Skip to content
This repository has been archived by the owner on Jan 19, 2021. It is now read-only.

Latest commit

 

History

History
584 lines (426 loc) · 22 KB

README.md

File metadata and controls

584 lines (426 loc) · 22 KB

⚠️⚠️ DEPRECATED ⚠️⚠️

This project, and it's related TestArmada projects, will no longer be supported. No further work from the owners will be done, and no PRs will be reviewed.

Magellan: Large-Scale Automated Testing

Build Status codecov Downloads

image

Magellan is a tool for massively-scaling your automated test suite, with added reliability. Run large test suites across many environments (multiple browsers or versions, or multiple native iOS or Android devices) at the same time, in parallel, with a friendly command-line workflow that is both local development and continuous-integration friendly. Magellan is compatible with mocha (wd.js, webdriver.io, appium) tests and Nightwatch.js tests ( example Nightwatch project ), and includes third party browser provider support such as SauceLabs. Through Magellan's mocha support, you can scale regular node.js test suites too.

Features

  • Parallel Test Runs
    • Worker allocation and management with failed test retry.
    • Network port management and testing, with isolated ports for mocking servers, individual (per-worker) selenium servers.
    • Configurable worker count.
    • Testing and debugging workflows. Run many tests at once, one test at a time, filter by tags, groups, etc.
    • Suite run control: Bail likely-failing suite runs early, or bail upon first failure.
    • Run many different parallel local browsers (eg: Chrome, Firefox, etc) all at the same time.
    • Run many different parallel remote (SauceLabs, Browserstack, etc.) browsers.
  • Integration Support
    • Status reporter API with events streamed from workers, with some included reporters.
    • Slack reporting support.
    • Admiral reporting support.
    • Plays well with CI (Jenkins, etc).
    • Runs test over the cloud like saucelabs via magellan executor (configurable and in parallel)
    • Can talk to a locks service to control saucelabs virtual machine usage (beta).

------------------BREAKING CHANGE in v10.0.0------------------

Magellan Executor

Executor is a mid layer between magellan and test framework to drive test run (via framework) based on a specific need (differentiated by executing environments). Magellan doesn't provide a default executor, so you need to pick at least one executor from the existing executor list, or implement one yourself.

What is an executor

  1. middle layer between magellan and test framework
  2. bridge to connect magellan and plugins

What can an executor do

  1. resolve profiles (env info, test info, capabilities for selenium test)
  2. patch setup and teardown event on the magellan test runner
  3. patch setup and teardown event on a magellan worker
  4. do some extra work in test's lifecycle
  5. communicate to a specific test env

Existing executors

Test Framework Compatibility and Installation

Magellan supports test frameworks like Mocha and Nightwatch via the usage of plugins.

Mocha

------------------BREAKING CHANGE in v10.0.0------------------

[email protected] doesn't support the mocha plugin for now. If you're using magellan version 9 or lower to run mocha test please don't upgrade. Or if you're seeking for mocha support please use magellan version 9 or lower.

All magellan mocha supports can be found here

Nightwatch

Plugin:

Boilerplate / example project:

Helper Library: (note: this is not required for nightwatch support)

Executor:

Installation:

npm install --save-dev testarmada-magellan
npm install --save-dev testarmada-magellan-nightwatch-plugin
npm install --save-dev testarmada-magellan-local-executor
npm install --save-dev testarmada-magellan-saucelabs-executor

magellan.json

{
  "framework": "testarmada-magellan-nightwatch-plugin",
  "executors": [
    "testarmada-magellan-local-executor",
    "testarmada-magellan-saucelabs-executor"
  ]
}

How Magellan Fits In

Magellan can best be described as a runner-runner. If you use mocha or nightwatch to run your current tests, then you can use Magellan to in turn run mocha or nightwatch and scale up your test runs.

When running tests with mocha, Magellan simply stacks on top of your existing suite:

mocha tests

magellan stack-mocha

mocha tests, with help from rowdy

Note: This is the preferred solution for working with mocha. If you want to spend less time handling selenium desiredCapabilities objects and starting/stopping Selenium, Magellan works much better with rowdy, which handles these tasks for you (for examples of this in action, see our example Mocha/wd project and example Mocha/webdriver.io project). That setup looks like this:

magellan stack-rowdy

Nightwatch.js tests

Magellan can also run Nightwatch.js test suites (please see our example Nightwatch project ), in which case your stack looks like this:

magellan stack-nightwatch

Plain node.js tests

Finally, Magellan can also scale regular ol' node.js tests (no browsers or devices) using Mocha:

magellan stack-nodejs

Example Developer Workflows

Magellan is a command line tool.

