1

Feature toggles (or feature flags) enable me to deploy unfinished features without comprehensive manual testing of features toggled off. We can toggle the feature on for certain users, try it out, and then choose to toggle it off or on for all users. This enables more frequent integration as suggested by version control strategies such as Trunk-Based Development, and more frequent deployment of the trunk (master) branch into user's hands.

Sarbanes-Oxley rules require that all features that are in user's hands be fully tested.

So if I have a SOX-compliant system, how do I proceed with using feature toggles without having to do full manual testing of the feature toggled ON? It makes sense to do regression testing with the feature toggled OFF, but the most common need is to integrate and deploy good code but not incur the manual testing cost of the "on" state for a feature toggled off by default.

So any suggestions on ways to have Sox-compliant feature toggles without comprehensive manual testing of the toggled-off feature on every release?

1 Answers1

1

It depends on how you're managing your feature flags. Ideally, you're using a flagging system that allows you to target specific users or segments of users so that an untested feature isn't inadvertently enabled for all your users. Some feature management systems (like LaunchDarkly) allow you to target users any way you want.

For example, a common workflow (for teams who use feature flags) is to release a new feature under a flag that's only turned on for the product team that's working on that feature. So, even if the feature hasn't been tested, it's guaranteed that only the right users have access to that new feature.

This allows your product team to continue to work on the feature and test it in production (if that's appropriate). Of course, a full set of tests should cover that feature before it's released, but it's nice that product teams have a way to iterate quickly without the hangups of a process that didn't involve feature flags.

manalang
  • 111
  • 2