Hacker News Reader

Comment by UmbertoNoEco | Hacker News Reader

What? Dont you like to learn 3/4 almost-useless DSLs, have to modify 3 config files for the smallest change and plumb more breaking things than Mario?

View on HN

Mario doesn't do any plumbing.

It’s-a-me, the SRE!

It solves a lot of scaling problems though

> It solves a lot of scaling problems though

Problems that most devs wish that they had but don't.

Infrastructure as Code is powerful; Terraform + k8s (with nginx-ingress/cert-manager/external-dns) and you never need to worry about infrastructure again...

At $work we use(d) the following:

- Heroku

- AWS ECS Fargate

- AWS EC2 (AutoScalingGroups) + Docker

- AWS EKS (managed K8s) + Helm charts

Everything (except Heroku) is/was IaaC.

They are listed in the ascending order of how many problems they cause, and how much we have to worry about them ... I will never ever touch K8s if I can possibly avoid it.

At my last job, heroku was the cause of at least an outage a month. The redis add-on was particularly troublesome, as I recall.

That's not to say the rest aren't bad, of course, but I'd much rather have control in my hands than wondering if the next time it goes down again how many angry phone calls we'll get and if there is even anything we can do about it.

Give me a few bare servers any day, but you take what you can get I guess.

Yeah, until your team starts using PaaS services and your terraform modules are a mess that barely supports a third of the product's configuration, and that process is going to accelerate more and more over time.

Which 99% of users don’t have and never will

So does a beefy server and a nicely parallelized backend service.

Problems that most of Kubernetes users don't have, and could have been solved with some unfashionable 2/3 servers and a load balancer.

Reply on news.ycombinator.com