We’re in the process of scaffolding out our Microsoft Azure environments using Terraform. While Terraform does an excellent job creating ...Terraform OSS
For almost two years, the RIMdev team has worked exclusively in and around GitHub tools to manage and deploy our applications. While GitHub has been fun, it hasn’t been perfect. In 2016, we are addressing some of our project management processes to deliver features, bug fixes, and all around awesomeness to our users. This post describes what changes we are making to our process.
GitHub has always lacked organizational tools. The unit of work with GitHub is clearly the repository. Most teams work across multiple repositories. Gaining visibility across multiple repositories and projects was growing more difficult.
We decided to adopt ZenHub into our process as it gives us some much-needed features we intend to use:
The biggest advantage of ZenHub is it layers right on top of GitHub via a browser extension. The tight integration with GitHub means zero context switching for team members.
labels across repositories were inconsistent and were doing two things: describing the content of the issue, and also defining the status of our issue. Our new approach to labels is to describe the contents of the issue only. Pipelines, through ZenHub, better represents the state of the issue.
The labels we’ve agreed on are the following:
We used Windows Azure staging slots for deployments for that same period, and we came to the realization they were not a good fit for us.
We treat our
master branch as the truth. The branch is what is in production at the time, and if any pull request gets merged into that branch, then it should be production ready.
We are adopting a strict release cadence for each project currently under our responsibility. Right now we have chosen a 2-week release cycle. That means whatever issues are deemed “done” will be deployed to our production environment. This approach is also known as the release train.
We previously found that waiting for a completed milestone would keep us from deploying features, putting us behind. It also meant there was an opportunity for other features to creep in and increase the size of a deployment. Bigger deployments make us feel uncomfortable.
We are currently customizing our version of Hubot to serve the team, communicate important events, and perform tedious tasks. While Slack integrations are nice, and we do utilize them, we wanted that extra level of configuration.
As a development team, we are constantly reevaluating our process. We want to deliver on our promise of a great user experience and to do it quickly and efficiently. These alterations are probably not the last time we change our process, but we feel it moves us and the business we service in a positive direction.