Originally, I thought I’d give a deep dive into the mechanics of some of our automated work flows at 2nd Watch, but I really think it’s best to start at the very beginning. We need to understand “why” we need to automate our delivery. In enterprise organizations, this delivery is usually setup in a very waterfall way. The artifact is handed off to another team to push to the different environments and to QA to test. Sometimes it works but usually not so much. In the “not so much” case, it’s handed back to DEV which interrupts their current work.
That back and forth between teams is known as waste in the Lean/Agile world. Also known as “Throwing it over the wall” or “Hand offs.” This is a primary aspect any Agile process intends to eliminate and what’s led to the DevOps movement.
Now DevOps is a loaded term, much like “Agile” and “SCRUM.” It has its ultimate meaning, but most companies go part way and then call it won. The changes that get the biggest positive effects are cultural, but many look at it and see the shiny “automation” as the point of it all. Automation helps, but the core of your automation should be driven by the culture of quality over all. Just keep that in mind as you read though this article, which is specifically about all that yummy automation.
There’s a Process
Baby steps here. You can’t have one without the other, so there are a series of things that need to happen before you make it to a fully automated process.
Before that though, we need to look at what a deployment is and what the components of a deployment are.
First and foremost, so long as you have your separate environments, development has no impact on the customer, therefore no impact to the business at large. There is, essentially, no risk while a feature is in development. However, the business assumes ALL the risk when there is a deployment. Once you cross that line, the customer will interact with it and either love it, hate it, or ignore it. From a development standpoint, you work to minimize that risk before you cross the deployment line – Different environments, testing, a release process etc. These are all things that can be automated, but only when that risk has been sufficiently minimized.
Step I: Automated testing
You can’t do CI or CD without testing, and it’s the logical first step. In order to help minimize the deployment risk, you should automate ALL of your testing. This will greatly increase your confidence that changes introduced will not impact the product in ways that you may not know about BEFORE you cross that risk point in deployment. The closer an error occurs to the time at which it’s implemented the better. Automated testing greatly reduces this gap by providing feedback to the implementer faster while being able to provide repeatable results.
Step II: Continuous Integration
Your next step is to automate your integration (and integration tests, right?), which should further provide you with confidence in the change that you and your peers have introduced. The smaller the gap between integrations (just as with testing), the better, as you’ll provide feedback to the implementers faster. This means you can operate on any problems while your changes are fresh in your mind. Utilizing multiple build strategies for the same product can help as well. For instance, running integration on a push to your scm (Source Control Management), as well as nightly builds.
Remember, this is shrinking that risk factor before deployment.
Step III: Continuous Deployment
With Continuous Deployment you take traditional Continuous Delivery a step further by automatically pushing the artifacts created by a Continuous Delivery process into production. This automation of deployments is the final step and is another important process in mitigating that risk for when you push to production. Deploying to each environment and then running that environment’s specific set of tests is your final check before you are able to say with confidence that a change did not introduce a fault. Remember, you can automate the environments as well by using infrastructure as code tooling around virtual technology (i.e. The Cloud).
Continuous Deployment is the ultimate goal, as a change introduced into the system triggers all of your confidence-building tools to minimize the risk to your customers once it’s been deployed to the production system. Automating it all not only improves the quality, but reduces the feedback to the implementer, increasing efficiency as well.
I hope that’s a good introduction! In our next post, we’ll take a more technical look at the tooling and automation we use in one of our work flows.
-Craig Monson, Sr Automation Architect (Primary)
-Lars Cromley, Director of Engineering (Editor)
-Ryan Kennedy, Principal Cloud Automation Architect (Editor)