My resume has the (dis)pleasure of converting all of Netflixes OSS projects to run in containers. JVM or not. I have golang services that run at 80Mb, I have rust services that run at 20Mb. Every JVM service? I have to literally define its memory consumption with command line flags (-jvmx? I forget, its my weekend) on each deployment. Running JVMs in k8s is hilarious. I'm not grey beard Linux level yet (soon) but my god. It's hilarious how poorly they perform.
Agree. I maybe wrong but I think it was Netflix first which normalized generating one web service call generating 1MB worth of logs. In our place finance is kind of whopping ass of teams which are generating extensive logging and monitoring as Splunk cost goes through roof.
