Working hard to be lazy

So there's a great writer, one of my favorites, UKLG. She's completely not lazy, so apologies to her for inclusion. She's here, though, because she invented something great: The ansible 🌌. And why is that relevant? We'll get there…

As any smart human person in charge of configuring and managing computers knows, I'm bad at my job. Or rather, my job is nearly impossible to be good at, given my human limitations. Why?

There are way too many things to keep track of to be good wrangling a computer, let alone more than one of them 🖥 𝔛 ⧝!

That's why smart humans invented ✨⚡️AUTOMATION⚡️✨.

Automation means: you keep track of all the random changes your make to your myriad .dotfiles in a central, version controlled place. You can run a convergent install.sh script that puts every alias, .ignore, .gdbinit, package, etc. you use on your computer, sets it all up, clones the repos you work on and curls and installs the things not easily package-managed, symlinks applicable files (.bashrc, etc.), and prints out instructions (that you keep up to date) of manual steps you must execute. You make the effort to improve this script and configuration by re-imaging your computer often and re-running it. You re-run it when its not fresh to test it's convergence. This way you get to be lazy when you set up a new laptop or workstation; you simply git clone dotfiles && cd $_ && ./install.sh. Nothing to remember, not much work to do.

Automation also means, and here we return to our opening image, making the effort to perform and document server setup via code. I happen to like Ansible for this, for a few reasons: its relatively lightweight, the manifest format is fairly straightforward, it doesn't require an agent to be installed on the machine being managed, and it works fairly hard to be idempotent[1]

Security, for example: There's an article kicking around about the first few minutes on a server, which is a great reference for a simple security baseline: lock things down, turn on updates, etc. I figured, why not do one better and automate this, since I run these same steps on basically every server I boot up (and should do on all of them, which automation makes easy!). So now I have a manifest on my github that can do this again, and I don't have to remember the exact steps. Again, I get to be lazy. I don't really have to look things up. I just know that I did at one point, codified that knowledge in some code that does the thing, and know where that code lives. Some documentation might be nice, because things break, but in general I expect that using the script over time will shake out bugs and make it run smoother every time.

And in general this is true. Documenting an obscure process that doesn't often run or a setup is complicated to get right doesn't have much as much value as the ability to achieve the desired end without really know how it happened. Sure, sometimes how it happened or how it works matters, such as when a dependency breaks you (but you're protecting yourself from that, right?) or you have to add something. In general though, I've been learning, you've always got more work to do. And reducing some of the toil[2] involved in your work frees you up to actually think at a more synthesizing level. That's the good stuff! 🜚

Computers in general function because we limn their absurd complexity via abstractions. Do you need to know about pipelines and interrupts and L2 caches to write CSS? No! That'd be a terrible waste of time. Do you need to know what to do to lock down a server, if that's your job. Yes! But make your job a little easier by adding in automation, and you'll start to operate strategically rather than always being caught up in the minutia.

So here's an exhortation: work hard to be lazy. Future you will be glad.


  1. basically you can keep running the steps as you're trying to get the whole thing to work and nothing break. Good for fiddlers… ↩︎

  2. https://landing.google.com/sre/book/chapters/eliminating-toil.html ↩︎