Spike on feature flags

As a developer I want to be able to keep my branches short lived so that I may achieve continuous integration and be ready to deliver a working system even when my feature work is still in-progress and un-complete.

To that end we'd like to utilize feature flags:

• Martin Fowler: https://martinfowler.com/articles/feature-toggles.html

• For trunk based development: https://trunkbaseddevelopment.com/feature-flags/

AC:

• Become familiar with trunk based development and the use of feature flags

• Hold a call with to hold a brainstorming and Q&A discussion and prepare for the technical forum discussion

• Spike on a code-sample of setting and using a feature flag, ideas:

  • Improve performance of SoH calc (part of ). Specifically do + and - calculations in db and avoid network IO by using Postgres SUM() and moving the reason + or - into table.

  • Fix a small bug or tech-debt - ideally one that involves the Database, the API. Extra points for the UI.

• Start a discussion on the technical forum and cover:

  • Does a feature flag to just be inside a service? Cross-service? Cross-project (e.g. reporting stack)?

  • Do we consider a feature flag to be tech-debt? How long should a feature flag exist?

  • How do feature flags work with long term storage (i.e. databases)?

  • How do the use-cases we discuss and the decisions above support better code quality and continuous delivery?

• Write up:

  • Guidelines for using feature flags

  • In conventions

  • In contribution guide (http://docs.openlmis.org/en/latest/contribute/contributionGuide.html#git-branching-pull-requests)

  • In how to deploy

• Present to technical committee how a developer uses these.