This document codifies the best practice for interacting with test flakes on Fuchsia.
Background: What is a flaky test?
A flaky test is a test that sometimes passes and sometimes fails, when run using the exact same revision of the code.
Flaky tests are bad because they:
- Risk letting real bugs slip past our commit queue (CQ) infrastructure.
- Devalue otherwise useful tests.
- Increase the failure rate of CQ, thereby increasing latency for modifying code.
This document is specific to test flakes, not infrastructure flakes.
Requirements: Goals for flaky tests
- Flakes should be removed from the critical path of CQ as quickly as possible.
- Since flakes present themselves as a failing test, flakes should not be ignored once taken off of CQ. They represent a real problem that should be fixed.
- Tests may flake at any time, and as a consequence, the observer of these bugs may not necessarily be the person best equipped to fix it. The process for reporting bugs must be fast, easy, and decoupled from diagnosis and patching.
The following provides the expected & recommended lifetime of a flake:
- A test flakes in CI or CQ.
- Identify: The test is automatically identified as a flake.
- Track: An issue is automatically filed for the identified flake under the Flake component.
- Remove: The test is removed from CQ immediately.
- Fix: The offending test is fixed offline and re-enabled.
A flake fetching tool is currently in use to identify the vast majority of flakes.
The tool looks for test failures in CQ where the same test succeeded when retried on the same patch set.
One should prioritize, above all else, removing the test from the commit queue. This can be achieved in the following ways:
- If the flake has been prompted by a recent patch: Submitting a revert of a patch which triggers this flake.
- Disable the test.
The above mechanisms are recommended because they remove the flaky test and prevent the commit queue from becoming unreliable. The first option (reverting code) is preferred, but it is not as easy as the second option (disabling test), which reduces test coverage. Importantly, neither of these options prevent diagnosis and fixing of the flake, but they allow it to be processed offline.
It is not recommended to attempt to fix the test without first removing it from CQ. This causes CQ to be unreliable for all other contributors, which allows additional flakes to compound in the codebase.
At this point, one can take the filed issue, locally re-enable the test, and work on reproducing the failure. This will enable them to find the root cause, and fix the issue. Once the issue has been fixed, the bug can be closed, and the test can be re-enabled. If any reverted patches need to re-land, they can re-land safely.
Improvements and Tooling
Ongoing efforts to improve tooling surrounding flakes are actively underway.
- Automatically assigning issues for resolving flakes, based on information present in OWNERs files.
- "Deflaking" infrastructure, to re-run tests in high volume before they are committed.
As improvements are made, this document will be updated with the latest policy.
Except as otherwise noted, the content of this page is licensed under the Creative Commons Attribution 4.0 License, and code samples are licensed under the Apache 2.0 License. For details, see the Google Developers Site Policies. Java is a registered trademark of Oracle and/or its affiliates.
Last updated 2019-12-12.