Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

It's tough to say. The last company I worked for (a self-described "unicorn") literally used one AWS account for everything: dev, test, and prod.


On the other hand, the company I’m working for now has over 50 accounts for different team envs and it’s a nightmare at times.

Terraform provider and workspaces don’t “scale” well when you need to juggle all those roles, for example.


Same in the previous company I worked for. All in one account, spending millions of dolards on it, terrible architecture. From the top of my head: Unecessary peerings everywhere, bad subnet configuration that provoked network conflicts in certain cases, no autoscaling, IAM access based exclusively on keys, no VPN etc And if that wasn't enough, everything was "done" with one of the most terrible examples of "automation" I've ever seen in my whole life: A huge Terraform mono repo with hundreds of files with hundreds of duplicated resource calls (because they didn't even use modules). It took them at least a week for just adding a simple IAM user without breaking anything else and the CD would take hours to finish. An absolute nightmare. As no one else in the company had experience with the cloud, the team managing all I mention was threated as some kind of supreme elite as everyone else though they were doing a great job. Beyond me.


That is a very bad idea.


I know! They set it up like that years ago. You’d think they’d at least have moved dev out of it.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: