Configuring test splitting

Test splitting is the process of partitioning a test suite to run in parallel across multiple Buildkite agents. Buildkite maintains its open source Test Splitter (test-splitter) tool. This tool uses your Buildkite Test Analytics test suite data to intelligently partition tests throughout your test suite into multiple sets, such that each set of tests runs in parallel across your agents. This process is known as orchestration and results in a test plan, where a test plan defines which tests are run on which agents. Currently, the Test Splitter tool only supports RSpec.

Dependencies

The Test Splitter relies on execution timing data captured by the Buildkite test collectors from previous builds to partition your tests evenly across your agents. Therefore, you will need to configure the Ruby test collector for your test suite.

Installation

The latest version of test-splitter can be downloaded from GitHub for installation to your agent/s. Binaries are available for both Mac and Linux with 64-bit ARM and AMD architectures. Download the executable and make it available in your testing environment.

Using the test splitter

Once you have downloaded the test-splitter binary and it's executable in your Buildkite pipeline, you'll need to configure some additional environment variables for the test splitter to function. You can then update your pipeline step to call test-splitter instead of calling RSpec to run your tests.

Configure environment variables

The Test Splitter tool uses a number of predefined, mandatory, and optional environment variables.

Predefined environment variables

By default, the following predefined environment variables are available to your testing environment and do not need any further configuration. If, however, you use Docker or some other type of containerization tool to run your tests, and you wish to use these predefined environment variables in these tests, you may need to expose these environment variables to your containers.

BUILDKITE_BUILD_ID #

The UUID of the pipeline build. Test Splitter uses this UUID along with BUILDKITE_STEP_ID to uniquely identify the test plan.

BUILDKITE_JOB_ID #

The UUID of the job in the pipeline's build.

BUILDKITE_ORGANIZATION_SLUG #

The slug of your Buildkite organization.

BUILDKITE_PARALLEL_JOB #

The index number of a parallel job created from a parallel build step.

Ensure you configure parallelism in your pipeline definition. Learn more about parallel build steps in Concurrency and parallelism.

BUILDKITE_PARALLEL_JOB_COUNT #

The total number of parallel jobs created from a parallel build step.

Ensure you configure parallelism in your pipeline definition. Learn more about parallel build steps in Concurrency and parallelism.

BUILDKITE_STEP_ID #

The UUID of the step group in the pipeline build. Test Splitter uses this UUID along with BUILDKITE_BUILD_ID to uniquely identify the test plan.

Mandatory environment variables

The following mandatory environment variables must be set.

BUILDKITE_SPLITTER_API_ACCESS_TOKEN #

Buildkite API access token with read_suites, read_test_plan, and write_test_plan scopes. You can create an API access token from Personal Settings > API Access Tokens in the Buildkite interface.

BUILDKITE_SPLITTER_SUITE_SLUG #

The slug of your Buildkite Test Analytics test suite. You can find the suite slug in the url for your test suite.

For example, the slug for the url: https://buildkite.com/organizations/my-organization/analytics/suites/my-suite is my-suite.

Optional environment variables

The following optional environment variables can also be used to configure the Test Splitter's behavior.

BUILDKITE_SPLITTER_DEBUG_ENABLED #

Default: false

A flag to enable more verbose logging.

BUILDKITE_SPLITTER_RETRY_COUNT #

Default: 0

The number of retries permitted. Test Splitter runs the test command defined in BUILDKITE_SPLITTER_TEST_CMD, and retries only the failing tests for a maximum of BUILDKITE_SPLITTER_RETRY_COUNT times. For RSpec, the Test Splitter runs BUILDKITE_SPLITTER_TEST_CMD with --only-failures as the retry command.

BUILDKITE_SPLITTER_SPLIT_BY_EXAMPLE #

Default: false

A flag to enable split by example. When this option is true, the Test Splitter will split the execution of slow test files over multiple partitions.

BUILDKITE_SPLITTER_TEST_CMD #

Default: bundle exec rspec {{testExamples}}

The test command to run your tests. The Test Splitter will replace and populate the {{testExamples}} placeholder with the test plan.

BUILDKITE_SPLITTER_TEST_FILE_EXCLUDE_PATTERN #

Default:

The glob pattern to exclude certain test files or directories. The exclusion will be applied after discovering the test files using a pattern configured with BUILDKITE_SPLITTER_TEST_FILE_PATTERN.

This option accepts the pattern syntax supported by the zzglob library.

BUILDKITE_SPLITTER_TEST_FILE_PATTERN #

Default: spec/**/*_spec.rb

The glob pattern to discover test files. You can exclude certain test files or directories from the discovered test files using a pattern that can be configured with BUILDKITE_SPLITTER_TEST_FILE_EXCLUDE_PATTERN.

This option accepts the pattern syntax supported by the zzglob library.

Update the pipeline step

With the environment variables configured, you can now update your pipeline step to use test-splitter instead of running RSpec. The following example pipeline step demonstrates how to partition your test suite across 10 nodes.

pipeline.yml
steps:
  - name: "RSpec"
    command: ./test-splitter
    parallelism: 10
    env:
      BUILDKITE_SPLITTER_SUITE_SLUG: my-suite
      BUILDKITE_SPLITTER_API_ACCESS_TOKEN: your-secret-token