[sruffatti]

Can someone help me understand why keeping metrics in memory, consuming resources, and having Prometheus ping your spring-boot app every 10 seconds is good design? What happens to this design at scale when you have thousands of spring boot microservices and Prometheus has to ping each instance to collect metrics.

[zbentley]

I have a lot of the same qualms and unhappinesses with the prom ecosystem. Also don't love how it sort of fiats that you have low-latency, quickly-consistent external service discovery to find out what to scrape.

It also seems like the prom ecosystem is still deciding how it wants to do push based metrics (you know...the pattern that has none of those problems) and that none of the tools in that area are especially good or mature.