Releases are scheduled during sprint planning. Normally pencilled in for the end of a sprint, they can be set mid-sprint to give flexibility for specific deliverables.
Who is responsible for what?
If you’ve been working on a feature during the current sprint, once you’ve had it reviewed by a colleague andsigned-off as ready then you should merge it into the current develop branch before the release is prepared.
Once your feature has been merged into the develop branch you should then update the The Changelog to contain a one line description of your newly merged in but unreleased feature. This allows the person who will be doing the release later to know exactly what they are releasing. You should add this one line description under the Unreleased section of the The Changelog, for example:
Changelog --------- Unreleased Fixed issue with the users view v0.5.0: Added full name field for users.Note: *requires db migration*. Implemented the new designs for the sign-in form.
This unreleased section will then be renamed during the release to the release version number.
Then it’s the job of one person in the team to create and deploy the release. That person is arbitrarily nominated - eventually it will be you!
Creating the release
Creation of a new release is done using Git Flow. Essentially all it does is to merge the current develop branch into master and then tag the code at that specific point. This documents changes to the code and enables easy rollback later if need be.
We use major.minor.patch version numbering to ‘bump’ the releases. For example:
v0.0.1 - The first release
v0.0.2 - A minor change to fix that release, e.g. a config variable was missed
v0.1.0 - A new feature was introduced
v1.0.0 - A new feature was introduced that brought about a fundamental change in the app. Usually meaning that any API this app provides is now broken for clients expecting it to work in the old way.
This version is set during the git flow release process:
> git flow release start v0.0.1
The changelog is then updated with a high-level description of the changes, before finishing the release with Git Flow.
Release to staging first
It is definitely prudent to release production-ready code to a pre-production or staging environment first, to ensure that whatever the changes are doing, they happen without any nasty surprises. This way you might see issues with your db migrations or with updates to third-party libraries when the code is deployed, before the users do.
Often, a release involves a set of procedures - updating code, migrating a db, creating a few symlinks, etc. The list of procedures grows and changes as each project grows over time. To avoid complications, every project makes use of deployment wrappers like Fabric or Capistrano. These allow the configuration and arbitrary procedures to be managed centrally, with a common deployment pattern, making it possible for any developer to be able to deploy code competently.
The release checklist
- The features have been signed off.
- The features have been merged to develop.
- The tests are all green.
- The release has been created.
- The release has been tested in the staging environment.