dovydas.xyz

Allow Developers To Deploy Straight To Production

In times when continous integration is a new fad there still exists legacy systems which are not suitable for this flexible way of applying changes. This means that your company will be stuck in the eternal purgatory of developers and devops working together on task that developer alone could of done. Not allowing for your developer to apply changes for himself is a stupid methodology.

It hinders personal growth

Learning is a process of making a mistake and learning from it. True lessons are learned when you have to think on your feet while being spammed with 10 emails per second about your feature, that broke whole production . And you know what “hot learning” is a nice thing. Because your developer will be the person who will fix the problem faster than anyone else, because he most likely formed an intuition on parts that may break. Do not outsource the most valuable learning resource – mistakes.

If you compartmentalize duties of your organization like cheap Arduino modules from eBay you actively create an organization of people who are overspecialized thus leading the whole company not having a holistic view on their own system, because your employees are focused on one small part and not whole project in general. This means that if something goes wrong there will be only one or two people who can fix something after a change has been deployed, but what happens if those two super employees are on a vacation? Some third superman will most likely emerge naturally so why not allow everyone to be natural from their first feature? Fires are best handled in houses of firefighters.

Bugs will are inevitable. Furthermore different data yields different bugs. This is more than true with Enterprise Resourse Planning (ERP) systems. From my own experience and shared experiences of many other ERP developers most of them seem to arrive to the same two conclusions: 1. clients manage to find complex bugs first, 2. data creates bugs no matter how thorough you are at hunting them. This is normal, because systems with such complexity can’t be (and shouldn’t be) understood by a single developer. So if developer deploys the code and some bugs emerge it will most likely be a bug that no one even thought about. Just get on with it, since you already trust a developer to write a piece of code not trusting him to deploy it seems like more dichotomy rather than a safety measure.

Cost per change is much higher

Some big companies need whole departments just to release software, this in an overkill. If you would look at majority of operation departments you would most likely see some old, crusty and manual deployment rituals that could be made more modern without “risking” the so required system stability. Locate parts in your system that require most manual work and replace them with automation tools or frameworks like: Jenkins, Ansible, Puppet, Fabric, Kubernetes…

Additionally, if some other department (or a overspecialized employee) is doing your deployment it creates an unpleasant dependency on other people no matter how streamlined your deployment process is, there will to be cases where developer will have to contact the operations and ask for them to do something “not found in the recipe book”, this often creates long threads of emails and answers saying I am not authorized to authorize you should ask authorization from your superior authority. Long threads of emails!

Finally, organizational modularity increases complexity due to lack of cross domain knowledge. Operations know very little about programming, while developer knows very little about deployment. This makes communication for these to parties hard from time to time, thus increasing complexity of deployment process.

Changing something takes much longer

A clunky deployment ritual will lead to releases. Releases depending on the project will occur daily, weekly or even monthly. This means that code developer wrote will be stuck in some sort of “abyss”, where it is not used nor tested if it was a big feature and release took one month and bugs were spotted in production, good look expecting that this bug will be fixed quickly, because it is very likely that dev who wrote the code barely remembers what that feature and assumptions he had during development of that feature. This is super unproductive. Clunky releases create another issue it realeases many features made by many developers this means that if something breaks (after a release) there will be name calling and finger pointing because no one wants to fix bugs since they are not entirely sure it is a bug they made. To many many cooks spoil the broth.

Okay, your “bug free” code has been peer reviewed all unit and function tests passed testing department tested your feature and says it can be released, this code gets released and BOOM critical bug in production emerges. Now the whole release gets reverted and you as a developer have to squash that darn thing. Just because a bug was critical does not mean it can’t be fixed in few minutes. But since your organization adheres to the dogma of releases you have to be put on hold and be forced to wait a day, week, month until your patch will be deployed again. This is also super unproductive, compared to developer deploying code on his own. By deploying code by yourself you most likely would of fixed it and deployed it again almost instantly and if you saw it is a hard case you would make a final decision whether this feature should be reverted or not. If a person is capable of making decision on his own don’t outsource that to another department, because your company and the developer will lose valuable time and energy.

Also, imagine a case where there let’s say 3 uncaught bugs appear in the release. This means that in order to fix everything developer on worst case scenario would have to wait 3 whole months just to fix 3 simple bugs he would of fixed quickly on his own without relying on operations to deploy it.

If things go wrong someone will most likely be bombarded with emails. Emails from multiple people saying the same thing: “Feature X broke after most recent release” this only distracts everyone involved in the process, because someone has to respond to these emails and if a developer is responding then he is wasting his valuable time rather than fixing a bug he made. This in turn lengthens the whole release process.

Conclusion

More individual responsibility for the developer yields greater individual growth and more productive organization in general. Allowing a developer to deploy his own code reduces number of employees needed to release software. Developer will be significantly faster when fixing and releasing his code if he is given an ability to deploy to production since bugs are inevitable.

If you work in a company who has an old and stubborn way of releasing software begin by trying defuse the barrier between developers and operations, let them talk and decide how to give more freedom & responsibility to person actually writing the code.

return 0

Proudly representing Lithuania 🇱🇹 as dovydasgulbinas