This software is great for cases where the business need demands maximum usage of available hardware resources (turn all your compute into one pool, because your company is too poor to accept that not every CPU needs to be maximized 24/7) and also provides a foundation for high availability, automatic failover, etc.
Most businesses do not need this. They don't need nine 9's of uptime. Neither do you. You could absolutely survive with short monthly or quarterly maintenance windows (except when a security update demands immediate attention).
Is having your "infrastructure as code" useful? Absolutely!
Does 100% of it need to be IaC? No.
Does the complexity of having 100% IaC and reproducible make it worth it? Not even close.
Unless you want your ops team to be disposable. If it's all IaC, you can replace everyone with juniors who only know e.g., Terraform. They don't have to have any real skills beyond that.
@sun@feld Personally I'm having the most trouble with what we're trying to do with k8s. It is unbelievably opaque. You control the thing by writing magic YAML files, and there is basically no way to know what critical variable to set, because there are literally tens of thousands of them. The doc team is absolutely overwhelmed. If anything goes wrong, the only way to fix it is to find a guy who knows a guy who either hacked on something related or saw it resolved and wrote it down. The k8s themselves seem almost benign in comparison.
@sun@feld My answer was to resign and retreat into basic infrastructure like filesystems. They are always necessary and that will butter my toast for a few years until I die. Also, I founded a manufacturing company.
I was concerned that I was becoming obsolete by not following the latest on Kubernetes and then I realized that it was futile. I don't know squat about blockchain or LLM technologies either, and I feel fine.