July 3, 2018
It is a popular notion that the alternative to DevOps is essentially a debacle. You might be hazy in the matter of what precisely is the reason for this software engineering culture. Well, in layman terms, you have to maintain what you developed, or you have to make sure that the component you developed runs accurately from the development stack to the production stack, and all the stages in between.
So with a specific end goal to guarantee this runs smoothly, we adopt DevOps as a software engineering practice integrating both software development(Dev) and software operation(Ops).
Let us say you have a breakthrough and you begin a venture to exhibit your ideas, and it also meets the need as well as the liking of your clients. In the beginning, everything is simple; your codebase is compact, you need not bother with any environments for testing or presentation, everything runs on your PC, you probably do not even require a version control system.
This bodes well since you need to proceed promptly and ensure your idea is worth your investors’ money. In the event that you spend excessively on other things as opposed to implementing your idea in the initial stage then you stand a chance of spending your money on something that will be of no value in the future.
Time passes by, it seems people like your software, and you start adding people to your project. You realize you need a version control system to collaborate with your team members. You also need an integration server, where you integrate and test your software components written by you and your colleagues to avoid the problem: “This works on my computer.” You need at least one environment which hosts a stable version of your software and where you can do your presentations, later you will need a production environment.
Of course, you do testing which means you need another environment because you do not want to do testing in the same environment where you do presentations. You find your software is causing you troubles as it grows as well. You have more services collaborating with each other, different API versions, different data schema versions, and other issues.
You deploy one version of a service, and then you realize it is not compatible with another service which your service depends on and the environment stops working. Good, if it is a testing environment, bad if it happens during the presentation.
You keep hiring testers to avoid bug leakage to the clients, but at the same time, you are under pressure to deliver a new feature to the client to stay competitive. So you keep hiring developers as well. At some point you realize, you cannot solve all your problems by hiring more and more people. This makes your situation even worse.
Chaos leaks in, because the more people and the more services you have, the more errors there will be made by people configuring and running these services. Instead of developing new features, people start losing time by fixing broken services and firefighting.
Surprisingly the problem is not in the number of people you have in your company but in the way how people cooperate and communicate during the work processes. To avoid communication issues, DevOps promotes tight collaboration between developers and operators and even more, developers are made responsible for their services on any environment and are given all the necessary tools to configure and monitor them.
Deployments are made as easy as committing code to a version control system and pressing a button. All the tests are run automatically without human intervention and if some test fails the whole build is rejected. Errors are caught as early as possible in the delivery pipeline, so they do not propagate into presentation and production environments.
Rollbacks or undo’s should be implemented as well. If some service fails to deploy, it should be possible to rollback using just one button click. By giving developers all the necessary tools, operations people are relieved from routing work and can use their time more productively.
Two additional important principles of DevOps are infrastructure being as agile as possible and, infrastructure as code. It should be possible to create and delete services on the go. Virtualization makes it easy. Prior to virtualization, it took days or even weeks to develop computer networks, install and configure software. Usually, servers were configured manually which made them harder to maintain, because changes were not tracked by anyone.
This, in turn, led to the famous configuration mismatches between different environments, like the preproduction and production environments. Virtualization solved the time problem, while infrastructure as code solved the configuration mismatch problem. Whatever changes you make to your environments you do it through special tools, by first committing a description of how you would like to see your environment and then letting the tool do it for you. By having these configuration descriptions in the version control, it is possible for everyone to find out who and committed the changes when and which is the current configuration of the environment.
There are other DevOps principles which are not mentioned herein, but I hope the main idea was clearly illustrated to the reader. DevOps helps you focus more on your initial idea, on the value you provide to the customer instead of spending time on endless error fixing and fiddling around with infrastructure.