Comment by mewpmewp2
Personally I think the successful automation ending would've been "too satisfactory and perfect". Real life is not that.
I felt the video was relatable to me in many frequent situations where I would like to automatize some sort of process, but I end up being overly enthusiastic, because I'm so biased towards automating routineus activities. I frequently end up spending more time automating something than it would've been worth to just do it couple of times. I end up creating PoCs that don't entirely work in situations like those and the demos can fail, because I got bored at some point and didn't check thoroughly enough, so this video made me cringe at myself.
I do feel however that this was just one layer of this video, and the other layer is about the other side where manual processes are preferred for irrational reasons over automation, so I think this video accurately reflected all sides of the story. Automating vs doing something routineusly is a balancing and tolerance act.
I have fought this battle for my entire career. The problem is, incentives are tough - often senior engineers on ops teams have built the system on duct tape and bubble gum, and have a very specific and unobtainable knowledge about all the quirks of it which frequently involve a ton of manual, esoteric shit to resolve without a lot of know-how.
Part of how I’ve made myself useful as a consultant sometimes in my career is by insisting on a strict set of processes that break down and identify bottlenecks like this. It’s almost always ops, it almost always can be automated, but there’s “no bandwidth or budget” to fix the problems so they expect superheroes to constantly save the day by pulling absurd hours or by inventing genius solutions that management buys into.
For me, as an IC or sometimes as a consultant, I try to chip away. We spend X minutes a week doing Y thing manually. what’s the easiest thing we can do to reduce X? and then slowly introduce whatever arises out of those processes, and then slowly “eat the elephant” one bite at a time.
when people come into a big messy manual legacy system they often, imho, make the mistake of wanting to tear it down and recreate it from scratch. having done several cloud architecture/IAC refactors in my career, I can tell you its definitely a way easier sell to management and other teams to whittle away at it than to recreate it entirely, and i think that’s also what the video is poking fun at. There’s always skeletons that your perfect solution will run into. Processes, standards, and monitoring will expose these dead bodies, but it’s not a one shot solution and I’m immediately skeptical when anyone proposes something in this space that is. Every situation is a snowflake.