If your job is to build and maintain the continuous delivery (CD) pipeline for your organization, do you feel you can’t keep up with your development counterparts in innovation? Everyone is raving about the new cool features in your company’s product. Who deserves the credit? People who wrote the code, right? It’s not that simple. Churning out new features requires a sound process and solid infrastructure. A well-designed CD pipeline is crucial to the developers’ success. It mitigates the inherent risks in a fast-paced, frequent iteration regime now prevalent in agile software shops. On the flip side, as software gets more and more complex and the appetite for a faster time to market grows, opportunities for innovation are in abundance for DevOps teams.
Multi-speed IT
So you’ve got two tactical teams with significantly different risk-reward equations, working at a different pace due to distinct objectives but marching toward the same strategic goal. You’re finding yourself in a multispeed IT environment. Development is focused on innovation and quick frequent iteration cycles. DevOps’ primary focus is to ensure the stability of the CD pipeline and keep a precise track record of what takes place inside the pipeline every step of the way. How do you make sure to stay ahead of the technological curve while minimizing risk?
Based on my experience on both sides of the DevOps isle, I recommend the following four steps to successfully build and maintain a continuous delivery pipeline:
1. Automate, Automate, Automate!
My personal motivations for rolling up my sleeves and scripting a task are:
- It became an annoying, mindless routine;
- It’s no longer a one-time thing and isn’t going away;
- My time can be spent on more important things; or
- I make mistakes and then waste time fixing them.
As with any type of code, organize the scripts into a sound logical structure easy to navigate and version-control it. You can always revise the code or completely rewrite it as the needs evolve.
2. Treat Automation Code as a First-Class Citizen
That’s right! As a practitioner, I learned the hard way that automation starts innocently small and simple. Then the requirements for more automated functionality begin to pile up rather quickly and things quickly become really messy. So write high-quality code from the get-go and reassess the design and implementation every step of the way. Remember that the visibility of your code is very high.
Apply the same standards to your code as your peer development teams do to theirs. Write tests—unit, integration, functional, end-to-end—as appropriate. Use the same CD pipeline to automatically build, package, deploy and test your automation code. This will give your pipeline another round of validation, your code will be of a much better quality, and your team’s credibility and reputation will soar.
3. Containerize the CD Pipeline Components
In the past, because the CD pipeline components such as build servers were manually provisioned, they often became the single points of failure, which made it hard to plan and roll out the modifications to homegrown, open source and commercial tools. It wasn’t unusual for some components to fall behind the latest release by a full year or more. In the meantime, a year’s worth of updates and patches could address a litany of issues, but the need for a maintenance window, usually on a weekend, and the uncertainty of the outcome often would outweigh the sought-after benefits. Containerizing the CD pipeline component can be the solution. Running a build server in a container, for example, will make it a lot easier to create an exact copy of the production version in a test environment and iron out the upgrade process more thoroughly ahead of the upgrade.
4. Test Commercial Tools
Commercial and open source projects test their software, of course. However, it could make a whole lot of sense for your organization to create your own integration, end-to-end and performance tests for the software you don’t write—especially if the software has a plug-in architecture like Jenkins or UrbanCode Deploy and you use community or homegrown plugins. Commercial tools typically don’t provide any guarantees that those plugins will work after the upgrade. Running your own automated tests that exercise the functionality of the tool in your environment and on your specific data gives the opportunity to validate updates as they become available and identify any regressions proactively.
The objectives of DevOps are not only to identify and eliminate the existing bottleneck in the organization’s continuous delivery process, but also to look out for any future potential bottlenecks and proactively address them before they become real issues.
So your team is DevOps. Your product is a CD pipeline. How do you continuously improve the CD pipeline for your product?
If want to learn more about how to effectively build and maintain your continuous delivery pipeline, please check out the newly released 2nd edition of the “Application Deployment and Release for Dummies” eBook.
About the author / Alex Seriy
Alex is an accomplished technologist and DevOps evangelist with over 25 years of experience under the belt. He is an automation enthusiast and usability guru, passionate about uncovering new use cases and innovating applications for platforms, frameworks and tools. At IBM, Alex is a subject matter expert in Continuous Integration, Delivery and Testing. He currently focuses on integration testing in the API Economy through service virtualization. Prior to joining IBM, Alex served as the Senior DevOps Architect at TIAA-CREF for 4 years. Before that, he was a consultant specializing in SCM, Release Engineering, QA and DevOps for over 10 years.