> Just realize it's not going to be anything like a giant tech firm, where you don't realize they actually do 99% of the job for you.
This. So much this.
All those teams of devops people, the SREs, the QA people, the network engineers and all the other unsung heroes don't exist in a startup yet the people who only ever worked at BigCo assume that they do and so forget about all of the problems involved in building a codebase that can scale in a meaningful way because it's always been someone else's job and they have been focused solely on their KPI or singular deliverable.
All that build infrastructure or coding standards or documentation or monitoring systems that you took for granted at Google or Facebook? None of it exists. Not only that - you won't be rewarded for building it, instead you'll be asked why the next person to come behind you got their work done faster than you did.
I almost got burned by this. The COO of the company basically asked my why I wasn't getting anything done. I had to be direct and say that I'm happy to spend 100% of my time coding new features, but he is going to have to answer the phone when a customer says the site is down, because I will be too busy coding.
I got burned by this. I worked at a Google spinoff as the first non-Google backend engineer. The documentation was atrocious and so was the onboarding process so I took the time to fix the problems for the next new hire.
I was then told the new hire got up to speed faster than I had (no shit) and that this reflected negatively upon me.
at least there is a customer. you need to prioritize in a startup and your COO is telling you that you have the wrong priority. it’s not useful to have a reliable monitored system that no one uses.
> All that build infrastructure or coding standards or documentation or monitoring systems that you took for granted at Google or Facebook? None of it exists. Not only that - you won't be rewarded for building it, instead you'll be asked why the next person to come behind you got their work done faster than you did.
Not necessarily like that, but yes, almost. For instance, if someone sat down and spent a month writing documentation for shit we might pivot from in two, I must fire myself for having let it happen. Just wasting money and (perhaps more importantly) time.
I've definitely seen what you're saying happen, though. Lots of FAANG engineers not experienced in how things come to be.
- "Why don't we have blue/green deploys?"
- "Great idea. What will it help us with and how can we make it happen?"
That's likely to happen. Someone else speccing out a deploy system for you with canaries or b/g or whatever so you can go write some code isn't going to. You're hopefully independent enough to make the case, build the thing fast enough that it's worth it to the business, and have it running.
It can get a bit cumbersome. Especially for people coming fresh into an agency environment. You are going to have to learn a dozen different build pipelines and tech stacks by the time you are done at the agency. We are absolutely aware of blue green deploys or whatever it may be, in fact client X had it, they are gone now though and our most recent inherited code base uses cloud formation to spin up feature specific environments instead, have fun learning that and then never using it again.
We work hard to normalize tech stacks and deploys as they come in, but it is always a cost/benefit tradeoff and very often "leave it alone" is the clients choice.
I agree with most of this, but services like CircleCI, Rollbar, and (now) Github Actions are just as good as the build infrastructure at many large companies, and they are free at low usage
That may be true, but they don't configure themselves and adapt over time as new demands or bugs arise.
And sure, doing so may be "easy" but it's yet another context to switch in and out of. The difficulty and time required also increases as the scale of the company, codebase, and infra grows.
I used to work for small companies where I was very hands-on with all of that infra/devops sort of stuff in addition to app/systems dev, but now at a not-quite-FAANG-but-you've-heard-of-them company realize how much I don't actually know is going on in our infra because I don't have to. In some ways I appreciate it, in other ways I feel disconnected.
This. So much this.
All those teams of devops people, the SREs, the QA people, the network engineers and all the other unsung heroes don't exist in a startup yet the people who only ever worked at BigCo assume that they do and so forget about all of the problems involved in building a codebase that can scale in a meaningful way because it's always been someone else's job and they have been focused solely on their KPI or singular deliverable.
All that build infrastructure or coding standards or documentation or monitoring systems that you took for granted at Google or Facebook? None of it exists. Not only that - you won't be rewarded for building it, instead you'll be asked why the next person to come behind you got their work done faster than you did.