I created [misterio](https://github.com/daitangio/misterio) to exactly solve the same issue described in the article.

So yes, if you have a simple setup you can use a "web" of docker servers.

But if your needs are just a bit more complex, or you have a mix of different technology (rubyonrails, python flask, java spring boot) and you want to standardize the communication, the security, the performance tracking and so on, K8s is the way.

The problem is the microservice architecture: the complexity is pushed to the infrastructure by the microservices, and K8s is a super-generic way to solve that complexity.

It is like unpacking (in a exploding-kitten mode) an application server on the your Intranet.

Nice, that is my current approach on my latest project (referring to what your linked repo does). Inlining it here for others: copy a .env and a docker-compose.yml, run `docker compose "$@"` on the instance via ssh. That lets me deploy, run one-off commands, and tail logs easily.