The Criteria for Automation
Is something really worth automating, or will you spend more time on it than it's worth?
One of my favorite quotes is from Twitter user @zhuowei: "Never spend 6 minutes doing something by hand when you can spend 6 hours failing to automate it." Automation has become such a pervasive buzzword in the IT industry that people no longer ask whether something should be automated, only whether it can be.
Let me walk you through two instances where I automated a process. One delivered real value, and the other missed the mark.
Missing the Mark
In my homelab, I run three servers and I’m fairly particular about my terminal setup. I use a custom message of the day (MOTD) to keep things interesting, a concise and colorful prompt that shows the machine hostname and current directory, and a simple Docker configuration to make docker ps more readable.
After configuring my first server I spent a significant of time setting up this repository to house both the default and tailored customizations for each server I added. With this in place, a new server could be customized to my liking with just two commands.
git clone "https://github.com/chase-roohms/ServerConfig.git" "./.ServerConfig" bash "./.ServerConfig/configure"
The problem? I only ever set up three servers. I realized quickly that I did not need to scale past that, and it would've taken far less time to just copy the MOTDs and .bashrc files across those three servers manually.
Delivering Value
As a hobby, I create poster collections for movie and TV franchises, it's a refreshing break from technical work that lets me exercise my creative side. You can see some of my work on TPDb.
Once a collection is finished, I share it on Reddit in a clean grid layout that makes it easy for both myself and others to view the entire set at a glance. I've made over 900 posters this year, and plenty of them were manually edited into a grid using Photoshop.
Last month I spent about 20 minutes writing a Python script using the image processing library Pillow (PIL) that would generate the grids for me, and I have used it daily since. Putting together the grids in Photoshop used to take me 20 minutes or sometimes even longer, now it takes around 10 seconds. You can find the script here.
What's the Difference?
Frequency and Complexity. There are two kinds of processes that should be automated: those you perform frequently, and those that are easy to get wrong.
My server customization fell into neither category. It wasn’t something I did often, in fact, it only happened three times. It also wasn’t particularly error-prone, as it amounted to copying and pasting five to ten files.
My poster grids, on the other hand, are created often and are surprisingly easy to mess up. I used to spend around half an hour on each one, carefully aligning every poster so the edges were pixel-perfect.
Conclusion
Next time you consider automating a process, ask yourself two simple questions: How often am I really doing this, and how easy is it to do incorrectly? The answer may lead you to realize that its more efficient to keep doing it manually.