chaotic_job: Resilience test helpers for Active Job

ChaoticJob provides a set of tools to help you test the reliability and resilience of your Active Jobs. It does this by allowing you to simulate various types of failures and glitches that can occur in a production environment, inspired by the principles of chaos testing and deterministic simulation testing

Install the gem and add to the application's Gemfile by executing:

If bundler is not being used to manage dependencies, install the gem by executing:

ChaoticJob should be used primarily by including its helpers into your Active Job tests:

The ChaoticJob::Helpers module provides 6 methods, 4 of which simply allow you to perform a job with retries in the proper way while the other 2 allow you to simulate failures and glitches.

A central concept in ChaoticJob is the glitch. A glitch is an error injected into the job execution flow via a TracePoint. Glitches are transient errors, which means they occur once and only once, making them perfect for testing a job's resilience to unpredictable failures that can occur while running jobs, like network issues, upstream API outages, rate limits, or infrastructure failure. By default, ChaoticJob raises a custom error defined by the gem (ChaoticJob::RetryableError), which the internals of the gem ensure that the job under test is configured to retry on; you can, however, raise specific errors as needed when setting up your scenarios. By forcing a retry via the error handling mechanisms of Active Job, glitches are a simple but effective way to test that your job is resilient to any kind of transient error that the job is configured to retry on.

When testing job resilience, you will necessarily be testing how a job behaves when it retries. Unfortunately, the helpers provided by ActiveJob::TestHelper are tailored to testing the job's behavior on the first attempt.

Specifically, when you want to perform a job and all of its retries, you would naturally reach for the perform_enqueued_jobs method.

But, this method does not behave as you would expect. Functionally, it overwrites the enqueue method to immediately perform the job, which means that instead of your job being performed in waves, the retry is performed within the execution of the original job. This both confuses the logs and means the behavior in your tests are not representative of the behavior in production.

In order to properly test job retries, you should use the perform_all_jobs method provided by ChaoticJob::Helpers:

This helper will perform the job and all of its retries in the proper way, in waves, just like it would in production.

If you need more control over which batches of jobs are performed, you can use the perform_all_jobs_before and perform_all_jobs_after methods. These are particularly useful if you need to test the behavior of a job that schedules another job. You can use these methods to perform only the original job and its retries, assert the state of the system, and then perform the scheduled job and its retries.

You can pass either a Time object or an ActiveSupport::Duration object to these methods. And, to make the code as readable as possible, the perform_all_jobs_before is also aliased as the perform_all_jobs_within method. This allows you to write the example above as perform_all_jobs_within(4.seconds).

The helper methods for correctly performing jobs and their retries are useful, but they are not the primary reason for using ChaoticJob. The real power of this gem comes from its ability to simulate failures and glitches.

The first helper you can use is the run_scenario method. A scenario is simply a set of glitches that will be injected into the specified code once. Here is an example:

In this example, the job being tested is defined within the test case. You can, of course, also test jobs defined in your application. The key detail is the glitch keyword argument.

A "glitch" is describes precisely where you would like the failure to occur. The description is composed first of the kind of glitch, which can be either before_line, before_call, or before_return. These refer to the three kinds of TracePoint events that the gem hooks into. The second element is the key for the code that will be affected by the glitch. This key is a specially formatted string that defines the specific bit of code that the glitch should be inserted before. The different kinds of glitches are identified by different kinds of keys:

As you can see, the before_call and before_return keys are formatted the same, and can identify any instance (#) or class (.) method.

What the example scenario above does is inject a glitch before the step_3 method is called, here:

If we wanted to inject a glitch right before the step_3 method finishes, we could define the glitch as a before_return, like this:

and it would inject the transient error right here:

Finally, if you need to inject a glitch right before a particular line of code is executed that is neither a method call nor a method return, you can use the before_line key, like this:

Scenario testing is useful to test the behavior of a job under a specific set of conditions. But, if you want to test the behavior of a job under a variety of conditions, you can use the test_simulation method. Instead of running a single scenario, a simulation will define a full set of possible error scenarios for your job as individual test cases.

More specifically, it will create a scenario injecting a glitch before every line of code executed in your job. So, in this example, the simulation will run 12 scenarios:

It generates all possible glitch scenarios by performing your job once with a TracePoint that captures every event executed as a part of your job running. The block that you pass to test_simulation will be called for each scenario, allowing you to make assertions about the behavior of your job under all scenarios.

If you want to have the simulation run against a larger collection of scenarios, you can capture a custom callstack using the ChaoticJob::Tracer class and pass it to the test_simulation method as the callstack parameter. A Tracer is initialized with a block that determines which TracePoint events to collect. You then call capture with a block that defines the code to be traced. The default Simulation tracer collects all events for the passed job and then traces the job execution, essentially like this:

To capture, for example, a custom callstack that includes all events within your application, you can use the ChaoticJob::Tracer class as follows:

If you passed this callstack to your simulation, it would test what happens to your job whenever a transient glitch is injected anywhere in your application code called as a part of executing the job under test.

Remember, in your application tests, you will want to make assertions about the side-effects that your job performs, asserting that they are correctly idempotent (only occur once) and result in the correct state.

After checking out the repo, run bin/setup to install dependencies. Then, run rake test to run the tests. You can also run bin/console for an interactive prompt that will allow you to experiment.

To install this gem onto your local machine, run bundle exec rake install. To release a new version, update the version number in version.rb, and then run bundle exec rake release, which will create a git tag for the version, push git commits and the created tag, and push the .gem file to rubygems.org.

Bug reports and pull requests are welcome on GitHub at https://github.com/fractaledmind/chaotic_job. This project is intended to be a safe, welcoming space for collaboration, and contributors are expected to adhere to the code of conduct.

The gem is available as open source under the terms of the MIT License.

Everyone interacting in the ChaoticJob project's codebases, issue trackers, chat rooms and mailing lists is expected to follow the code of conduct.