IMHO, Prometheus is not complicated. In fact, it was designed to be simple and dumb. I mean, it does not provide any production grade scalability out of box, an

IMHO, Prometheus is not complicated. In fact, it was designed to be simple and dumb. I mean, it does not provide any production grade scalability out of box, and people have to implement their own add-ons and external storage solutions (thanos, cortex, M3DB...) to make Prometheus even usable in big clusters. Just because you can buy DataDog and be done doesn't mean the paid product itself is easy to understand. What I like most about Prometheus, are PromQL, pull-based model, and its open source ecosystem. I use it both at work and in my homelab (which is just a bunch of Intel NUC bare metal boxes), and it has been great. I can't speak for any other individuals or companies, but there are tons of learning materials and active community where you can get help. Also, when it comes down to application level metrics, you have to educate your developers to add instrumentation in their own code anyways. And I think exposing a small HTTP server is far easier than pushing data to a remote server.