I have a large Elastic Search cluster that I want to tune better. I have a few different things I want to try, including some different jvm parameters. What might be a good way to test the parameters that provides a valid confirmation if the jvm parameters are better.
I was thinking of bring up a new node that has the new jvm parameters and benchmarking the cpu/gc/speed against the other nodes. Would that be a valid test? I read that for other types of tuning like the Java version, you shouldn't run different versions in the same cluster. Is that true for jvm parameters as well?