A big goal of DevOps is to increase productivity. Yet many organizations still struggle with the efficiency of their software release processes. In fact, 54% of engineers stated they have too many manual steps within their deployment process, according to an Armory survey. And with the pervasiveness of multi-cloud, developers now face an escalating number of deployment environments, each carrying their own nuances.
Development teams want to move as quickly as possible without barriers. Yet, there are often hurdles to seamless deployments. For example, the code might be 100% ready to deploy, yet someone still must validate it before it reaches production. Manual bottlenecks like this in the CI/CD process can make developers feel like their code isn’t worthy of deploying and may even amplify feelings of burnout.
I recently met with Anna Daugherty, senior product marketing manager at Armory, to see what technology leaders can do to remove barriers within their software release cycles and increase productivity. According to Daugherty, many of these strategies boil down to having empathy for developers and their needs. “Developers aren’t machines; they’re people, and people deserve to have a quality of life when they’re doing their jobs.” Below, we’ll highlight tips and tricks to keep in mind to improve release fluidity.
1. Identify Where Your Manual Checks Are
The first step to speeding up CI/CD is identifying where manual steps exist within the development workflow. The DevOps life cycle, often visually depicted as an infinity loop, includes many states, such as design, build, test, deploy, assess and monitor. But if each step has one or more manual steps, you accrue a lot of manual debt, said Daugherty.
Therefore, engineering leads should seek to remove manual checks whenever possible. Of course, this may not be realistic within certain situations, such as financial institutions that require manual approvals at the last mile. However, everything should be smoothed out and automated up to that point, said Daughtery.
2. Shift From Continuous Delivery to Deployment
Depending on who you ask, the ‘CD’ in ‘CI/CD’ may stand for continuous delivery or continuous deployment. As part of reducing friction in the software release process, Daughtery encourages teams to begin to embrace the latter whenever possible. If deployment remains inhibited by unnecessary manual steps, it can decrease scalability and interoperability.
3. Increase Confidence With Testing
Of course, if we introduce more automation into the CI/CD pipeline, we must ensure the software is rigorously tested for errors and vulnerabilities. Daugherty encourages development to embrace automated, continuous testing approaches at every juncture, whether pre-production, inter-production, or throughout the application lifecycle. There are also advanced deployment strategies to consider, such as canary testing, which rolls out new changes only to a small subset of users or blue-green deployment can deploy to different places at different times.
These deployment strategies can help build confidence while reducing the risk of disrupting a majority of your user base with untested production updates. “The more you can integrate testing into your process, the more confident your process will become,” said Daughtery.
4. Chose The Right Toolsets
So, how do we implement more automation into the DevOps cycle? And what sort of tools should we look for? According to Daugherty, when scouting for DevOps tools, technology leaders should prioritize open solutions that aren’t closed-off environments. Instead, they should offer flexibility and the ability to integrate with the tools you already use every day. “You can have all the tools in the world, but if they don’t fit together, that can drag down the process.”
5. Consider Open Source Software (OSS)
On that note, consider open source software (OSS) where possible. OSS like Argo or Spinnaker has the potential to empower developers and bridge gaps in the CI/CD process. Other projects from the Cloud-Native Cloud Native Computing Foundation (CNCF) and Continuous Delivery Foundation (CDF) are mature with proven stable adoptions.
And although there is some weariness around OSS, particularly around unknown vulnerabilities and project maintenance concerns, “you can still be successful, secure, and scalable with open source,” said Daugherty. Open source also delivers community benefits and helps engineers feel like they are part of a bigger picture. “[OSS] is the opposite of burnout. It brings people together. It’s a community builder.”
6. Become Developer-First
Transitioning into a developer-first organization is another way to increase developer satisfaction and efficiency. But to become a developer-first organization, developers will need more say in the processes—this will involve balancing top-down requirements with total developer control. For example, leaders should suggest tools to enhance productivity but accept feedback on an ongoing basis to see what works best.
To support these efforts, the concept of developer experience (DX) has entered the lexicon in recent years. Building with developer empathy can result in more usable, frictionless software that increases productivity. DX is crucial within the burgeoning area of platform engineering, as the goal is to create reusable, internal self-service platforms for developers to leverage.
Interestingly, Netflix extolls what it calls full-cycle development, in which a developer oversees the building and operational aspects of a single component throughout its entire life cycle. “It’s an interesting way of addressing burnout because you feel more connected to what you create,” said Daugherty.
Final Thoughts: DevOps Isn’t Dead
Above, we’ve highlighted some potential strategies to increase developer productivity. By replacing manual bottlenecks with more automation, you can increase deployment fluidity. Considering how tools are acquired and iterating upon feedback from engineers can also help improve consistency while retaining quality talent.
Looking to the future, Daughtery hopes that similar roles such as DevOps, site reliability engineering and platform engineering can come together and join forces. Because, at the end of the day, they share many goals. And contrary to some grabby headlines, DevOps is far from becoming obsolete. DevOps is not dead. It’s very much alive and increasingly important.
