Anybody has some experiences about druid metrics?

I had produced druid metrics data in a kafka topic, but there are too many monitor data, such as jvm/mem/used, segment/used and so on.
I would monitor the druid cluster if it is healthy, such as whether there is node down, which node full gc frequently, which queries time out or took a long time and so.

So which metrics we should choose or create some dashboards? Anybody has some experiences about this? please give me some examples and guides? Thank you so much.