@marcingrzejszczak @spencergibb @ryanjbaxter help~

@marcingrzejszczak @spencergibb @ryanjbaxter help~

This had been a really good improvement. It can today take ages to fetch all config from ZooKeeper if there are lot of keys stored, even if they are only on one level. The need to attempt to fetch children for every node becomes really expensive if you have couple of hundreds of keys stored.

Another approach could be to fetch the stats for each path together with the data and check if the node has any children
and only attempt to fetch if that is true. That reduces the number of calls to the ZooKeeper cluster.

@marcingrzejszczak @spencergibb @ryanjbaxter CC