currently , the zookeeper holds loaded segments metadata ,when a server holds too many segments, it’s possible that the correspond zookeeper znode will take too much storage size to hold too much metadata. I think this will be a potential performance bug that will cause low read latency of zookeeper. is there a plan to solve this ?
yes, see this - https://github.com/druid-io/druid/pull/3902
however, there is no fixed timeline as of now when zookeeper will be completely replaced by this. This will be an opt-in feature in a recent future release for experimentation purposes on real clusters.