Hey folks! I'm looking into recommendations or insights into how you deal with the automatic cluster scaling in your prod environment. Particularly, how to scale query and data nodes as there are some dependency on each other from the configuration perspective, like for example:

druid.broker.http.numConnections

on broker and

druid.server.http.numThread

on historical . For proper scaling when we see the broker is a bottleneck, it looks like we need to update historical first and only then broker (config + vertically or horizontally scale), which means it will be hard to do it automatically. (We are hosting them on ec2)