Note: The following examples assume you have ./node_modules/.bin in your PATH. If you don't have this rule in PATH, or are unable to add it, you can also run any of the examples below like this:

$ ./node_modules/.bin/magellan

Quick Reference Guide for Command-Line Use

Running Many Tests in Parallel (Default)

By default, magellan will try to run your test suite the fastest way possible, in parallel. Default is 3 workers, but this can be set using "--max_workers=n" command line param.

You can also run parallel tests on a real local browser (with magellan-local-executor):

# launch several instances of phantomjs at once and run tests in parallel
$ magellan --local_browser=phantomjs

# launch several instances of Firefox at once and run tests in parallel
$ magellan --local_browser=firefox

Testing in Multiple Browsers

magellan can run your test suite across multiple browsers with one command:

# Run tests locally in Chrome,phantomjs and Firefox
$ magellan --local_browsers=chrome,firefox,phantomjs

Controlling Which Tests Run

Tag Support in Nightwatch.js-based Tests

In Nightwatch-based tests, Magellan supports the standard Nightwatch convention of tags. To define a test that Magellan can match with --tags=commerce or --tags=smoke, we would write:

module.exports = new Test({

  tags: ["commerce", "smoke"],

  "Load the order page": function (client) {
    ...
  },

  "Submit the purchase": function (client) {
    ...
  }

});
Selecting Tags at Runtime

To filter by one or more tags, run magellan with the --tag or --tags option:

# Specify one tag:
$ magellan --tag=customer

# Specify multiple tags:
$ magellan --tags=customer,guest

To limit the your tests by a file path prefix, use the --group option:

$ magellan --group=tests/Smoke

The above filter will match tests/product/tests/Smoke*

Filter options can be combined together. For example, --tags and --group work together:

$ magellan --group=tests/Smoke --tags=customer,guest

To run one specific test, use the --test flag with a path to the test file:

$ magellan --test=path/to/my/test.js

To run tests defined in a json file as array, use the --testFile flag with a path to the test file (must use [email protected] or highter):

$ magellan --testFile=path/to/my/tests.json

Example file:

[
  "path/to/my/test1.js",
  "path/to/my/test2.js",
  "path/to/my/test3.js"
]

Running Tests One at a Time (Serial Mode)

You can run your entire suite serially (i.e. one at a time) and get live console output with the --serial option:

$ magellan --serial

Note that --serial can be combined with --tag, --group and --test filters, as well as different --browser options.

Early Test Run Termination

Test suites that have one or more failing tests can be terminated faster to conserve resource usage (i.e. in CI) or obtain results on failing tests faster (i.e. during a developer workflow).

To terminate a run early as soon as any test has failed, use --bail_fast:

$ magellan --bail_fast

To terminate a run early if 10% of total test runs have failed (and at least 10 test runs have occurred), use --bail_early.

$ magellan --bail_early

This option allows for a build to run for longer but still terminate if a significant portion of the suite is failing. Use this option if you want to avoid running a long test run, but want to be able to potentially spot trends in failing tests.

You can control how long Magellan waits for a test to execute before explicitly killing it by supplying a bail_time argument. For example, to set bail time to 60 seconds:

$ magellan --bail_early --bail_time=60000

A bail option does not have to be used to set bail time. For example:

$ magellan --bail_time=60000

The bail_time setting can also be written to Magellan configuration. See Saving Configuration

Saving Configuration

To use the same arguments on every magellan run, create a file called magellan.json and use the same arguments as magellan takes, but in JSON format. For example:

{
  "bail_fast": true,
  "max_workers": 5
}

To supply a path to a different magellan configuration file, use the --config argument:

$ magellan --config=./alternate_config.json

Custom Reporters

Magellan supports custom reporters. A custom reporter is a module that inherits from Magellan's base reporter class and can listen to a number of events and streams for monitoring test activity.

