot_id: 34

In order to use a system as a pennyworth host it needs to be prepared like this:

  1. The root partition needs to be a Btrfs partition

  2. Snapper needs to be installed and configured for /:

    1. Enable snapper for root by running:

      snapper create-config /

    2. Remove more than 5 Pennyworth snapshots and disable the timeline for the new config:

      snapper set-config NUMBER_LIMIT=5 TIMELINE_CREATE=no

    3. Disable the automatic zypper snapshots by removing the plugin:

      zypper rm snapper-zypp-plugin

    For further details: Snapper Automatic Snapshot Documentation

  3. There can't be any subvolumes below / besides .snapshots

  4. It's usually helpful to exclude */.ssh from the rollback so that SSH access is retained

    echo "*/.ssh" > /etc/snapper/filters/ssh.txt

  5. There needs to be a snapper snapshot of the defined state which will be configured in the hosts.yaml:

    snapper create --description "Initial Snapshot"

    The according snapshot id can be retrieved using

    snapper list

Accessing test systems

Boxes, images and systems have the following requirements:

  • ssh port is configured to be open in the firewall
  • activated sshd service
  • the public ssh key of the user running pennyworth/rspec tests in /root/.ssh.authorized_keys

For boxes handled by pennyworth the ssh key is copied into the target when creating the box, for images or hosts this has to be done manually by e.g. running ssh-copy-id root@<HOST>.

The start_system method returns a VM instance, which can be used to interact with the running machine (via SSH). It supports the following methods:

  • stop

    Stops or disconnects the system. This stops running boxes or images and disconnects from running systems.

  • run_command(command, *args, options = {}) run_command(command_and_args, options = {})

    Executes a command on the running machine. A VM::CommandResult instance is returned which can be used with special RSpec matchers to define the expected behavior. Examples:

    Expect exit code 0 and no stderr

    expect(result).to succeed

    Expect exit code 0, ignoring any stderr

    expect(@vm.run_command("ls -l")).to succeed.with_or_without_stderr

    Expect a certain stdout

    expect(result).to succeed.and have_stdout(/foo.*bar/) expect(result).to succeed.and have_stdout("foo bar")

    Expect exit code 0, but some stderr output

    expect(result).to succeed.with_stderr

    Expect the stdout to include a certain string

    expect(result).to include_stdout("foo bar")

    Expect exit code != 0

    expect(result).to fail

    Expect a specific exit code

    expect(result).to fail.with_exit_code(15)

    Expect specific stderr

    expect(result).to have_stderr(/This.*error/) expect(result).to have_stderr("This is an error")

    Expect stderr to include a certain string

    expect(result).to include_stderr("Warning"

  • inject_file(source, destination)

    Injects a file from the machine running the specs into the VM.

  • inject_directory(source, destination, opts = {})

    Injects a file from the machine running the specs into the VM. The :owner and :group options can be used to set the owner and the group of injected files.

  • extract_file(source, destination)

    Extracts a file from the VM into the machine running the specs.

All machines started by start_system are stopped when the RSpec example group containing the call is finished.

Terminology

(Vagrant) Box : Base image used as a package format for Vagrant environments. Provides an identical working environment on any platform. For more information please visit the Vagrant documentation

VM : Virtual machine ran in KVM. Pennyworth supports running VMs described in a Vagrantfile as well as non vagrant managed ones.

Release Cycle

As for now we don't have a time based release cycle. We update and publish the gem as needed. However don't hesitate to report bugs or better yet submit fixes as pull requests. We will make sure to update the gem with the latest code.

Further information

Further information like a FAQ or a Troubleshooting guide can be found in the Pennyworth Wiki.

About

Test framework for running integration tests using virtual machines

Resources

License

Stars

Watchers

Forks

Packages

No packages published