Adding Custom Reporters

To include a custom reporter, add it to magellan.json like this:

{
  "bail_fast": true,
  "max_workers": 5,
  "reporters": [
    "./path/to/my/reporter",
    "my_reporter_module"
  ]
}

In the example above, ./path/to/my/reporter refers to a module in your test suite directory tree, whereas my_reporter_module is a module you have included in package.json

Example Custom Reporter

var BaseReporter = require("testarmada-magellan").Reporter;
var util = require("util");

var Reporter = function () {
};

util.inherits(Reporter, BaseReporter);

Reporter.prototype.listenTo = function (testRun, test, source) {
  // Stream stdout and stderr directly to stdout, assuming this source is
  // a process that has those properties.
  if (source.stdout) {
    source.stdout.pipe(process.stdout);
  }
  if (source.stderr) {
    source.stderr.pipe(process.stderr);
  }
};

module.exports = Reporter;

When magellan is preparing a test run, it will call your reporter's implementation of listenTo(). Arguments are as follows:

  • source is a node event emitter that optionally has properties stdout and stderr.
  • test is the actual test being run. This contains information about how long the test has been running or already took to run, as well as how many attempts have been made on this test already.
  • testRun is an object with information about the specific test run this source is associated with. It has the following properties:
    • locator: An object describing the identity of the test, whether it be filename or name (note: the uniqueness of the test is often a composition of locator properties, because a test with a given name might live in a given file, and that name might not uniquely identify it across an entire suite). This object is guaranteed to have a toString() function which returns a human readable representation of the identity of the test, but is not guaranteed to have any other properties.
    • buildId: The id of the whole suite run.
    • tempAssetPath: The path to place various temporary assets (logs, screenshots, configuration files, etc) generated by or for a specific test run.

Reporter Events

At the moment, there is one event available, worker-status, which indicates whether a Magellan worker has started executing a test or finished executing a test. The example below illustrates how to unpack this event:

source.on("message", function (msg) {
  if (msg.type === "worker-status") {
    if (msg.status === "started") {
      console.log("Test started: " + msg.name);
    } else if (msg.status === "finished") {
      console.log("Test finished: " + msg.name);
      console.log("    Pass/fail: " + msg.passed);
    }
  }
});

Asynchronous Reporter Initialization

Some custom reporters need to initialize themselves asynchronously before listenTo can be called. To do this, the a reporter class can define an initialize() method, and return a Q promise:

  initialize: function () {
    var self = this;
    var deferred = Q.defer();

    console.log("Initializing reporter..");

    createJob(function(err, jobId) {
      if (err) {
        deferred.reject(err);
      } else {
        self.jobId = jobId;
        deferred.resolve();
      }
    });

    return deferred.promise;
  },

In the above example, a reporter needs to create a job entry and obtain the id of the job before it can send information about running tests. Magellan will wait until initialize() resolves this promise before starting any tests.

Optional Reporter Loading

Your project may have reporter modules that you only want to run in certain circumstances, like a CI environment, and you may not want to require them to be installed on a given system (eg: a developer machine). If you want to reference a reporter but still run tests if that reporter is not installed or not found, use optional_reporters in magellan.json:

{
  "reporters": [
    "./path/to/my/reporter",
  ],
  "optional_reporters": [
    "my_optional_reporter_module"
  ]
}

Setup and Teardown

If you need to run setup and teardown tasks before and after your test suite runs, respectively, it's recommended to write npm tasks for that purpose and simply sandwich your magellan call between them.

Sometimes, however, you need a setup and teardown process that constructs state within Node and then cleans up afterwards. For this use case, Magellan also supports a global setup and teardown hook that will run before and after your suite. To register a setup and teardown, add a path to your magellan.json:

{
  "setup_teardown": "path/to/setup_teardown.js",
}

Magellan will require() these modules and run them as functions that should call a callback. Setup and teardown modules should look like this:

var Q = require("q");

var SetupTeardown = function () {
};

SetupTeardown.prototype = {
  initialize: function () {
    var deferred = Q.defer();

    // do asynchronous setup stuff here. Resolve (or reject) promise when ready.
    doAsyncStuff(function (){
      deferred.resolve();
    });

    return deferred.promise;
  },

  flush: function () {
    var deferred = Q.defer();

    // do asynchronous teardown stuff here. Resolve (or reject) promise when ready.
    doAsyncStuff(function (){
      deferred.resolve();
    });

    return deferred.promise;

  }
};

module.exports = SetupTeardown;

Note: Magellan should ensure that flush() always runs even if your test suite fails.

Local Browser Profiles

Sometimes it's useful to specify a list of environments with differing resolutions or orientations. For this case, Magellan supports profiles stored in magellan.json:

{
  "profiles": {
    "my_profile_1": [
      { "browser": "chrome_42_Windows_2012_R2_Desktop", "resolution": "2560x1600" },
      { "browser": "chrome_42_Windows_2012_R2_Desktop", "resolution": "800x600" },
      { "browser": "ipad_8_2_iOS_iPad_Simulator", "orientation": "landscape" },
      { "browser": "ipad_8_2_iOS_iPad_Simulator", "orientation": "portrait" },
      { "browser": "iphone_8_2_iOS_iPhone_Simulator", "orientation": "portrait" }
    ],
    "tier_2_browsers": [
      { "browser": "safari_7_OS_X_10_9_Desktop" },
      { "browser": "IE_8_Windows_2008_Desktop" },
      { "browser": "IE_9_Windows_2008_Desktop" }
    ]
  }

Notice that resolution and orientation are optional. Multiple definitions of the same browser running in different resolutions or orientations are permitted. To select these profiles, we call:

$ magellan --profile=tier_2_browsers
# or
$ magellan --profile=my_profile_1
# or specify multiple profiles
$ magellan --profile=tier_1_browsers,tier_2_browsers

Hosted Browser Profiles

If you're using profiles to reflect browser tiers in a large organization, you may wish to centralize your profiles on a web server somewhere and have magellan load them remotely. To do this, specify a profile URL to source profiles from, and the profile you want to use after a # (hash):

$ magellan --profile=http://example.com/testing/browser_profiles.json#tier2

Multiple profiles can be specified the same way, comma delimited:

$ magellan --profile=http://example.com/testing/browser_profiles.json#tier1,tier2

Where browser_profiles.json should have a structure similar to placing profiles{} in magellan.json:

{
  "profiles": {
    "tier_2": [
      { "browser": "safari_7_OS_X_10_9_Desktop" },
      { "browser": "IE_8_Windows_2008_Desktop" },
      { "browser": "IE_9_Windows_2008_Desktop" },
      { "browser": "IE_10_Windows_2008_Desktop" }
    ]
  }

Magellan's Strategies

Since 10.1.0 magellan supports strategies. Strategy is a rule which tells magellan when to do what. There are two strategies that magellan allows for now

Bail strategy

Bail strategy is a rule which tells magellan when to fail the whole test suite when there are certain failures in your test run, you can tell magellan to terminate your test run early via a certain bail strategy.

Current �supported bail strategies: magellan-early-bail-strategy and magellan-fast-bail-strategy.

Please refer to the readme of each repo for more details.

Resource strategy

Resource strategy tells magellan what to do if required resources are not available for the test.

Setting Up Setup and Teardown Tasks for CI

If you have setup and teardown tasks that need to run before and after magellan is called (for compilation, standing up a mocking server, cleanup, etc), then you can expose the following npm task flow using the scripts example below:

$ npm run magellan:setup
$ magellan
$ npm run magellan:teardown

and:

$ npm run magellan:setup
$ magellan --sauce
$ npm run magellan:teardown

Here's an example scripts block that implements these tasks in your package.json:

  "scripts": {
    "magellan:setup": "./path/to/my/setup.sh",
    "magellan:teardown": "./path/to/my/teardown.sh"
  },

Using Magellan's Port Acquisition and Scanning Facilities

If you need to be able to consume ports during a magellan run and want to be absolutely certain they won't conflict with those that magellan gives to its worker pool, you can use the portUtils submodule like this in your setup and teardown tasks:

var magellan = require("testarmada-magellan");

// calls callback with arguments [null, 12000] if successful
magellan.portUtil.acquirePort(function(err, port) {
  if (err) {
    console.log("error!", err);
  } else {
    console.log("Got port " + port + " for safe usage.");
  }
});

License

Documentation in this project is licensed under Creative Commons Attribution 4.0 International License. Full details available at https://creativecommons.org/licenses/by/4.0